Psychosocial Factors and Workers’ Health and Safetydownloads.hindawi.com/journals/specialissues/909274.pdfelevated basal systolic blood pressure, reduced basal heart rate variability,

BioMed Research International

Psychosocial Factors and Workers’ Health and Safety

Guest Editors: Sergio Iavicoli, Giancarlo Cesana, Maureen Dollard, Stavroula Leka, and Steven L. Sauter

Psychosocial Factors and Workers’

Health and Safety

BioMed Research International

Psychosocial Factors and Workers’

Health and Safety

Guest Editors: Sergio Iavicoli, Giancarlo Cesana,

Maureen Dollard, Stavroula Leka, and Steven L. Sauter

Copyright © 2015 Hindawi Publishing Corporation. All rights reserved.

�is is a special issue published in “BioMed Research International.” All articles are open access articles distributed under the CreativeCommons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the originalwork is properly cited.

Contents

Psychosocial Factors andWorkers’ Health and Safety, Sergio Iavicoli, Giancarlo Cesana, Maureen Dollard,Stavroula Leka, and Steven L. SauterVolume 2015, Article ID 628749, 3 pages

Verbal Aggression from Care Recipients as a Risk Factor among Nursing Sta�: A Study on Burnout in

the JD-R Model Perspective, Sara Viotti, Silvia Gilardi, Chiara Guglielmetti, and Daniela ConversoVolume 2015, Article ID 215267, 17 pages

�eRelationship of On-Call Work with Fatigue, Work-Home Interference, and Perceived Performance

Di�culties, Carla M. Ziebertz, Madelon L. M. van Hoo�, Debby G. J. Beckers, Wendela E. Hoo�man,Michiel A. J. Kompier, and Sabine A. E. GeurtsVolume 2015, Article ID 643413, 10 pages

Psychosocial Work Factors and Musculoskeletal Pain: A Cross-Sectional Study among Swedish Flight

Baggage Handlers, Eva L. Bergsten, S. E. Mathiassen, and E. VingårdVolume 2015, Article ID 798042, 11 pages

�eAssociation between Job Strain and Atrial Fibrillation: Results from the Swedish WOLF Study,Eleonor I. Fransson, Magdalena Stadin, Maria Nordin, Dan Malm, Anders Knutsson, Lars Alfredsson,and Peter J. M. WesterholmVolume 2015, Article ID 371905, 7 pages

Burnout Is Associated with Reduced Parasympathetic Activity and Reduced HPA Axis Responsiveness,

Predominantly in Males, Wieke de Vente, Jan G. C. van Amsterdam, Miranda Ol�, Jan H. Kamphuis,and Paul M. G. EmmelkampVolume 2015, Article ID 431725, 13 pages

Job Strain and Self-Reported Insomnia Symptoms among Nurses: What about the In�uence of

Emotional Demands and Social Support?, Luciana Fernandes Portela, Caroline Kröning Luna,Lúcia Rotenberg, Aline Silva-Costa, Susanna Toivanen, Tania Araújo, and Rosane Härter GriepVolume 2015, Article ID 820610, 8 pages

Consequences of Job Insecurity on the Psychological and Physical Health of Greek Civil Servants,Dimitra Nella, Efharis Panagopoulou, Nikiforos Galanis, Anthony Montgomery, and Alexis BenosVolume 2015, Article ID 673623, 8 pages

Towards a Job Demands-Resources Health Model: Empirical Testing with Generalizable Indicators of

Job Demands, Job Resources, and Comprehensive Health Outcomes, Rebecca Brauchli, Gregor J. Jenny,Désirée Füllemann, and Georg F. BauerVolume 2015, Article ID 959621, 12 pages

�eContext, Process, and Outcome EvaluationModel for Organisational Health Interventions,Annemarie Fridrich, Gregor J. Jenny, and Georg F. BauerVolume 2015, Article ID 414832, 12 pages

Associations between Distal Upper Extremity Job Physical Factors and Psychosocial Measures in a

Pooled Study, Matthew S.�iese, Kurt T. Hegmann, Jay Kapellusch, Andrew Merryweather, Stephen Bao,Barbara Silverstein, and Arun GargVolume 2015, Article ID 643192, 9 pages

Models of Workplace Incivility: �e Relationships to Instigated Incivility and Negative Outcomes,Kristo�er Holm, Eva Torkelson, and Martin BäckströmVolume 2015, Article ID 920239, 10 pages

E�ects of aWorkplace Intervention Targeting Psychosocial Risk Factors on Safety andHealth Outcomes,Leslie B. Hammer, Donald M. Truxillo, Todd Bodner, Jennifer Rineer, Amy C. Pytlovany, and Amy RichmanVolume 2015, Article ID 836967, 12 pages

An Evaluation of the Policy Context on Psychosocial Risks andMental Health in theWorkplace in the

European Union: Achievements, Challenges, and the Future, Stavroula Leka, Aditya Jain, Sergio Iavicoli,and Cristina Di TeccoVolume 2015, Article ID 213089, 18 pages

Workplace Bullying as a Risk Factor for Musculoskeletal Disorders: �eMediating Role of Job-Related

Psychological Strain, Michela Vignoli, Dina Guglielmi, Cristian Balducci, and Roberta Bon�glioliVolume 2015, Article ID 712642, 8 pages

Estimating the Impact of Workplace Bullying: Humanistic and Economic Burden amongWorkers with

Chronic Medical Conditions, A. Fattori, L. Neri, E. Aguglia, A. Bellomo, A. Bisogno, D. Camerino,B. Carpiniello, A. Cassin, G. Costa, P. De Fazio, G. Di Sciascio, G. Favaretto, C. Fraticelli, R. Giannelli,S. Leone, T. Maniscalco, C. Marchesi, M. Mauri, C. Mencacci, G. Polselli, R. Quartesan, F. Risso, A. Sciaretta,M. Vaggi, S. Vender, and U. VioraVolume 2015, Article ID 708908, 12 pages

Prognostic Factors of Returning to Work a�er Sick Leave due to Work-Related CommonMental

Disorders: A One- and�ree-Year Follow-Up Study, Bo Netterstrøm, Nanna Hurwitz Eller,and Marianne BorritzVolume 2015, Article ID 596572, 7 pages

Do Italian Companies ManageWork-Related Stress E�ectively? A Process Evaluation in Implementing

the INAIL Methodology, Cristina Di Tecco, Matteo Ronchetti, Monica Ghelli, Simone Russo,Benedetta Persechino, and Sergio IavicoliVolume 2015, Article ID 197156, 10 pages

EditorialPsychosocial Factors and Workers’ Health and Safety

Sergio Iavicoli,1 Giancarlo Cesana,2 Maureen Dollard,3

Stavroula Leka,4 and Steven L. Sauter5

1Department of Occupational and Environmental Medicine Epidemiology and Hygiene, INAIL, Monte Porzio Catone,00040 Rome, Italy2Research Center for Public Health, University of Milano-Bicocca, 20126 Milan, Italy3Asia Pacific Centre for Work Health and Safety, A World Health Organization Collaborating Centre in Occupational Health,University of South Australia, Adelaide, SA CA1-05, Australia4Centre for Organizational Health & Development, A World Health Organization Collaborating Centre in Occupational Health,University of Nottingham, Nottingham NG7 2RD, UK5Northern Kentucky University, Highland Heights, KY 41099, USA

Correspondence should be addressed to Sergio Iavicoli; [email protected]

Received 30 September 2015; Accepted 5 October 2015

Copyright © 2015 Sergio Iavicoli et al. This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Over the last decades significant developments in the eco-nomic, political, technological, and social landscape havecontributed to changes in the nature of work and the way bywhich people work. Moreover, significant demographic andsocial changes have had an impact on working conditionscontributing to the emergence of new risks for health atwork. In this scenario, psychosocial risks have attracted theattention of occupational safety and health researchers, policymakers, and practitioners. Work-related psychosocial risksemerge from the design, content, andmanagement of work aswell as its social context that can have a hazardous influenceon employees’ health.They are considered as a contemporarychallenge for health due to their close link with stress at work.There is evidence about the detrimental impact of work-related stress on workers’ health and safety, particularly inrelation to cardiovascular diseases andmental,musculoskele-tal, and chronic degenerative disorders. Consequently, theseissues are the primary focus of the current special issue.

Following a peer review process involving a broadgroup of international experts, out of over 60 submissionsreceived, 17 contributions were accepted in this special issue.The papers selected represent a good collection of originalresearch and review articles, with a wide geographical distri-bution. The contributions focus on the following: (a) workand psychosocial risks in the field of occupational healthand safety, exploring the impact of psychosocial hazards in

terms of workers’ health, well-being, and performance and(b) policy as well as company level interventions. Thus, all ofthem provide new evidence-based insights in occupationalhealth and well-being. A brief summary of each paper ispresented below.

A review article “An Evaluation of the Policy Contexton Psychosocial Risks and Mental Health in the Workplacein the European Union: Achievements, Challenges, and theFuture” by S. Leka et al. offers a review of hard and soft lawpolicies in the European Union in relation to mental healthand psychosocial risks in the workplace, to identify strengths,weaknesses, and gaps to be addressed in the future. Ninety-four policies included in the review revealed several gaps,especially in relation to binding in comparison to nonbindingpolicies, and recommendations are offered for future actionsin this area.

The paper “Burnout Is Associated with Reduced Para-sympathetic Activity andReducedHPAAxis Responsiveness,Predominantly in Males” by W. de Vente et al. showsthe presence of a dysregulation of the sympathetic-vagalbalance and the HPA axis in burnout, as indicated byelevated basal systolic blood pressure, reduced basal heartrate variability, and a trend for elevated cardiac output inthe burnout group as compared to the healthy referencegroup. Gender differences in cardiovascular functioningand in cortisol reactivity to a psychosocial stressor and in

Hindawi Publishing CorporationBioMed Research InternationalVolume 2015, Article ID 628749, 3 pageshttp://dx.doi.org/10.1155/2015/628749

http://dx.doi.org/10.1155/2015/628749

2 BioMed Research International

basal alpha-amylase in the context of burnout were alsofound.

In their paper “Towards a JobDemands-ResourcesHealthModel: Empirical Testing with Generalizable Indicators ofJob Demands, Job Resources, and Comprehensive HealthOutcomes” R. Brauchli et al. expand the logic of the origi-nal job demands-resources model from the original healthimpairment/motivational processes to simultaneously study-ing and improving pathogenic and salutogenic health devel-opment processes at work. The paper offers evidence on theapplicability of this model in diverse economic sectors andprofessional groups and its usefulness for population-basedpublic health interventions in the general working popu-lation.

A paper by D. Nella et al. entitled “Consequences ofJob Insecurity on the Psychological and Physical Healthof Greek Civil Servants” provided an estimation of short-term consequences of job insecurity associated with a newlyintroducedmobility framework inGreece in terms of anxiety,depression, and psychosomatic and musculoskeletal symp-toms. Their findings showed immediate detrimental effectsof job insecurity on the physical, psychological, and socialfunctioning of employees.

A study reported by C. M. Ziebertz et al. in “TheRelationship of On-Call Work with Fatigue, Work-HomeInterference, and Perceived PerformanceDifficulties” focuseson the effects of the offsite on-call duties on employees’recovery from work. According to the effort-recovery model,a long lasting situation of incomplete recovery from loadeffects is critical for workers’ health and well-being. Althoughthe variation in the amount of exposure to on-call work wasnot systematically related to a lack of recovery from work,the experience of being on-call was related to fatigue, strain-based and time-based work home interference, and on-callperformance difficulties.

L. F. Portela et al. explored the effect of perceived stresson insomnia symptoms, in a large sample of nurses, in theirpaper “Job Strain and Self-Reported Insomnia Symptomsamong Nurses: What about the Influence of EmotionalDemands and Social Support?” Given the high emotionaldemands of the nursing profession, which requires caringpersonal service, the role of social support in relation towork-related sleep disturbance was particularly confirmed for theemotional demand control model.

A challenge for nursing staff is exposure to verbal aggres-sion. The paper “Verbal Aggression from Care Recipients asa Risk Factor among Nursing Staff: A Study on Burnout inthe JD-R Model Perspective” by S. Viotti et al. is a cross-sec-tional study that examines the association between verbalaggression and burnout also considering the role of job con-tent, social resources, and organizational resources in reduc-ing the negative impact of verbal aggression. Authors foundan association between verbal aggression and burnout thatwas facilitated by the job content level resources (e.g., jobautonomy, role clarity, and skill discretion). It provides aninteresting comparison between general nurses and nurses’aides highlighting the role of different resources in protectingnursing staff from the detrimental effects of verbal aggressionon health.

In the paper “Models of Workplace Incivility: The Rela-tionships to Instigated Incivility and Negative Outcomes” K.Holm et al. investigated workplace incivility as a social pro-cess. Different components of work incivility (experienced,witnessed, and instigated incivility) were examined also inrelation to negative outcomes of workplace incivility. Wit-nessing coworker incivility emerged as the most importantdimension to explain instigated incivility. Moreover, giventhe moderating role of support, organizational factors wereidentified as a key component to be included in future studiesin this field.

Linking psychosocial risk exposure to productivity isimportant to draw the attention of managers and policymakers. A. Fattori et al. in their paper “Estimating theImpact of Workplace Bullying: Humanistic and EconomicBurden among Workers with Chronic Medical Conditions”demonstrate the negative impact of workplace bullying onquality of life and productivity among workers with commonand severe chronic diseases. Particularly, authors found asignificant association between workplace bullying and allcomponents of productivity loss as well as an associationwithworse health-related quality of life in comparison with otherconcurrent medical conditions.

The previously underresearched area of the link betweenworkplace bullying and physical health problems is tackedin the paper “Workplace Bullying as a Risk Factor for Mus-culoskeletal Disorders: the Mediating Role of Job-RelatedPsychological Strain” by M. Vignoli et al. The researchersshowed the mediating role of work-related strain in therelationship between bullying and musculoskeletal disordersof the low back, upper back, and neck, but not the shoulders.The strain process emerged as one of the elements to considerin understanding the detrimental effect of bullying on thevictims’ health, even though bullying remained a significantrisk factor for musculoskeletal disorders.

A novel investigation by E. L. Bergsten et al. in thepaper “Psychosocial Work Factors andMusculoskeletal Pain:a Cross-Sectional Study among Swedish Flight BaggageHandlers” offers an investigation of the relationship betweenpsychosocial exposures and musculoskeletal health amongflight baggage handlers. Findings showed an associationbetween severity of pain and pain interfering with work andpsychosocial factors at work (work organization, job content,interpersonal relationships, and leadership). Findings suggestthe inclusion of the psychosocial work environment as arelevant target for interventions in this occupation.

The paper “Associations between Distal Upper ExtremityJob Physical Factors and Psychosocial Measures in a PooledStudy” by M. S. Thiese et al. provides an exploratory analysison the relationship between quantified job physical expo-sure and psychosocial outcomes in a large sample. Multipleassociations between physical exposure and occupationaland nonoccupational psychosocial factors were found afteradjustment for age, body mass index, and gender. Moreoverthe study provides a quantification of this association includ-ing the effect on occupational injuries and illness.

The paper “TheAssociation between Job Strain andAtrialFibrillation: Results from the Swedish WOLF Study” by E. I.Fransson et al. provides additional knowledge about different

BioMed Research International 3

risk factors related to work stress and atrial fibrillation (AF)through a two-time study.The association between job strainand AF was found to be time-dependent, since long-termexposure to job strain was more strongly associated with AFrisk than shorter exposure.

Getting people back to work after sick leave for mentalhealth problems is crucially important. In the paper “Prog-nostic Factors of Returning to Work after Sick Leave dueto Work-Related Common Mental Disorders: A One- andThree-Year Follow-Up Study” B. Netterstrøm et al. assess theprognostic factors of return to work after one year and threeyears among workers after sick leave due to occupationalstress. While the role of psychosocial factors in predictingreturn to work disappears over time, the severity of thedisorder (full time sick leave and self-rated work ability) wasfound to be a crucial predictor in the long run.

Providing a new framework for evaluating organizationalhealth interventions in their paper “The Context, Process,and Outcome Evaluation Model for Organisational HealthInterventions” A. Fridrich et al. proposed the CPO model asa basis for a structured evaluation of combined occupationalhealth interventions. Findings support the effectiveness ofa CPO evaluation model as a shared mental model forthe complex intervention evaluation process in the fieldof occupational health. The use of shared terminologiescan facilitate the development of a common language forimproving the comparability of evaluation study results.

An interesting preliminary study, “Effects of a WorkplaceIntervention Targeting Psychosocial Risk Factors on Safetyand Health Outcomes” by L. B. Hammer et al., offers a firstlook at the effectiveness of a workplace intervention targetingwork-life stress and safety-related psychosocial factors onhealth and safety outcomes. The study gives evidence of theneed for focusing interventions on support training and teameffectiveness for planning and problem solving to improveworkers’ health.

In the paper “Do Italian Companies Manage Work-Related Stress Effectively? A Process Evaluation in Imple-menting the INAIL Methodology” C. Di Tecco et al. offera process evaluation on interventions to assess and managerisks related to work-related stress, using a methodologicalpath proposed by INAIL. Findings highlight that key aspectsof process and contentmay be considered as recurrent factorswhich might account for the differences in the results duringthe assessment phases and in the perception of the usefulnessof the method.

Sergio IavicoliGiancarlo CesanaMaureen DollardStavroula LekaSteven L. Sauter

Research ArticleVerbal Aggression from Care Recipients asa Risk Factor among Nursing Staff: A Study onBurnout in the JD-R Model Perspective

Sara Viotti,1 Silvia Gilardi,2 Chiara Guglielmetti,3 and Daniela Converso1

1Dipartimento di Psicologia, Universita degli Studi di Torino, Via Verdi 8, 10124 Torino, Italy2Dipartimento di Scienze Sociali e Politiche, Universita degli Studi di Milano, Via del Conservatorio 7, 20122 Milano, Italy3Dipartimento di Economia, Management e Metodi Quantitativi, Universita degli Studi di Milano,Via Conservatorio 7, 20122 Milano, Italy

Correspondence should be addressed to Sara Viotti; [email protected]

Received 9 February 2015; Revised 7 April 2015; Accepted 4 May 2015

Academic Editor: Stavroula Leka

Copyright © 2015 Sara Viotti et al. This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Among nursing staff, the risk of experiencing violence, especially verbal aggression, is particularly relevant. The present study,developed in the theoretical framework of the Job Demands-Resources model (JD-R), has two main aims: (a) to examine theassociation between verbal aggression and job burnout in both nurses and nurse’s aides and (b) to assess whether job content,social resources, and organizational resources lessen the negative impact of verbal aggression on burnout in the two professionalgroups. The cross-sectional study uses a dataset that consists of 630 workers (522 nurses and 108 nurse’s aides) employed inemergency andmedical units. High associationswere found between verbal aggression and job burnout in both professional groups.Moderated hierarchical regressions showed that, among nurses, only the job content level resources moderated the effects of theverbal aggression on job burnout. Among nurse’s aides, the opposite was found. Some resources on the social and organizationallevels but none of the job content level resources buffered the effects of verbal aggression on workers burnout. The study highlightsthe crucial role of different types of resources in protecting nursing staff from the detrimental effects of verbal aggression on jobburnout.

1. Introduction

In the workplace, nursing staff are exposed to various factorsthat are likely to jeopardize their health and safety. Amongthese, the risk of experiencing violence is particularly rel-evant. Work-related violence includes both physically andpsychologically violent incidents in which staff members areabused, threatened, or assaulted. It can be defined as “anythreat, physical, and/or psychological, that is directed towarda person while at work” [1].

More specifically, in the health care sector, the mostcommon violence is the so-called Type II category, describedas the following in the Californian Occupational Safety andHealth Administration classification [2, 3]: events involvingaggressions by someonewho is either the recipient of a serviceprovided by the affected workplace or the victim.

Europe is recently witnessing a progressive increase ofType II violence, which is considered an “emerging epidemic”[4, 5]. In a study across 10 European countries, Camerino etal. [6] found that 9.9% of nurses face violence from patientsor patients’ relatives at least once a week (countries over theEuropean average: France, 19.5%; UK, 12.3%; Germany, 11.5%;and Italy, 10.3%). This violence mainly consists of verbalaggression, including loud and demanding verbal hostility orverbal threats of the intent to do harm [7, 8].

Because a higher level of violence is expected in thoseunits where patients may initiate more verbal or physicalthreats (i.e., psychiatric wards or elderly patient areas) orwhere emergencies and workload are massive (i.e., out-of-hours primary care, emergency, and ICU units), most of theresearch has been conducted in these specific contexts [9–13].The existing literature mainly highlights the negative effects


http://dx.doi.org/10.1155/2015/215267


of violence exposure in terms of physical and psychologicalhealth, and it rarely investigates the protective factors or thepositive resources that workers possess/adopt to buffer them[14–16].

The present study focused on verbal aggression, whichis one of the most common forms of Type II violenceincurred by nursing staffs. Using the framework of the JobDemands-Resources model (JD-R [17–19]), the study intendsto investigate the relationship between verbal aggression andburnout among two categories within the nursing profession,namely nurses and nurse’s aides. The study also examineswhether and which different kinds of job resources are ableto buffer the impact of verbal aggression on burnout amongthe two subsamples considered.

Verbal aggression is a formof direct psychological aggres-sion that includes yelling at the service provider or makingsarcastic or offensive remarks [20]. According to the JD-Rmodel [17–19], verbal aggression can be considered a jobdemand because it is a psychological aspect of the job thatrequires sustained psychological effort and is therefore asso-ciated with certain physiological and/or psychological costs.The present paper focuses on a specific psychological cost,namely, job burnout, which, as the literature has extensivelyhighlighted, represents a particularly relevant concern withinthe nursing profession [21, 22]. According to Green et al. [23],burnout is a syndrome recognizable by two core dimensions:emotional exhaustion, which refers to the depletion of theenergy process, and depersonalization, which indicates ahighly detached attitude toward patients.

The fact that being exposed to verbal aggression may leadto burnout is suggested by one of the main assumptions ofthe JD-R model [17–19], the health impairment hypothesis.In accordance with this assumption and the Consarvationof Resources theory (COR) [24], verbal aggression maydeplete workers’ energy, activating a loss cycle that can leadto exhausting employees’ mental and physical resources. Inaddition, as a consequence of perceiving contact with thepatient as a threat, the workers may adopt an attitude ofavoidance, such as depersonalization. From an empiricalpoint of view, many studies carried out both within thecustomer service workers’ population [25–28] and withinthe health sector workers’ population, in particular [29–34],confirmed the positive association between verbal aggressionand burnout. Based on that, in the present study, it is expectedto find a significant and positive relationship between verbalaggression and respectively emotional exhaustion and deper-sonalization among nurses (H1a) and nurse’s aides (H1b).

The buffering assumption of the JD-R model [17–19]states that job resources may buffer the impact of verbalaggression on job burnout. Job resources refer to thosephysical, psychological, social, or organizational aspects thathelp achieve work goals, reduce job demands, and lessen theassociated physiological and psychological costs. As statedabove, according to the COR theory [24], verbal aggressionis generally perceived to be losses because meeting suchdemands requires the investment of valued resources, whichare viewed as gains [35]. By contrast, the presence of resourcesin the workplace may interrupt the loss cycle and lead toboosting the motivational process by sustaining the workers

in successfully coping with job demands [19, 34]. In thisperspective, it is important to understandwhich resources areuseful for dealing with verbal aggression and moderating thedevelopment of burnout symptoms.

However, whereas the research is well-developed formostjob demands and provides evidence in that direction, asregards verbal aggression, the attention on the variables ofthe workplace that may buffer its detrimental effects is quitelimited [19]. Particularly, the research needs to be expandedin the direction of examining and comparing the roles ofdifferent kinds of job resources in buffering the adverse effectof verbal aggression. According to the literature, three typesof resources may be available in the workplace: job contentresources, social resources, and organizational resources[19]. Rarely in the literature there are studies available thattake into consideration, all together, resources from thesethree levels to test their buffering effects and comparetheir function in a unique sample. Studies examining all ofthese resources may advance the literature by indicating thelevel (job content, social, or organizational levels) to whichintervention would be most appropriate [36].

In that direction, the present study includes eight specificresources at the job content, social, and organizational levels.The choice was driven by previous research that recognizedthe importance of these job characteristics in moderating theeffects of the various job demands, including verbal aggres-sion on job burnout, both among the general population andamong nursing professionals [15, 35, 37–39].

At the job content level, skill discretion, job autonomy,role clarity, and work meaning were taken into account.According to Karasek [40], skill discretion and autonomyexpress the extent to which workers are capable of controllingtheir tasks and general work activities. Skill discretion refersto a person’s opportunity to use specific job skills in the workprocess. Job autonomy refers to the extent to which a personis autonomous in task-related decisions, such as timing andmethod control. Broadly speaking, it is plausible that havingwide margins of discretion may stimulate workers to exer-cise creativity in finding successful strategies for managingaggressive patient behaviour, thus lessening exhaustion anddepersonalization symptoms caused by exposure to verbalaggression. As regards autonomy, some evidence supportsits moderating effect on the relationship between verbalaggression and burnout [15, 38], whereas no studies werefound in the literature regarding skill discretion.

Role clarity refers to the degree to which the task and theobjectives of a job are clearly defined [41]. This job resourcehas been found to work as a moderator on the relationshipbetween several job demands andworkers’ outcomes [42, 43].Even though no studies focus on its role in moderating therelationship between verbal aggression and burnout, it isplausible that role clarity may increase the opportunity toeffectively manage the relationship with patients in severalways. For example, workers may be placed in the position togive adequate feedback to patients.

Work meaning refers to the degree to which the workis perceived meaningful, important, and constructive [41].It may work as a buffer of the perceived verbal aggressionon burnout by leading the workers to consider the episodes


of verbal aggression as learning opportunities for improvingcare service rather than just as adverse events.

At the social level, support from colleagues and supportfrom supervisors were considered. Karasek and Theorell [44]defined social support at work as “overall levels of helpfulsocial interaction available on the job from co-workers andsupervisors” (page 69). Both supervisors and colleagues mayplay a role in buffering the burnout symptoms caused bybeing exposed to patients’ verbal aggression by providingboth instrumental (i.e., helping workers manage the relation-ships with patient/relatives) and affective (i.e., giving affectivesupport and not blaming workers for what happens withpatients) support. This explanation is consistent with thefindings provided by the study fromXanthopoulou et al. [38],which found that social support moderated the detrimentaleffect of patient harassment on both emotional exhaustionand cynicism in a sample of home care nurses.

At the organizational level, organizational support, fair-ness, and social utility of the service were considered in thepresent study. Organizational support refers to the degree towhich the organization values workers’ contributions and theextent it cares about workers’ well-being [45]. In a sampleof Canadian postal workers, Schat and Kelloway [37] foundthat organizational support moderated the effects of violenceand aggression on emotional well-being and physical health.Based on that, the presence of supportive procedures that helpworkers when they are victims of aggressive behaviors mayhelp contain the development of burnout.

According to Maslach and Leiter [46], fairness reflectsorganizational justice and can be defined as the extent towhich the organization has consistent and equitable rulesfor all employees. Even if no study specifically explored themoderating role of fairness between verbal aggression andburnout, the literature suggests that it may matter. Elovainoet al. [47] proposed that fairness matters to people becauseit helps them deal with uncertainty, suggesting that peopleespecially need fair judgments when they are concerned withpotential problems associated with social interdependenceand socially based identity processes.

Social utility of the service refers to the degree to whichworkers perceive that the organization provides useful andhigh-quality services for the community [48]. The literaturefocused poorly on this kind of resource. However, especiallyin sectors such as health care, in which the link with thecommunity is important, it may play a central role. Indeed,the perception that the service provided by the organizationhas a positive return for the community may support theworkers in keeping a positive self-image, even if somepatientsshow disapproval for their job or the service.

According to the buffering assumption [17–19], it isexpected that all the resources considered in this study mod-erate the burnout symptoms among both nurses (H2a) andnurse’s aides (H2b). In particular, the relationship betweenverbal aggression and emotional exhaustion and depersonal-ization is expected to be stronger when job resources are lowrather than when job resources are high.

The literature developed in the framework of the JD-Rmodel regarding the nursing context [17–19], rarely paid spe-cific attention to the various subcategories within the nursing

profession, such as nurses and nurse’s aides, when the effectsof job demands and job resources on psychological healthwere examined. In particular, previous studies, in most cases,chose to merge these two job categories without verifyingthe presence of any difference between them despite the factthat nurses and nurse’s aides, even if they share the sameworkplace, significantly differ in educational background,types of tasks they perform, and position in the hierar-chy. Nurses have specialized, formal, post-basic education,and they perform more complex tasks such as developingand implementing nursing care plans, maintaining medicalrecords, and administering care to patients. By contrast,nurse’s aides have little or no formal training or educationandusually assist nurses by carrying out basic, nonspecializedtasks in the care of patients, such as bathing, feeding, andtransporting patients under the supervision and the directionof a nurse [49].

Empirical evidence also suggests that merging thesegroups may obscure the specificity that each category hasregarding job stress experience. For example, Seago andFaucett [50] and Morgan et al. [51], using the framework ofthe JobDemand-Controlmodel (JDC, [40]), found thatwhilenurses fall into the category of active strain (showing highdemand and high control), nurse’s aides are in the high-straincategory (having high demand and low control). Also Fia-bane et al. [52] found significantly different distributions onthe perception of several work-related psychosocial factorsacross these two job categories. For these reasons, in the beliefthat it may be useful to advance the understanding of thephenomenon of job stress in the nursing context, the analyseswill be performed separately on the nursing and nursing aidessubsamples in the present study to highlight any differencesbetween the two job categories. Due to the exploratory natureof the aim, no expectations can be stated on this point.

The present study may advance the past knowledge onthe buffering role of job resources in the demands-burnoutrelationship because it focuses on some aspects neglectedin the previous literature: (a) it considers a wide range ofresources (i.e., task level, social level, and organizational level)as possible moderators of the relationship between verbalaggression and burnout and (b) it analyses the bufferingmechanism separately within the categories of nurses andnurse’s aides.

2. Method

Data were collected during a multi-centre intervention-research conducted in four hospitals in Northwest Italyin 2012. Hospital administrations evaluated, endorsed, andauthorized the research, allowing researchers to use the datafor scientific purposes. Upon approval, department chiefsand nurse coordinators from each ward were asked forauthorization to administer the questionnaire to the nurses.An additional ethical approval was not required because nomedically invasive diagnostics or procedureswere involved tocause psychological or social discomfort for the participants,nor were the patients the subjects of the data collection.However, the research conforms to the provisions of theDeclaration of Helsinki in 1995 (as revised in Edinburgh


Table 1: Sociodemographic and professional characteristics of nur-ses and nurse’s aides.

Nurses Nurse’s aides𝑛 % 𝑛 %

GenderFemale 429 82.2 87 80.6Male 90 17.2 19 17.6

Age≤40 288 55.2 33 30.6>41 234 44.8 75 69.4

Marital statusMarried/living with partner 297 56.9 64 59.3Single/divorced/widowed 221 42.3 43 39.8

WardEmergency 220 42.1 28 25.9Medicine 302 57.9 80 74.1

Years in the health sector≤15 307 58.8 79 73.1>16 215 41.2 29 26.9

2000), and all ethical guidelines were followed as required forconducting human research, including adherence to the legalrequirements of the study country (Italy).

Participants volunteered for the research and were notasked to sign consent forms, but returning the question-naire implied consent. The cover sheet clearly explainedthe research aim, the voluntary nature of participation, theanonymity of the data, and the elaboration of the findings.

The sample consisted of 630 workers: 522 (82.90%) nursesand 108 (17.10%) nurse’s aides. The majority were women(81.9%, 𝑛 = 516) aged between 21 and 62 years (𝑚 = 37.97, sd= 8.76). 57.30% were married or living with partners, 32.20%were single, .90% were divorced, and .60% were widowed.

The average period during which participants had beenworking in the health-care sector was 13.31 years (sd = 9.02)and ranged from 1 month to 39 years. They were employedin emergency (40.30%) andmedical (59.70%) units. Sociode-mographic and profession details for nurses and nurse’s aidesare reported in Table 1.

The data were obtained by means of a self-reportedquestionnaire that included two sections. The first sectioncollected sociodemographic (gender, age, and marital status)and professional (occupation, units, and years in the healthsector) data. The second section included scales aimed atmeasuring job demand, job resources, and worker outcomes.

2.1. Job Demand. Customer verbal aggression that was mea-sured by the subscale coming from the Customer-RelatedSocial stressors (CSS) inventory was developed by Dormannand Zapf [20]. The subscale consists of four items (e.g., item:“Patients get angry at us even over minor matters.”) andreports aCronbach’s alpha (𝛼) of .92. Responseswere given ona four-point scale with a range between 1 (“strongly disagree”)and 4 (“strongly agree”).

2.2. Job Resources. Three categories of factors referring tothe job content, the social, and the organizational levelswere considered. At the job content level, we included foursubscales:workmeaning (5 items,𝛼 = .761, e.g., item: “Is yourwork meaningful?”), role clarity (3 items, 𝛼 = .72, e.g., item:“Does your work have clear objectives?”), skill discretion (5items, 𝛼 = .61, e.g., item: “My job requires that I learn newthings.”), and job autonomy (3 items, 𝛼 = .82, e.g., item: “Myjob allows me to make a lot of decisions on my own.”). Theformer two were drawn from the Copenhagen PsychosocialQuestionnaire by Kristensen et al. [41], and the latter twowere taken from the Job Content Questionnaire (JCQ [53]).To measure social resources, two subscales of JCQ [53]were employed. They respectively investigate support fromsuperiors (5 items, 𝛼 = .83; e.g., item “My supervisor is helpfulin getting the job done.”) and from peers and colleagues (6items, 𝛼 = .82; e.g., item: “People I work with are competentin doing their jobs.”). Three organizational resources wereincluded in the questionnaire. The Organizational CheckupSystem (OCS [46, 54, 55]) measured fairness (6 items, 𝛼 =.65; e.g., item: “In my organization, job resources are equallydistributed.”). Organizational support is a scale included ina recent revision of the Job Content Questionnaire (JCQ[53, 56]) (4 items, 𝛼 = .80; e.g., item “My organization reallycares about my well-being.”). Social utility of the service is ascale drawn from Multidimensional Organizational HealthQuestionnaire (MOHQ, [48]) and (4 items,𝛼 = .69; e.g., item:“The organization in which I work provides good service forthe community”).

Responses on all subscales were given on a four-pointscale with a range between 1 (“strongly disagree”) and 4(“strongly agree”).

2.3. Outcomes. Job burnout was measured thought two sub-scales from the Italian version of Maslach Burnout Inventory(MBI [57–59]): emotional exhaustion (EE, 9 items, e.g., item:“I feel emotionally drained from my work”) and depersonal-ization (DP, 5 items; e.g., item “I feel I treat some patients asif they were impersonal objects”). Both subscales reported agood internal consistency (𝛼EE = .82; 𝛼DP = .77). Responseswere given on a seven-point scale (ranging from 0 = “never”to 6 = “every day”).

2.4. Control Variables. Gender (0 = male; 1 = female), age,marital status (0 = not living with partner; 1 = living withpartner), job seniority, and type of ward (0 = nonacute careward; 1 = acute care ward) are potential confounders forburnout [57, 58, 60, 61]. In view of that, they were taken intoconsideration as control variables.

Table 2 reports descriptive statistics (means and standarddeviations) and Pearson’s correlations for all subscales con-sidered in the study.

All the analyses were performed using SPSS 21. Moder-ated hierarchical regression analyses were employed to exam-ine the main effect of verbal aggression and of job resourceson job burnout, as well as the moderating (buffering) role ofjob resources on the relationship between verbal aggressionand burnout. For each moderated hierarchical regression


Table2:Descriptiv

estatistic

s(means

andsta

ndarddeviations)a

ndPearson’s

correlations

fora

llsubscalesc

onsid

ered

inthes

tudy.

M(ds)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

(1)V

erbalaggression

1.88(.7

7)1

(2)M

eaning

ofthew

ork

3.38

(.48)

−.11∗∗

1(3)R

olec

larity

3.34

(.53)

−.11∗∗

.49∗∗

1(4)S

killdiscretio

n3.36

(.55)

−.03

.57∗∗

.37∗∗

1(5)Job

autono

my

2.68

(.60)

−.08

.36∗

.30∗∗

.39∗∗

1(6)S

uppo

rtfro

msuperio

rs2.84

(.60)

−.13∗∗

.24∗∗

.27∗∗

.14∗∗

.36∗∗

1(7)S

uppo

rtfro

mcolleagues

3.07

(.59)

−.18∗∗

.30∗∗

.28∗∗

.22∗∗

.30∗∗

.36∗∗

1(8)F

airness

2.36

(.50)

−.13∗∗

.15∗∗

.20∗∗

.08

.27∗∗

.41∗∗

.35∗∗

1(9)S

uppo

rtfro

morganizatio

n2.51

(.62)

−.19∗∗

.18∗∗

.19∗∗

.11∗∗

.41∗∗

.50∗∗

.32∗∗

.57∗∗

1(10)S

ocialutility

2.73

(.54)−.26∗∗

.43∗∗

.39∗∗

.22∗∗

.31∗∗

.36∗∗

.36∗∗

.43∗∗

.43∗∗

1(11)E

motionalexh

austion

2.06

(1.28)

.41∗∗

−.21∗∗

−.21∗∗

−.13∗∗

−.27∗∗

−.35∗∗

−.40∗∗

−.29∗∗

−.38∗∗

−.40∗∗

1(12)D

epersonalization

1.21(1.18)

.43∗∗

−.19∗∗

−.23∗∗

−.08∗

−.12∗∗

−.23∗∗

−.20∗∗

−.18∗∗

−.21∗∗

−.33∗∗

.55∗∗

1Note:∗∗

<.001;∗<.05.


performed, predictor variables were entered within threesuccessive steps. In the first step, demographical (gender, age,and marital status) and occupational (years in the healthsector and type of unit) variables were entered as controlvariables. In the second step, standardized indexes of verbalaggression and job resources were entered. In the thirdstep, the interaction term, which is the product betweenverbal aggression and job resource, was entered. Whenthe interaction term showed significant value, the simpleslope procedure recommended by Aiken and West [62] wasadopted to further examine the pattern of the relationship.

The risk of multicollinearity between independent vari-ables was controlled by standardizing all indexes. Analysesindicated that there were no signs of multicollinearity in anyof the regression models. For each independent variable, thetolerance index (1/VIF) never exceeded the score of .70 (cut-off < .20 [63]).

3. Results and Discussion

3.1. Nurses. Table 3 reports the results of the moderatedhierarchical regressions in which emotional exhaustion wasentered as a dependent variable. In the third step, allthe models reported significant 𝑅2 and showed a varianceexplained that ranged from 20% (model 3; JR: skill discretion)to 31% (model 6; JR: support from colleagues). Concerningcontrolling variables, gender showed a significant associationwith emotional exhaustion only in model 1 (JR: meaning ofthe job). The type of unit was found significant in all ninemodels, indicating that nurses employed in medical unitsare more prone to develop emotional exhaustion than nursesin the emergency units. Verbal aggression was found to besignificant in all models, and its 𝛽 coefficients ranged from.35 to .44.

Regarding the main effect, all the resources we con-sidered, except skill discretion, helped lessen emotionalexhaustion.The smallest 𝛽 coefficient was found for meaningof work with .12, and the largest was found for support fromcolleagues with .35.

The interaction effect between verbal aggression and jobresources was found to be significant in models 1, 3, 5, and 7,suggesting that meaning of work (𝛽 = −.11), skill discretion(𝛽 = −.11), support from superiors (𝛽 = −.12), andfairness (𝛽 = −.11) buffer the effects of verbal aggression onemotional exhaustion.

In all these cases, the simple slope analysis (see Figures1–4) showed that when the job resources were high (+1standard deviation, SD), verbal aggression was positively andsignificantly related to emotional exhaustion. However, whenthe job resources were low (−1 SD), the relationship wasstronger (𝛽 = .63, 𝑡 = 7.63, 𝑝 = .00). In particular, for workmeaning, the slope at +1 DS showed a 𝛽 of .39 (𝑡 = 4.67,𝑝 = .00), whereas at −1 DS, the 𝛽 value reached .63 (𝑡 = 7.63,𝑝 = .00). Similarly, the association between verbal aggressionand emotional exhaustion was weaker when skill discretionwas high (𝛽 = .61, 𝑡 = 2.65, 𝑝 = .01), rather than whenskill discretion was low (𝛽 = .85, 𝑡 = 5.19, and 𝑝 = .00).Concerning support from superiors, the value of 𝛽 at −1 SD

5

4.5

4

3.5

3

2.5

2

1.5

1

Emot

iona

l exh

austi

on

Low verbal aggression High verbal aggression

Low meaning of the workHigh meaning of the work

Figure 1: Interaction between verbal aggression andmeaning of thework for emotional exhaustion among nurses.

2.5

2

1.5

1

0.5

0

Emot

iona

l exh

austi

on


Low skill discretionHigh skill discretion

Figure 2: Interaction between verbal aggression and skill discretionfor emotional exhaustion among nurses.

was equal to .73 (𝑡 = 8.76, 𝑝 = .00), whereas at +1 SD,𝛽 was equal to .43 (𝑡 = .43, 𝑝 = .00). Finally, regardingfairness, the value of 𝛽 at −1 SD was equal to .77 (𝑡 = 8.60,𝑝 = .00), whereas at +1 SD, 𝛽 was equal to .53 (𝑡 = 4.83,𝑝 = .00). Therefore, the slope tests further supported thatthese resources moderated the effect of verbal aggression inincreasing emotional exhaustion in the expected direction.

Table 4 shows the results for depersonalization. Withincontrol variables, gender (in all models) and marital status(in some) were significant. Based on these results, menand people who do not have a partner have more risk ofdeveloping depersonalization. Verbal aggression significantlypredicted depersonalization in all the models. All resources


Table3:Mod

erated

hierarchicalregressio

nsto

measure

mainandinteractioneffectsof

verbalaggressio

nandjobresourceso

nem

otionalexh

austionam

ongnu

rses.

Nurses

M1JR

M2JR

M3JR

M4JR

M5JR

M6JR

M7JR

M8JR

M9JR

Meaning

ofthew

ork

Rolecla

rity

Skill

discretio

nJobautono

my

Supp

ortfrom

superio

rSupp

ortfrom

colleagues

Fairn

ess

Organizationalsup

port

Organizational

socialutility

Emotiona

lexh

austion

Step

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

(1) Gender(1=

female)

.09∗

1.98∗

.09

1.87

.09

1.85

.80

1.72

.07

1.48

.08

1.77

.09

1.81

.07

1.51

.07

1.67

Age

(1≥40

).05

.95

.05

.84

.05

.89

.03

.07

.05

.89

.06

1.16

.04

.67

.40

.70

.05

.85

Marita

lstatus(1=

marrie

d/partnered=1)−.02

−.32

−.01−.09−.02−.40−.04−.87−.05−1.0

8−.04

−.85−.40−.82−.04

−.91

−.17

−.37

Year

health

sector

(1≥15)

.03

.55

.05

.92

.04

.71

.04

.75

.02

.45

.05

1.05

.04

.75

.01

.34

.03

.55

Type

ofun

it(1=em

ergency)

−.14∗∗

−2.85∗∗

−.14

2∗∗

−2.82∗∗

−.14∗∗

−2.73∗∗

−.12∗

−2.30∗

−.10∗

−2.19∗

−.14∗∗

−3.01∗∗

−.11∗

−2.17∗

−.11∗

−2.35∗

−.13∗∗

−2.61∗∗

(2) Ve

rbalaggressio

n.42∗∗∗

8.55∗∗∗

.43∗∗∗

8.71∗∗∗

.44∗∗∗

9.11∗∗∗

.41∗∗∗

8.50∗∗∗

.36∗∗∗

7.45∗∗∗

.35∗∗∗

7.3∗∗∗

.40∗∗∗

7.9∗∗∗

.37∗∗∗

7.72∗∗∗

.38

7.80

Jobresource

−.12∗

−2.45∗

−.13∗∗

2.59∗∗

−.077−1.5

8−.26∗∗∗

−5.59∗∗∗

−.25∗∗∗

−5.40∗∗∗

−.35∗∗∗

−7.5

1∗∗∗

−.16∗∗

−3.46∗∗

−.29∗∗∗

−6.08∗∗∗

−.28∗∗∗

−6.01∗∗∗

(3) Ve

rbalaggressio

n×jobresource

−.11∗

−2.25∗

−.44−.86−.11∗

−2.38∗

.02

.40

−.12∗∗

−2.64∗∗

−.06

−1.2

5−.11∗

−2.30∗

−.01

−.16

4−.07−1.4

6(2)v

ersus(1)Δ𝑅2

.21∗∗∗

.21∗∗∗

.19∗∗

.25∗∗∗

.25∗∗∗

.31∗∗∗

.28∗∗∗

.26∗∗∗

.26∗∗∗

(3)v

ersus(2)Δ𝑅2

.01∗

.00

.01∗

.00

.01

.00

.09∗∗

.00

.00

Adj𝑅2

.21∗∗∗

.21∗∗∗

.20∗∗∗

.25∗∗∗

.27∗∗∗

.31∗∗∗

.22∗∗∗

.25∗∗∗

.27∗∗∗

Note:∗

.05≤𝑝≤.011;∗∗

.01≤𝑝≤.001;∗∗∗

=.00.


Table4:Mod

erated


nsto

measure


verbalaggressio

nandjobresourceso

ndeperson

alizationam

ongnu

rses.

Nurses

M1JR

M2JR

M3JR

M4JR

M5JR

M6JR

M7JR

M8JR

M9JR

Meaning

ofthew

ork

Rolecla

rity

Skill

discretio

nJobautono

my

Supp

ortfrom

superio

rSupp

ortfrom

colleagues

Fairn

ess

Organizationalsup

port

Organizational

socialutility

Depersona

lization

Step

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

(1) Gender(1=

female)

−.17∗∗∗

−3.7∗∗∗

−.16∗∗∗

−3.50∗∗∗

−.15∗∗

−3.23∗∗

−.17∗∗∗

−3.66∗∗∗

−.19∗∗∗

−3.98∗∗∗

−.17∗∗∗

−3.7∗∗∗

−.18∗∗∗

−3.71∗∗∗

−.18∗∗∗

−3.72∗∗∗

.19∗∗∗

−4.08∗∗∗

Age

(1≥40

).04

.74−.00−.02

.01

.12−.01−.18

.02

.27

.01

.25

−.01−.16

.03

.43

.01

.08

Marita

lstatus

(1=marrie

d/partnered=1)−.10∗

−2.14∗

−.07−1.3

8−.09−1.9

1−.11∗∗

−2.25∗∗

−.12∗

−2.43∗

−.11∗

−2.12∗

−.12∗

−2.37∗

−.11∗

−2.17∗

−.10

−1.7

8

Year

health

sector

(1≥15)

−.03

−.63

−.00−.07−.02−.40−.01−.21

−.04−.65

−.01

−.16

−.03−.63

−.04

−.71

−.02−.28

Type

ofun

it(1=em

ergency)

.08

1.52

.07

1.48

.07

1.36

.07

1.43

.06

1.19

.03

.51

.05

1.05

.06

1.17

.06

1.26

(2) Ve

rbalaggressio

n.29∗∗∗

5.90∗∗∗

.27∗∗∗

5.62∗∗∗

.31∗∗∗

6.27∗∗∗

.28∗∗∗

5.53∗∗∗

.27∗∗∗

5.24∗∗∗

.27∗∗∗

5.13∗∗∗

.30∗∗∗

5.72∗∗∗

.30∗∗∗

5.98∗∗∗

.25∗∗∗

5.06∗∗∗

Jobresource

−.17∗∗∗

−3.45∗∗∗

−.19∗∗∗

−3.74∗∗∗

−.14∗∗

−2.75∗∗

−.18∗∗∗

−3.71∗∗∗

−.19∗∗∗

−3.18∗∗∗

−.18∗∗

−3.50∗∗

−.20∗∗∗

−4.00∗∗∗

−.17∗∗

−3.49∗∗

−.26∗∗∗

−5.32∗∗∗

(3) Ve

rbalaggressio

n×Jobresource−.19∗∗∗

−3.92∗∗∗

−.12∗

−2.41∗

−.15∗∗

−3.08∗∗

−.04−.73

−.06−1.2

9−.01

−.10

−.01−.10

.02

.33−.08−1.7

2(2)v

ersus(1)Δ𝑅2

.14∗∗∗

.15∗∗∗

.12∗∗∗

.13∗∗∗

.13∗∗∗

.12∗∗∗

.14∗∗∗

.13∗∗∗

.16∗∗∗

(3)v

ersus(2)Δ𝑅2

.03∗∗∗

.01∗

.02∗

.00

.00

.00

.00

.00

.01

Adj𝑅2

.23∗∗∗

.21∗∗∗

.18∗∗∗

.17∗∗∗

.18∗∗∗

.16∗∗∗

.18∗∗∗

.18∗∗∗

.21∗∗∗

Note:∗

.05≤𝑝≤.011;∗∗

.01≤𝑝≤.001;∗∗∗

=.00.


2.5

2

1.5

1

0.5

0

Emot

iona

l exh

austi

on

Low support from superiorHigh support from superior


Figure 3: Interaction between verbal aggression and support fromsuperior for emotional exhaustion among nurses.

2.5

2

1.5

1

0.5

0

Emot

iona

l exh

austi

on

Low fairnessHigh fairness


Figure 4: Interaction between verbal aggression and fairness foremotional exhaustion among nurses.

were found to be significantly related to depersonalization; 𝛽coefficients indicated that, of these variables, only meaningof work (𝛽DP = −.17 to 𝛽EE = −.12), role clarity (𝛽DP = −.19to 𝛽EE = −.13), and skill discretion (𝛽DP = −.14 to 𝛽EE =−.07) have a stronger relationship with depersonalizationthan emotional exhaustion. In the third step, entering theinteraction term produced a significant incremental changeof 𝑅2 only for three content-level resources: meaning of work(Δ𝑅2 = .03), role clarity (Δ𝑅2 = .01), and skill discretion(Δ𝑅2 = .02).

Figures 5–7 clearly suggest that meaning of work, roleclarity, and skill discretion act as buffers in the relationshipbetween verbal aggression and depersonalization. Further

2.5

2

3

1.5

1

0.5

0


Low meaning of the workHigh meaning of the work

Dep

erso

naliz

atio

n

Figure 5: Interaction between verbal aggression andmeaning of thework for depersonalization among nurses.


Dep

erso

naliz

atio

n

Low role clarityHigh role clarity

2

1.8

1.6

1.4

1.2

1

0.8

0.6

0.4

0.2

0

Figure 6: Interaction between verbal aggression and role clarity fordepersonalization among nurses.

confirmationwas provided by the slope test analyses. Accord-ing to these, when work meaning was high, the associationbetween verbal aggression and depersonalization was notsignificant (𝛽 = .12, 𝑡 = 1.61, and 𝑝 = .11), whereas in thecase of low work meaning, the relationship between verbalaggression and depersonalizationwas positive and significant(𝛽 = .51, 𝑡 = 6.70, 𝑝 = .00). As regards role clarity, theassociation between verbal aggression and depersonalizationwas significant in both conditions. However, the relationshipwas weaker in conditions of high role clarity (𝛽 = .19, 𝑡 =2.18, and𝑝 = .03), rather than in conditions of low role clarity(𝛽 = .63, 𝑡 = 7.10, 𝑝 = .00). Similarly, for skill discretion,



Dep

erso

naliz

atio

n

1.8

1.6

1.4

1.2

1

0.8

0.6

0.4

0.2

0

Low skill discretionHigh skill discretion

Figure 7: Interaction between verbal aggression and skill discretionfor depersonalization among nurses.

2.5

2

1.5

1

0.5

0

Emot

iona

l exh

austi

on



Figure 8: Interaction between verbal aggression and support fromsuperior for emotional exhaustion among nurse’s aides.

the value of 𝛽 at −1 SD was equal to .62 (𝑡 = 6.57, 𝑝 = .00),whereas at +1 SD, 𝛽 was equal to .31 (𝑡 = 3.20, 𝑝 = .00).

These results confirm H1a because, among nurses, verbalaggression was found significantly associated with both emo-tional exhaustion and depersonalization in all themodels. Onthe other hand, H2a is partially confirmed because the buffereffect of the resource was found in four cases for emotionalexhaustion and three cases for depersonalization.

3.2. Nurse’s Aides. Table 5 shows the results of moderatedhierarchical regressions for emotional exhaustion. Controlvariables showed significant values in none of the cases.


1.8

1.6

1.4

1.2

1

0.8

0.6

0.4

0.2

0

2

Emot

iona

l exh

austi

on

Low support from colleaguesHigh support from colleagues

Figure 9: Interaction between verbal aggression and support fromcolleagues for emotional exhaustion among nurse’s aides.

2.5

2

1.5

1

0.5

0

Emot

iona

l exh

austi

on



Figure 10: Interaction between verbal aggression and fairness foremotional exhaustion among nurse’s aides.

Concerning verbal aggression, nurse’s aides results aresimilar to the nurses’: 𝛽 coefficients in all models showedsignificant values with the lowest value of .31 and the highestof .50, indicating that verbal aggression positively predictsemotional exhaustion. No content level resources displayeda direct effect on emotional exhaustion. On the contrary,support from superiors (𝛽 = −.26) and peers (𝛽 = −.26),fairness (𝛽 = −.27), organizational support (𝛽 = −.27),and utility of the service (𝛽 = −.32) showed a negativesignificant association with emotional exhaustion. In all ofthese models, with the exception of the social utility, theinteraction terms were also significant. Graphs reported inFigures 8–11 indicated the presence of a buffering effect for


Table5:Mod

erated


nsto

measure


verbalaggressio

nandjobresourceso

nem

otionalexh

austionam

ongnu

rse’s

aides.

Nurse’saides

M1JR

M2JR

M3JR

M4JR

M5JR

M6JR

M7JR

M8JR

M9JR

Meaning

ofthew

ork

Rolecla

rity

Skill

discretio

nJobautono

my

Supp

ortfrom

superio

rSupp

ortfrom

colleagues

Fairn

ess

Organizationalsup

port

Organizational

socialutility

Emotiona

lexh

austion

Step

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

Β𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

(1) Gender(1=

female)

.01

.09

.02

.24

.04

.35

.01

.11−.02−.21

.01

.14−.08−.85−.04

−.42

−.05−.54

Age

(1≥40

).01

.09

.01

.08

.01

.09

.01

.12−.03−.27

−.07

−.66

.01

.06

−.06

−.46

.03

.29

Marita

lstatus(1=

marrie


−.86−.07−.68−.10−1.0

3−.10−.99−.06−.62

−.10

−1.0

6−.11−1.2

3−.06

−.66

−.07−.78

Year

health

sector

(1≥15)

.07

.71

.06

.59

.08

.75

.09

.90

.101.0

4.12

1.24

.08

91.05

.50

.111.13

Type

ofun

it(1=em

ergency)

−.10

−.94−.09−.97−.07−.70−.06−.67−.01−.14

.03

.39.05

.57

−.04

−.50

−.07−.79

(2) Ve

rbalaggressio

n.49∗∗∗

4.61∗∗∗

.45∗∗∗

4.69∗∗∗

.50∗∗∗

4.39∗∗∗

.39∗∗

3.32∗∗

.38∗∗∗

4.04∗∗∗

.31∗∗

3.10∗∗

.31∗∗

3.21∗∗

.31∗∗

3.19∗∗

.32∗∗

2.99∗∗

Jobresource

−.02

−.23−.10−.99−.16−1.5

0−.16−1.6

0−.26∗∗

−2.78∗∗

−.26∗∗

−2.62∗∗

−.27∗∗

−2.90∗∗

−.27∗∗

−2.95∗∗

−.32∗∗

−3.24∗∗

(3) Ve

rbalaggressio

n×Jobresource

.04

.42

.171.8

0.14

1.21−.07−.68−.19∗

−1.9

4∗−.26∗∗

−2.68∗∗

−.33∗∗

−3.53∗∗

−.27∗∗

−2.81∗∗

−.06−.69

(2)v

ersus(1)Δ𝑅2

.21∗∗∗

.22∗∗

.22∗∗∗

.24∗∗∗

.31∗∗∗

.31∗∗∗

.28∗∗∗

.27∗∗∗

.30∗∗∗

(3)v

ersus(2)Δ𝑅2

.00

.03

.01

.00

.03∗

.05∗∗

.09∗∗

.06∗∗

.00

Adj𝑅2

.19∗∗∗

.22∗∗∗

.21∗∗∗

.21∗∗∗

.32∗∗∗

.34∗∗∗

.35∗∗∗

.31∗∗∗

.28∗∗∗

Note:∗

.05≤𝑝≤.011;∗∗

.01≤𝑝≤.001;∗∗∗

=.00.


2.5

2

1.5

1

0.5

0

Emot

iona

l exh

austi

on


Low support from organizationHigh support from organization

Figure 11: Interaction between verbal aggression and organizationalsupport for emotional exhaustion among nurse’s aides.

all these job resources in the relationship between verbalaggression and burnout among nurse’s aides.

Further evidence of the moderating role of these jobresources was provided by the slope test. As regards supportfrom colleagues, the relationship between verbal aggressionand emotional exhaustion was significant at −1 SD (lowsupport from colleagues; 𝛽 = .66, 𝑡 = 4.27, 𝑝 = .00) butnot at +1 SD (high support from colleagues; 𝛽 = .21, 𝑡 = 1.09,𝑝 = .27). Similar results were obtained for fairness (−1 SD:𝛽 = .82, 𝑡 = 4.84, and 𝑝 = .00; +1 SD: 𝛽 = .15, 𝑡 = .71,and 𝑝 = .47) and organizational support (−1 SD: 𝛽 = .78,𝑡 = 7.54, and 𝑝 = .00; +1 SD: 𝛽 = .22, 𝑡 = 1.00, and 𝑝 = .31).Concerning support from superiors, the association betweenverbal aggression and emotional exhaustion was significantin both conditions; however, it was weaker in conditions at+1 SD (𝛽 = .44, 𝑡 = 2.43, and 𝑝 = .02) rather than at −1 SD(𝛽 = .77, 𝑡 = 7.40, and 𝑝 = .00).

Table 6 reports results for depersonalization. Gender wassignificant only in themodel inwhich fairness, organizationalsupport, and social utility were entered. Any other controlvariables resulted in no significance in the models. Also, inthis case, results highlighted that verbal aggression negativelypredicted depersonalization (.22 ≤ 𝛽 ≤ .47) in all models.

On the contrary, no resources, except for social utility,showed a direct effect in lessening the depersonalization levelamong nurse’s aides. As highlighted in step three, supportfrom superiors (𝛽 = −.32), colleagues (𝛽 = −.38), and theorganization (𝛽 = −.31) and fairness (𝛽 = −.40) have a rolein moderating the negative effect of verbal aggression. As it ispossible to see inmodel 9, social utility is the unique resourcethat reported both a direct (𝛽 = −.30) and a moderating(𝛽 = −.28) effect on depersonalization.

According to the slopes test (see Figures 12–16), allthese resources exercise a buffer effect, thus moderating thenegative effect of verbal aggression in increasing nurse’s aides


Dep

erso

naliz

atio

n

1.8

1.6

1.4

1.2

1

0.8

0.6

0.4

0.2

0


Figure 12: Interaction between verbal aggression and support fromsuperior for depersonalization among nurse’s aides.


Dep

erso

naliz

atio

n

1.8

1.6

1.4

1.2

1

0.8

0.6

0.4

0.2

0

Low support from colleaguesHigh support from colleagues

Figure 13: Interaction between verbal aggression and support fromcolleagues for depersonalization among nurse’s aides.

depersonalization. Particularly when support from superiorswas high, the association between verbal aggression anddepersonalization was not significant (𝛽 = .21, 𝑡 = 1.66, and𝑝 = .09), whereas in the case of low support from superiors,the association was positive and significant (𝛽 = .51, 𝑡 =4.65, and 𝑝 = .00). Also regarding support from colleagues,the relationship between verbal aggression and emotionalexhaustion was significant at −1 SD (low support from col-leagues; 𝛽 = .66, 𝑡 = 4.27, and 𝑝 = .00) but not at +1 SD (highsupport from colleagues; 𝛽 = .54, 𝑡 = 5.72, and 𝑝 = .00).As suggested by Figure 14, when fairness was high (+1 DS),verbal aggression was positively and significantly related todepersonalization (𝛽 = .32, 𝑡 = 4.18, and 𝑝 = .00). However,


Table6:Mod

erated


nsto

measure


verbalaggressio

nandjobresourceso

ndeperson

alizationam

ongnu

rse’s

aides.

Nurse’saides

M1JR

M2JR

M3JR

M4JR

M5JR

M6JR

M7JR

M8JR

M9JR

Meaning

ofthew

ork

Rolecla

rity

Skill

discretio

nJobautono

my

Supp

ortfrom

superio

rSupp

ortfrom

colleagues

Fairn

ess

Organizationalsup

port

Organizational

socialutility

Depersona

lization

Step

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

𝛽𝑡

(1) Gender(1=

female)

−.14

−1.3

5−.15−1.5

6−.17−1.6

6−.16−1.6

3−.20−2.22

−.12

−1.3

4−.25∗−2.59∗

−.20∗

−2.12∗

−.21∗−2.25∗

Age

(1≥40

).08

.73

.08

.80

.11.98

.09

.86

.111.0

9−.01

−.10

.111.0

7.08

.74.09

.93

Marita

lstatus(1=

marrie


−.36

.02

.23−.01−.14−.01−.08−.03−.29

.04

.37−.45−.48−.02

−.25

−.02−.19

Year

health

sector

(1≥15)

.04

.33.01

.02

.02

.23

.04

.37−.05−.58

.00

.04

.02

.21

−.07

−.73

.07

.80

Type

ofun

it(1=em

ergency)

.02

.21

.02

.18.03

.28

.02

.17.08

.93

.131.4

0.11

1.21

.03

.31.03

.30

(2) Ve

rbalaggressio

n.39∗∗∗

3.72∗∗∗

.42∗∗∗

4.41∗∗∗

.47∗∗∗

4.15∗∗∗

.39∗∗

3.26∗∗

.35∗∗∗

3.82∗∗∗

.28∗∗

2.91∗∗

.31∗

3.16∗

.34∗∗

3.32∗∗

.22∗

2.11∗

Jobresource

−.09

−.99−.08−.85−.02−.18−.03−.26−.13−1.3

8−.14

−1.4

3−.10

−1.0

4−.06

−.65

−.30∗∗

−3.13∗∗

(3) Ve

rbalaggressio

n×Jobresource

−.10

−1.0

0−.15−1.5

6.06

.47−.09−.78−.32∗∗

−3.32∗∗

−.38∗∗∗

−3.85∗∗∗

−.40∗∗∗

−4.27∗∗∗

−.31∗∗

−3.09∗∗

−.28∗∗

−3.02∗∗

(2)v

ersus(1)Δ𝑅2

.20∗∗∗

.19∗∗∗

.19∗∗∗

.19∗∗∗

.13∗∗∗

.24∗∗∗

.19∗∗∗

.19∗∗∗

.24∗∗∗

(3)v

ersus(2)Δ𝑅2

.01

.03

.00

.00

.00

.11∗∗∗

.14∗∗∗

.07∗∗

.07∗∗

Adj𝑅2

.20∗∗∗

.21∗∗∗

.18∗∗∗

.18∗∗∗

.32∗∗∗

.34∗∗∗

.33∗∗∗

.26∗∗∗

.30∗∗∗

Note:∗

.05≤𝑝≤.011;∗∗

.01≤𝑝≤.001;∗∗∗

=.00.



1.8

1.6

1.4

1.2

1

0.8

0.6

0.4

0.2

0

2


Dep

erso

naliz

atio

n

Figure 14: Interaction between verbal aggression and fairness fordepersonalization among nurse’s aides.


1.8

1.6

1.4

1.2

1

0.8

0.6

0.4

0.2

0

Dep

erso

naliz

atio

n

Low support from organizationHigh support from organization

Figure 15: Interaction between verbal aggression and support fromorganization for depersonalization among nurse’s aides.

when fairness was low (−1 SD), the association was consid-erably stronger (𝛽 = .70, 𝑡 = 9.08, and 𝑝 = .00). As regardssupport from the organization, the relationship between ver-bal aggression and depersonalization was significant at −1 SD(low support; 𝛽 = .59, 𝑡 = 3.25, and 𝑝 = .00) but not at +1 SD(high support; 𝛽 = .12, 𝑡 = .26, and 𝑝 = .79). Similar resultswere obtained for organizational social utility (−1 SD:𝛽 = .51,𝑡 = 3.86, and 𝑝 = .00; +1 SD: 𝛽 = .07, 𝑡 = .42, and 𝑝 = .66).

The results confirm H1b because verbal aggression wassignificantly associated with both emotional exhaustion anddepersonalization in all models carried out among nurse’saides. On the other hand, H2b is partially confirmed becausethe buffer effect of the resource was found in five cases foremotional exhaustion and in four cases for depersonalization.


Dep

erso

naliz

atio

n

1.4

1.2

1

0.8

0.6

0.4

0.2

0

Low organizational social utilityHigh organizational social utility

Figure 16: Interaction between verbal aggression and organizationalsocial utility for depersonalization among nurse’s aides.

4. Conclusions

Thefirst aimof the present studywas to verify the relationshipbetween verbal aggression and job burnout. The high andsignificant associations found in both professional groupsconfirmed the hypothesis that verbal aggression is a predictorof burnout (H1a, H1b). These results suggested that not onlyin emergency and psychiatry units, as usually pointed out bythe literature [64, 65], but also in medical units, dealing withverbal aggression from patients and relatives can be a crucialissue which represents an important emotional demand thatcontributes to increased burnout levels among nursing staff.

The second aim of the study was to explore whether anyjob content, social, and organizational level resources arecapable of moderating the effect of the exposure to verbalaggression on burnout. The hypothesis that the resourcesconsidered moderate the relationship between verbal aggres-sion and the burnout symptoms was only partially confirmed(H2a, H2b). Overall, in 45% of the cases, the cross-productbetween verbal aggression and the resource was found tobe significant. From a general point of view, the findingsobtained contribute to enforce the buffering hypothesis ofthe Job Demands-Resources Model (JD-R, [17–19]), becausethe interactions found were all in the expected direction.However, it suggests that not all these resources, even ifimportant for reducing burnout (in all cases, job resourcesshowed significant direct and negative associationswith emo-tional exhaustion and in most cases with depersonalization),are useful to cope with verbal aggression. Indeed, resultshighlight profession-specific patterns in the two occupationalsubgroups considered.

Considering the job content level among nurses, mostof the resources work as moderators of the effect of verbalaggression on burnout. On the contrary, no job contentresources work as buffers among nurse’s aides. These resultscould be attributed to the different nature of the work of these


two categories. The nurses’ work, at the job content level,is richer and more complex than that of nurse’s aides and,thus, may offer more resources to successfully deal with theaggressive patients.

These results are also in accordance with those studieswhich, in the Job Demand Control (JDC) perspective, high-lighted that nurses fall into the active strain category, whereasnurse’s aides are in the high strain category [50, 51]. However,the present study suggests that for workers who have “poor”job control at the content level, such as nurse’s aides, otherjob resources at the social level and the organizationallevel may be available and buffer the negative effect of jobdemand. Indeed, at the social level, among nurse’s aides,both forms of support (from peers and superiors) moderatedemotional exhaustion and depersonalization. Similarly, at theorganization level, most of the resources worked as buffers ofverbal aggression among nurse’s aides.

On the other hand, it is also interesting to note that amongnurses, in most cases, social and organizational resources(with the exception of support from superiors and fairness)did not moderate burnout. These results are difficult tointerpret because previous literature rarely pays attention tothese aspects. However, an explanation of these results canbe found in the Job Characteristic Model by Hackman andOldham [66]: Aggressiveness may lead workers to developdoubts concerning the worth of their job because patientsdo not show appreciation for the efforts provided. Richer jobcharacteristics, as in the case of nurses, may allow them todraw energy from the job per se, thus making the resources ofthe job content level available for coping with aggressiveness.This may also be because motivation comes from the workper se and not from rewards from patients.This psychologicalmechanism does not work with nurse’s aides, for whom thework per se is poorer. Therefore, for them, other aspectsof the context such as the social and the organizationalenvironment (i.e., in terms of social and organizationalsupport, opportunity for positive identification in the serviceprovided by the organization, etc.) may be more salient anduseful for coping with verbal aggression from patients.

Finally, it is interesting to note that the findings do notsupport the matching principle by De Jonge and Dormann[67]. According to this principle, resources are most likely tomoderate the relationship between demands and outcomesif resources, demands, and psychological outcomes all match(e.g., are all at the emotional level). In the present study, it wasfound that verbal aggression (social stressor) interacted withskill discretion (cognitive resource) in predicting emotionalexhaustion (emotional outcome). This finding is in line withsome previous studies [38] and suggests that, more than thematching principle, aspects of thework context, including thetype of job (e.g., nurses versus nurse’s aides), may matter indetermining which resources may act as moderators in therelationship between any type of demand and any type ofoutcome.

Further studies should look more deeply at the differenceof the mechanisms that lead to burnout among the two sub-categories. Moreover, another suggestion concerns the explo-ration of the “positive side” of the patient-nurse relationshipas a resource able to buffer specifically the “negative side”

represented by verbal aggression and exceeding demands[48, 49].

The present study contributes to enlarging empiricalevidence developed in the framework of the JD-R model, inparticular, by focusing on understudied demands (i.e., verbalaggression) and considering a wide range of resources as itspotential moderators.

Moreover, it indicates that more attention should be paidto the study of the stress phenomenon among and acrossnurses and nurse’s aides because the mechanism that leads toburnout seems to be partially different, especially as regardsthe functioning of job resources as moderators. From a stressmanagement perspective, the present study suggests thatwhereas job content level resources should be reinforced tohelp nurses copewith aggressiveness frompatients, as regardsnurse’s aides, the attention should be focused on the socialand organizational levels.

The present study is not without limitations. One concernis that a nonrandomized sampling procedure was used. Evenif the sample is quite large, it can limit the generalizability ofthe results founded. Another important limitation is its cross-sectional design. Therefore, caution must be exercised in theinterpretation of the observed associations. It is assumedthat job demands and resources are antecedents of burnout,but the opposite could also be true. In fact, elevated ratesof burnout could lead workers to develop negative attitudestoward jobs, workplace contexts, and organizations.

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper.

References

[1] NIOSH (National Institute forOccupational Safety andHealth),Violence: Occupational Hazard in Hospitals, no. 101, 2002,http://www.cdc.gov/niosh/docs/2002-101/pdfs/2002-101.pdf.

[2] OSHA (Occupational Safety and Health Administration),Guidelines for Preventing Workplace Violence for Health Careand Social Service Workers, Department of Labor, Washington,DC, USA, 1996.

[3] M. Braverman, Preventing Workplace Violence: A Guide forEmployers and Practitioners, Sage, Thousand Oaks, Calif, USA,1999.

[4] B. Beech and P. Leather, “Workplace violence in the health caresector: a review of staff training and integration of trainingevaluation models,” Aggression and Violent Behavior, vol. 11, no.1, pp. 27–43, 2006.

[5] S. van den Bossche, T. Taris, I. Houtman, P. Smulders, and M.Kompier, “Workplace violence and the changing nature of workin Europe: trends and risk groups,” European Journal of Workand Organizational Psychology, vol. 22, no. 5, pp. 588–600, 2013.

[6] D. Camerino, M. Estryn-Behar, P. M. Conway, B. I. J. M. vander Heijden, and H.-M. Hasselhorn, “Work-related factors andviolence among nursing staff in the European NEXT study:a longitudinal cohort study,” International Journal of NursingStudies, vol. 45, no. 1, pp. 35–50, 2008.


[7] EU-OSHA (European Agency for Safety and Health at Work),Workplace Violence and Harassment: A European Picture, Pub-lications Office of the European Union, Luxembourg City,Luxembourg, 2010.

[8] J. E. Arnetz, L. Hamblin, L. Essenmacher, M. J. Upfal, J. Ager,and M. Luborsky, “Understanding patient-to-worker violencein hospitals: a qualitative analysis of documented incidentreports,” Journal of Advanced Nursing, vol. 71, no. 2, pp. 338–348, 2015.

[9] J. A. Duxbury, “An evaluation of staff and patient views ofand strategies employed to manage inpatient aggression andviolence on onemental health unit: a pluralistic design,” Journalof Psychiatric and Mental Health Nursing, vol. 9, no. 3, pp. 325–337, 2002.

[10] J. A. Duxbury and R.Whittington, “Causes andmanagement ofpatient aggression and violence: staff and patient perspectives,”Journal of Advanced Nursing, vol. 50, no. 5, pp. 469–478, 2005.

[11] T. S. Joa and T. Morken, “Violence towards personnel in out-of-hours primary care: a cross-sectional study,” ScandinavianJournal of Primary Health Care, vol. 30, no. 1, pp. 55–60, 2012.

[12] G. L. Gillespie, D. M. Gates, and P. Berry, “Stressful incidents ofphysical violence against emergency nurses,”TheOnline Journalof Issues in Nursing, vol. 18, no. 1, article 2, 2013.

[13] M. Bernaldo-De-Quiros, A. T. Piccini, M. M. Gomez, and J.C. Cerdeira, “Psychological consequences of aggression in pre-hospital emergency care: cross sectional survey,” InternationalJournal of Nursing Studies, vol. 52, no. 1, pp. 260–270, 2015.

[14] N. Magnavita, “Experience of prevention activities in localhealth units. Assaults andmusculoskeletal disorders,” LaMedic-ina del Lavoro, vol. 100, no. 4, pp. 24–28, 2009.

[15] C. Guglielmetti, S. Gilardi, L. Accorsi, and D. Converso, “Larelazione con i pazienti in sanita: quali risorse lavorativeper attenuare l’impatto degli stressor sociali?” Psicologia dellaSalute, vol. 2, pp. 121–138, 2014.

[16] N. Swain and C. Gale, “A communication skills interventionfor community healthcare workers reduces perceived patientaggression: a pretest-postest study,” International Journal ofNursing Studies, vol. 51, no. 9, pp. 1241–1245, 2014.

[17] E. Demerouti, F. Nachreiner, A. B. Bakker, and W. B. Schaufeli,“The job demands-resources model of burnout,” Journal ofApplied Psychology, vol. 86, no. 3, pp. 499–512, 2001.

[18] C. Balducci, L. Avanzi, and F. Fraccaroli, “Emotional demandsas a risk factor for mental distress among nurses,” La Medicinadel Lavoro, vol. 105, no. 2, pp. 100–108, 2014.

[19] A. B. Bakker, E. Demerouti, and A. I. Sanz-Vergel, “Burnoutand work engagement: the JD–R approach,” Annual Review ofOrganizational Psychology and Organizational Behavior, vol. 1,no. 1, pp. 389–411, 2014.

[20] C. Dormann and D. Zapf, “Customer-related social stressorsand burnout,” Journal of Occupational Health Psychology, vol.9, no. 1, pp. 61–82, 2004.

[21] G. A. C.-D. la Fuente, C. Vargas, C. San Luis, I. Garcıa, G. R.Canadas, and E. I. de la Fuente, “Risk factors and prevalenceof burnout syndrome in the nursing profession,” InternationalJournal of Nursing Studies, vol. 52, no. 1, pp. 240–249, 2015.

[22] A. Malach-Pines, “Nurses’ burnout: an existential psychody-namic perspective,” Journal of Psychosocial Nursing and MentalHealth Services, vol. 38, no. 2, pp. 23–31, 2000.

[23] D. E. Green, F. H. Walkey, and A. J. W. Taylor, “The three factorstructure of the Maslach Burnout Inventory,” Journal of SocialBehavior and Personality, vol. 6, pp. 453–472, 1991.

[24] M. Westman, S. E. Hobfoll, S. Chen, O. B. Davidson, and S.Laski, “Organizational stress through the lens of conservation ofresources (COR) theory,” in Exploring Interpersonal Dynamics,vol. 4 ofResearch inOccupational Stress andWell-Being, pp. 167–220, Emerald Group Publishing Limited, Bradford, UK, 2005.

[25] K. A. Rogers and E. K. Kelloway, “Violence at work: personaland organizational outcomes,” Journal of Occupational HealthPsychology, vol. 2, no. 1, pp. 63–71, 1997.

[26] M. M. LeBlanc and E. K. Kelloway, “Predictors and outcomesof workplace violence and aggression,” Journal of AppliedPsychology, vol. 87, no. 3, pp. 444–453, 2002.

[27] A. A. Grandey, D. N. Dickter, and H.-P. Sin, “The customer isnot always right: customer aggression and emotion regulationof service employees,” Journal of Organizational Behavior, vol.25, no. 3, pp. 397–418, 2004.

[28] I. J. H. van Emmerik, M. C. Euwema, and A. B. Bakker, “Threatsof workplace violence and the buffering effect of social support,”Group andOrganizationManagement, vol. 32, no. 2, pp. 152–175,2007.

[29] C. Mayhew and D. Chappell, “Workplace violence: an overviewof patterns of risk and the emotional/stress consequences ontargets,” International Journal of Law and Psychiatry, vol. 30, no.4-5, pp. 327–339, 2007.

[30] K. Aquino and S. Thau, “Workplace victimization: aggressionfrom the target’s perspective,” Annual Review of Psychology, vol.60, pp. 717–741, 2009.

[31] S. J. Johnson, L. Holdsworth, H. Hoel, and D. Zapf, “Customerstressors in service organizations: the impact of age on stressmanagement and burnout,” European Journal of Work andOrganizational Psychology, vol. 22, no. 3, pp. 318–330, 2013.

[32] S. Winstanley and R. Whittington, “Anxiety, burnout andcoping styles in general hospital staff exposed to workplaceaggression: a cyclical model of burnout and vulnerability toaggression,”Work & Stress, vol. 16, no. 4, pp. 302–315, 2002.

[33] S. Hahn, M. Muller, I. Needham, T. Dassen, G. Kok, and R. J.G. Halfens, “Factors associated with patient and visitor violenceexperienced by nurses in general hospitals in Switzerland: across-sectional survey,” Journal of Clinical Nursing, vol. 19, no.23-24, pp. 3535–3546, 2010.

[34] S. Gascon, M. P. Leiter, E. Andres et al., “The role of aggressionssuffered by healthcareworkers as predictors of burnout,” Journalof Clinical Nursing, vol. 22, no. 21-22, pp. 3120–3129, 2013.

[35] J. Y. Huynh, A. H. Winefield, and D. Xanthopoulou, “Socialsupport moderates the impact of demands on burnout andorganizational connectedness: a two-wave study of volunteerfirefighters,” Journal of Occupational Health Psychology, vol. 18,no. 1, pp. 9–15, 2013.

[36] S. I. Giga, C. L. Cooper, and B. Faragher, “The development of aframework for a comprehensive approach to stressmanagementinterventions at work,” International Journal of Stress Manage-ment, vol. 10, no. 4, pp. 280–296, 2003.

[37] A. C. H. Schat and E. K. Kelloway, “Reducing the adverseconsequences of workplace aggression and violence: the buffer-ing effects of organizational support,” Journal of OccupationalHealth Psychology, vol. 8, no. 2, pp. 110–122, 2003.

[38] D. Xanthopoulou, A. B. Bakker, M. F. Dollard et al., “When dojob demands particularly predict burnout?Themoderating roleof job resources,” Journal of Managerial Psychology, vol. 22, no.8, pp. 766–786, 2007.

[39] T. Heponiemi, A. Kouvonen, M. Virtanen, J. Vanska, andM. Elovainio, “The prospective effects of workplace violence


on physicians’ job satisfaction and turnover intentions: thebuffering effect of job control,” BMC Health Services Research,vol. 14, article 19, 2014.

[40] R. A. Karasek, “Job demands, job decision latitude, and mentalstrain: implications for job redesign,” Administrative ScienceQuarterly, vol. 24, no. 2, pp. 285–308, 1979.

[41] T. S. Kristensen,M. Borritz, E. Villadsen, and K. B. Christensen,“The Copenhagen burnout inventory: a new tool for theassessment of burnout,” Work & Stress, vol. 19, no. 3, pp. 192–207, 2005.

[42] P. D. Bliese and C. A. Castro, “Role clarity, work overload andorganizational support: multilevel evidence of the importanceof support,”Work & Stress, vol. 14, no. 1, pp. 65–73, 2000.

[43] J. Lang, J. L.Thomas, P. D. Bliese, and A. B. Adler, “Job demandsand job performance: the mediating effect of psychological andphysical strain and the moderating effect of role clarity,” Journalof Occupational Health Psychology, vol. 12, no. 2, pp. 116–124,2007.

[44] R. A. Karasek and T.Theorell,HealthyWork: Stress Productivity,and the Reconstruction of Working Life, Basic Books, New York,NY, USA, 1990.

[45] L. Rhoades and R. Eisenberger, “Perceived organizational sup-port: a review of the literature,” Journal of Applied Psychology,vol. 87, no. 4, pp. 698–714, 2002.

[46] C. Maslach and M. P. Leiter, “Early predictors of job burnoutand engagement,” Journal of Applied Psychology, vol. 93, no. 3,pp. 498–512, 2008.

[47] M. Elovainio, K. van den Bos, A. Linna et al., “Combinedeffects of uncertainty and organizational justice on employeehealth: Testing the uncertainty management model of fairnessjudgments among Finnish public sector employees,” SocialScience & Medicine, vol. 61, no. 12, pp. 2501–2512, 2005.

[48] F. Avallone, A. Paplomatas, and Salute Organizzativa, Psicologiadel benessere nei contesti lavorativi, Raffaello Cortina Editore,Milano, Italy, 2005.

[49] P. Harris, S. Nagy, N. J. Vardaxis, and N. Vardaxis, Mosby’sDictionary of Medicine, Nursing and Health Professions, ElsevierAustralia, 2009.

[50] J. A. Seago and J. Faucett, “Job strain among registered nursesand other hospital workers,” Journal of Nursing Administration,vol. 27, no. 5, pp. 19–26, 1997.

[51] D. G. Morgan, K. M. Semchuk, N. J. Stewart, and C. D’Arcy,“Job strain among staff of rural nursing homes: a comparisonof nurses, aides, and activity workers,” Journal of NursingAdministration, vol. 32, no. 3, pp. 152–161, 2002.

[52] E. Fiabane, I. Giorgi, C. Sguazzin, and P. Argentero, “Workengagement and occupational stress in nurses and other health-care workers: the role of organisational and personal factors,”Journal of Clinical Nursing, vol. 22, no. 17-18, pp. 2614–2624,2013.

[53] R. Karasek, Job Content Instrument Questionnaire and User’sGuide, Version 1.1, Department of Industrial and Systems Engi-neering, University of Southern California, Los Angeles, Calif,USA, 1985.

[54] M. P. Leiter and C. Maslach, Organizational Checkup Survey,John Wiley & Sons, Jossey-Bass, San Francisco, Calif, USA,2000.

[55] L. Borgogni, D. Galati, L. Petitta, and Petitta and Centro Forma-zione Schweitzer, Il questionario Checkup organizzativo, Man-uale dell’adattamento italiano, Organizzazioni Speciali, Firenze,Italy, 2005.

[56] S. I. Cho, S. H. Lim, K. D. Um et al., “A pilot study for JCQ2.0 among Korean subway workers,” in Proceedings of the 2ndICOH International Conference on Psychosocial Factors atWork,Okayama, Japan, August 2005.

[57] C. Maslach and S. E. Jackson, Maslach Burnout Inventory,Consulting Psychologists Press, Palo Alto, Calif, USA, 2ndedition, 1986.

[58] S. Sirigatti and C. Stefanile, Adattamento italiano del MBI-Maslach burnout inventory, vol. 1, Organizzazioni Speciali,Florence, Italy, 1993.

[59] B. Loera, D. Converso, S. Viotti, and S. Federici, “Evaluatingthe psychometric properties of the maslach burnout inventory-human services survey (MBI-HSS) among italian nurses: howmany factors must a researcher consider?” PLoS ONE, vol. 9,no. 12, Article ID e114987, 2014.

[60] G. E. R. Tummers, J. A. Landeweerd, and G. G. van Merode,“The diversity of work: differences, similarities and relation-ships concerning characteristics of the organisation, the workand psychological work reactions in intensive care and non-intensive care nursing,” International Journal of Nursing Studies,vol. 39, no. 8, pp. 841–855, 2002.

[61] S. Viotti, D. Converso, and B. Loera, “Job satisfaction, jobburnout and their relationships with work’ and patients’ char-acteristics: a comparison between intensive care units (ICU)and not-intensive care units (non-ICU),” Giornale Italiano diMedicina del Lavoro ed Ergonomia, vol. 34, no. 2, pp. B52–B60,2012.

[62] L. S. Aiken and S. G. West, Multiple Regression: Testing andInterpreting Interactions, Sage, Newbury Park, Calif, USA, 1991.

[63] A. Field, Discover Statistics Using SPSS, Sage, London, UK, 3rdedition, 2009.

[64] G. L. Gillespie, D. M. Gates, M. Miller, and P. K. Howard,“Emergency department workers’ perceptions of security offi-cers’ effectiveness during violent events,” Work, vol. 42, no. 1,pp. 21–27, 2012.

[65] F. I. Gunasekara, S. Butler, T. Cech et al., “How do intoxi-cated patients impact staff in the emergency department? Anexploratory study,” The New Zealand Medical Journal, vol. 124,no. 1336, pp. 14–23, 2011.

[66] J. R. Hackman and G. R. Oldham, Work Redesign, vol. 72,Addison-Wesley, Reading, Mass, USA, 1980.

[67] J. de Jonge and C. Dormann, “The DISC model: demand-induced strain compensation mechanisms in job stress,” inOccupational Stress in the Service Professions, pp. 43–74, Taylor& Francis, London, UK, 2003.

Research ArticleThe Relationship of On-Call Work with Fatigue, Work-HomeInterference, and Perceived Performance Difficulties

Carla M. Ziebertz,1 Madelon L. M. van Hooff,1 Debby G. J. Beckers,1 Wendela E. Hooftman,2

Michiel A. J. Kompier,1 and Sabine A. E. Geurts1

1Behavioural Science Institute, Radboud University, Montessorilaan 3, 6525 HR Nijmegen, Netherlands2Netherlands Organisation for Applied Scientific Research (TNO), Schipholweg 77-89, 2316 ZL Leiden, Netherlands

Correspondence should be addressed to Carla M. Ziebertz; [email protected]

Received 3 February 2015; Revised 8 May 2015; Accepted 24 May 2015

Academic Editor: Sergio Iavicoli

Copyright © 2015 Carla M. Ziebertz et al. This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.

Objectives. This study examined the relationship between on-call duty exposure (active and total on-call hours a month, numberof calls per duty) and employees’ experiences of being on-call (stress due to unpredictability, ability to relax during inactive on-callperiods, restrictions during on-call duties, on-call work demands, and satisfaction with compensation for on-call duties) on theone hand and fatigue, strain-based and time-based work-home interference (WHI), and perceived on-call performance difficulties(PPD) on the other hand. Methods. Cross-sectional survey data were collected among a large heterogeneous sample of Dutchemployees (𝑁 = 5437).The final sample consisted of 157 on-call workers (23–69 years, 71%males). Data were analyzed bymeans ofhierarchical regression analyses (controlling for age and job characteristics). Results. Differences in on-call work exposure were notsystematically related to fatigue, WHI, and PPD (all p’s > 0.50). The experience of being on-call explained a medium proportion ofthe variation in fatigue and strain-basedWHI and a medium to large proportion of the variation in time-basedWHI and PPD overand above the control variables. Conclusions. Our results suggest that it is employees’ experience of being on-call, especially theexperience of stress due to the unpredictability, rather than the amount of exposure, that is related to fatigue, WHI, and perceivedon-call performance difficulties.

1. Introduction

On-call work refers to work done on an “as needed basis,”meaning that employees must be available at certain timesto be called to work if required by the employer. Typically,this form of scheduling is used to provide 24/7 coveragein facilities where emergencies that need to be dealt withimmediately can occur [1]. On-call work occurs in a widevariety of occupations, such as firemen, police officers,doctors, midwives, utility workers, engineers, informationtechnologists, and airline pilots [1, 2].

Previous research has shown that on-call work can havenegative effects on employees’ well-being and work-relatedoutcomes such as performance and turnover intentions [3–8]. However, most studies have focused on medical staff withon-site standby duties (i.e., duties during which employeesremain at the workplace and that count as working time).

The aim of the present study was to gainmore insight into theconsequences of another type of on-call work: “off-site” on-call duties duringwhich employees do not remain at work butcan be called to work in case of an emergency. In order to doso, we first examined how exposure to off-site on-call dutiesrelates to fatigue, work-home interference, and performancedifficulties. Second, we examined employees’ experience ofthis on-call exposure. Occupational health research hasshown that exposure to work can be especially detrimentalwhen employees’ experiences of the work are unfavorable(e.g., [9–12]). Therefore, we studied not only “objective”exposure to on-call duties in relation to fatigue, WHI, andperformance difficulties, but also how employees’ experiencesof being on-call relate to these outcome measures.

In the following, the potential consequences of (i) on-call duty exposure and (ii) on-call duty experiences will bediscussed.


http://dx.doi.org/10.1155/2015/643413


1.1. Exposure to Off-Site On-Call Duties. Exposure to on-callwork is likely to affect employees’ recovery fromwork. InTheNetherlands and other European countries, off-site on-callduties are officially considered rest time, not working time[13]. Only when an employee is called to work, the activehours are legally considered working hours. This means thatemployees can be on-call in between regular work-hours,that is, during time that is usually reserved for recoveryfrom work. Recovering from work-related load reactions iscritical for employees’ well-being and health [14]. Accordingto the effort-recovery model [15], a long-lasting situationof incomplete recovery from load effects (e.g., fatigue) thatunavoidably build up due to effort expenditure at workeventually results in chronic load reactions, which, accordingto allostatic load theory [16], lead to impaired health (e.g.,[17, 18]).

There are several ways in which exposure to on-call workcan affect recovery. First of all, being called to work during anon-call duty means an interruption of employees’ free time,an extension of exposure to work demands, and, thus, lesstime for recovery. Second, restrictions with regard to locationand activities during on-call duties (e.g., having to stay withina certain radius from the workplace and to abstain fromalcohol) may interfere with employees’ leisure activities andcause work-home interference and thereby impair recovery[6, 10]. Third, it is likely that the restrictions and thepossibility of being called to work interfere with the ability topsychologically detach fromwork. Psychological detachmentrefers to mentally disengaging from work and not thinkingabout work-related issues [9]. A lack of detachment relatesto negative recovery-related outcomes such as fatigue, work-home interference, and emotional exhaustion (e.g., [19–21]).

In addition to the potential negative effects on employees’well-being, such as fatigue and work-home interference, on-call work might also have negative effects on work perfor-mance. Previous research has shown that frequent standbyduties are associated with, for example, decreased visualmemory, reaction times, vigilance, and clinical performance(e.g., [3, 22–24]). In driving simulation experiments, employ-ees’ performance after a standby duty has been found to becomparable to the effect of a 0.05% [23] or even 0.1% bloodalcohol concentration [25, 26]. There are several reasonsto assume that performance is also reduced during off-site on-call duties. First, during on-call duties, employeescan be called to work while not fully recovered from priorwork efforts, which may result in a suboptimal condition toperform well. Second, in case of a call, workers are quiteabruptly drawn from a nonwork ambiance into a high effortwork situation. This “switch” might make it more difficult tooptimally direct one’s attention towards work-related tasks.

Taken together, there are reasons to believe that expo-sure to on-call work may have negative consequences suchas fatigue, work-home interference, and performance dif-ficulties. Based on the effort-recovery model [15] our firsthypothesis is that the amount of exposure to on-call duties ispositively related to (a) fatigue, (b) strain-basedWHI (occurswhen tensions built up at work are transferred to private life[27]), (c) time-based WHI (occurs when work obligationsmake it timewise impossible tomeet obligations in the private

domain [27]), and (d) on-call performance difficulties. In thepresent study, exposurewas operationalized as (i) the numberof hours a month employees are on call, (ii) the number ofactive on-call hours (i.e., working hours) a month, and (iii)the average number of calls per on-call duty.

1.2. Experience of Being On-Call. As mentioned above,besides factual exposure, we also took employees’ experienceof the on-call duties into account. Previous research hasshown that exposure to work can be especially detrimentalwhen employees’ work experiences are unfavorable (e.g., [9–12]). Off-site on-call duties may be experienced unfavorablyfor several reasons.

First of all, on-call duties can be experienced as stressfuldue to the high unpredictability and the lack of controlover whether or not one will be called to work. Previousresearch has shown that unpredictability can indeed causestress [28]. Second, the unpredictability and restrictions mayalso interfere with employees’ ability to relax, which in turnis important for psychological detachment and recovery[9]. Another cause of stress may be that employees areusually only called in case of a critical incident when thereis no one else to deal with it. As such, the work may beexperienced as taxing and demanding, which in turn maylead to stress [15, 29]. Finally, since on-call duties officiallycount as rest time, only the actual working hours have tobe compensated for [13]. Employees’ satisfaction with thecompensation they receive is likely to affect the consequencesof on-call duties. According to Siegrist’s [30] effort-rewardimbalance model, perceived imbalance between employees’amount of effort and the rewards they receive for their effortsleads to negative consequences such as fatigue. Furthermore,in case of satisfactory compensation, employees may bemorelikely to experience their on-call duties as working time,which might make the restrictions more acceptable and on-call duties potentially even desirable.

Based on the effort-recoverymodel [15], the effort-rewardimbalancemodel [30], and previous research on the relevanceof the psychological experience ofwork (e.g., [6, 10, 11, 31, 32]),our second hypothesis is that the more employees experiencetheir on-call duties as unfavourable, the more (2a) fatigue,(2b) strain-based WHI, (2c) time-based WHI, and (2d) on-call performance difficulties they will show. In the presentstudy, experience of on-call duties was operationalized as(i) the experience of stress due to the lack of control overwhether or not being called to work, (ii) the ability to relaxduring inactive on-call periods, (iii) the perceived on-callwork demands, (iv) restrictions, and (v) the satisfaction withthe compensation for on-call duties.

2. Method

2.1. Procedure and Participants. The data were collected bymeans of an online questionnaire in autumn 2013. Thesample was derived from an earlier questionnaire study(Netherlands Working Conditions Survey (NWCS), year2010) conducted by theNetherlandsOrganisation forAppliedScientific Research (TNO) and Statistics Netherlands [33].The link to the questionnaire of the present study was sent


to all respondents of the earlier questionnaire who hadvolunteered to participate in subsequent studies (𝑁 = 5437).Reminders were sent two and three weeks after the initialinvitation. The response rate was 33.1% (𝑁 = 1798). Outof the 1798 employees, 203 (11.3%) indicated working on-calland were therefore relevant to the present study.

Respondents who indicated working less than 24 hours aweek (𝑛 = 19) were excluded from the present study becausework should constitute a substantial part of participant’s liveswhen studying the consequences of work. Furthermore, oneparticipant who indicated working more than 48 hours aweek on contract (i.e., more than the legal maximum) wasexcluded. Likewise, respondents who indicated working lessthan one or more than 336 hours a month on-call (i.e., morethan the legal maximum) were excluded from the presentstudy (𝑛 = 21). Finally, five respondents were excluded dueto missing data on several study variables. This resulted in afinal sample of 157 participants (70.7% males).

On average, participants worked 36.2 hours a week oncontract (SD = 4.91). Their mean age was 45.0 years (SD =11.07) with a range from 23 to 69 years. Most of the partici-pants had completed higher vocational education (38.9%) orvocational training (26.1%) and 17.8% had a university degree.About 16.0% of the participants had completed secondaryschool and 1.2% had completed elementary school or didnot complete any education. Seventy-seven percent of theparticipants were married or cohabiting and 44.6% hadchildren living in their household. Most respondents workedas social workers or in health care professions (22.9%), asspecialists (i.e., IT specialists, engineers; 21.7%) or in serviceprofessions (14.6%).

2.2. Measures

2.2.1. On-Call Duty Exposure. Respondents were asked toindicate the frequency of on-call duties per month, theaverage duration of those duties in hours, and the timespent working during an average on-call duty. Based onthese exposure items, the number of on-call hours per month(product of the frequency of on-call duties a month andthe duration of one on-call duty) and the number of activeon-call hours per month (product of the frequency of on-call duties per month and the average amount of time spentworking during an average on-call duty) were computed.Theaverage number of calls per on-call duty was assessed with thefollowing item: “on average, how many times are you calledto work during one on-call duty?”

2.2.2. Experience of Being On-Call. Due to a lack of validatedscales, all on-call work experiences were assessed with self-developed items. Response scales were based on the well-known Dutch grade notation system ranging from 1 to 10[34]. Satisfaction with compensation for on-call duties wasassessed with the following item: “how satisfied are you withthe compensation of your on-call duties?” (1 = extremelyunsatisfied, 10 = extremely satisfied). On-call work demandswere assessed with the item: “how taxing is the work thatyou have to do when being called during an on-call duty?”(1 = not at all taxing, 10 = extremely taxing). The experience

of restrictions (on-call restrictions) was assessed with theitem: “to what extent do you feel restricted during on-callduties (e.g., in choosing leisure activities)?” (1 = not at allrestricted, 10 = extremely restricted). The ability to relax (on-call relaxation) was assessed with the item “how well can yourelax during periods in which you do not have to work duringon-call duties?” (1 = not at all, 10 = extremely well). Finally,the experience of stress due to unpredictability (on-call stress)was assessed with the following: “how stressful do you findnot having control over whether or not you will be calledduring an on-call duty?” (1 = not at all stressful, 10 = extremelystressful).

2.2.3. Fatigue. Fatigue was assessed with a shortened version(four items) of the Fatigue Assessment Scale (FAS) [35].An example item is “I suffer from fatigue.” Answers wereprovided on a five-point Likert scale (1 = almost never; 2 =sometimes; 3 = regularly; 4 = often; 5 = almost always).Cronbachs’ 𝛼 for this scale was 0.84.

2.2.4. Work-Home Interference (WHI). Work-home interfer-ence (WHI) was measured with a shortened version ofthe negative work-home interference scale of the SWINGquestionnaire [36]. A distinction was made between strain-based WHI (3 items; e.g., “how often does it occur that youare irritable at home because your work is demanding?”; 𝛼 =0.83) and time-basedWHI (3 items; e.g., “how frequently doesit occur that your work schedule makes it difficult for you tofulfill your domestic obligations?”; 𝛼 = 0.78). Responses wereprovided on a four-point Likert scale (1 = almost never; 2 =sometimes; 3 = often; 4 = almost always).

2.2.5. Perceived On-Call Performance Difficulties (PPD). PPDwere assessed with four items which were based on theDutch scale for measuring experienced load (Schaal ErvarenBelasting, [29]). An example item is “how much effort doesit cost you to complete your tasks without errors when youare called to work during on-call duty?” Respondents had toindicate their answer on a scale from 1 (= no effort at all) to10 (= extremely much effort). Cronbachs’ 𝛼 for this scale was0.94. It should be noted that fatigue andWHIwere assessed asgeneral items (i.e., not in relation to on-call duties), whereasPPD were on-call duty specific.

2.2.6. Control Variables. In order to avoid potential con-founding, we included respondents’ gender (0 = male, 1 =female), age (years), children in the household (0 = no, 1 = yes),and cohabiting or marital status (0 = single, 1 = cohabitingor married) as control variables in the questionnaire. Fur-thermore, to examine whether the on-call duty variables canexplain a significant amount of variance in fatigue, WHI, andPPDover and above general job characteristics, we controlledfor three important job characteristics [30]. Job demandswere assessed with four items from the Questionnaire onthe Experience and Assessment of Work (VBBA, [37]). Anexample item is “do you have to work extra hard?” Responseswere provided on a four-point Likert scale (1 = almost never;4 = almost always; 𝛼 = 0.84). Autonomy was assessed withthree items based on the Job Content Questionnaire [38, 39].


Table 1: Descriptive statistics of the main variables (𝑛 = 157).

Range Min Max Mean SD Percentiles25% 50% 75%

On-call duty exposureNumber of hours on-call per month 1.00 336.00 69.31 71.97 12.00 48.00 100.00Number of active on-call hours per month 0.00 180.00 13.78 22.90 1.00 6.00 16.00Average number of calls per duty 0.00 30.00 1.74 3.49 0.50 1.00 2.00

Experience of being on-callOn-call relaxation 1–10 1.00 10.00 6.94 2.46 5.00 7.00 9.00On-call stress 1–10 1.00 10.00 3.95 2.57 1.50 3.00 6.00On-call work demands 1–10 1.00 10.00 4.78 2.60 3.00 6.00 8.00On-call restrictions 1–10 1.00 10.00 5.86 2.78 2.00 5.00 7.00Satisfaction with compensation for on-call work 1–10 1.00 10.00 5.76 2.83 3.50 6.00 8.00Strain-based WHI 1–4 1.00 3.67 1.74 0.61Time-based WHI 1–4 1.00 3.67 1.75 0.59Fatigue 1–5 1.00 5.00 2.15 0.82Perceived on-call performance difficulties 1–10 1.00 8.00 2.43 1.66

Control variablesAge 23.00 69.00 45.01 11.07Compensationa 0-1 0.00 1.00 0.77 0.42Job demands 1–4 1.25 4.00 2.57 0.58Autonomy 1–3 1.00 3.00 2.46 0.59Social support 1–4 1.00 4.00 3.15 0.64

a0 = no compensation, 1 = compensation.

An example item is “can you decide how your work isexecuted on your own?” Answers were provided on a three-point Likert scale (1 = yes, regularly, 2 = yes, sometimes,and 3 = no). The average score was mirrored for the easeof interpretation, so that higher scores indicate a higherdegree of autonomy. Cronbachs’ 𝛼 for this scale was 0.80.Social support by the supervisor was assessed with three itemsbased on the TNOWork Situation Survey (TAS, [40, 41]). Anexample item is “my supervisor has an eye for the well-beingof his/her employees.” Respondents could indicate the extentto which they agree with the items on a Likert scale from 1 (=totally disagree) to 4 (= totally agree). Cronbachs’ 𝛼 for thisscale was 0.87.

Finally, in the analyses concerning the experiences ofbeing on-call which included employees’ satisfaction withthe compensation they receive for their on-call duties, wecontrolled for compensation for on-call duties. This variableincluded compensation in money, extra free time, or bothand was measured with the following item: “do you receivecompensation for your on-call duties?” (1=no, 2 = yes, forthe actual working hours, and 3 = yes, for the entire on-call duty). Employees who did receive compensation werealso asked whether they received compensation in extrafree time, money, or both. Preliminary analyses (the resultsof these analyses can be requested from the first author)revealed no significant differences between the different typesof compensation with regard to fatigue, WHI, and PPD. Assuch, the variable compensation for on-call dutieswas dummycoded (0 = no compensation, 1 = compensation).

3. Results

3.1. Descriptive Statistics of On-Call Duty Exposure and theExperience of Being On-Call. As can be seen in Table 1,variance in on-call duty exposure was large. On average,employees were 69.31 hours on-call per month (SD = 71.97)with a range from 1 to 336. Half of the respondents spent 48hours or less on-call per month and only 7.6% were on-callmore than 168 hours (i.e., more than one week per month).The mean number of monthly active on-call hours was 13.78(SD = 22.90) hours with a range from 0 to 180. Nearly one-quarter (24.8%) of the employees indicated that, during anaverage on-call duty, they are not called to work and 45.9%indicated that, on average, they are called once.

A majority of the participants (57.5%) felt at least some-what restricted during their on-call duties (score 6 or higher).Nearly one-third (30.6%) of the participants experienced“not having control over whether or not they will be calledto work during their on-call duties” as at least somewhatstressful (score 6 or higher) and 26.1% found it at leastsomewhat difficult to relax during on-call duties (score 5 orlower). More than a third (43.3%) experienced the work theyhave to do when called as (reasonably) taxing (score 6 orhigher). With regard to satisfaction with compensation foron-call work, 45.2% scored 5 or lower, indicating that theywere at least somewhat dissatisfied. The descriptive statisticsand percentiles of the experience variables are presented inTable 1.

Correlations among all variables under examinationare displayed in Table 2. Except for a significant negative


Table2:Correlations

amon

gthev

ariables

undere

xamination(𝑛=157).

12

34

56

78

910

1112

1314

1516

1718

19(1)#

ofon

-callh

oursperm

onth

(2)#

ofactiv

eon-callho

ursp

ermon

th.33∗∗

(3)#

ofcalls

perd

uty

.21∗∗

.21∗∗

(4)O

n-callrelaxatio

n.01

.05−.06

(5)O

n-callstr

ess

−.00

.05

.13−.46∗∗

(6)O

n-callworkdemands

.06

.07

.13−.17∗

.43∗∗

(7)O

n-callrestric

tions

.07−.01

.16−.29∗∗

.50∗∗

.27∗∗

(8)S

atisfactio

nwith

compensation

−.13−.05−.18∗

.16∗

−.23∗∗

−.18∗

−.21∗∗

(9)F

atigue

−.09−.10

.05−.21∗∗

.26∗∗

.01

.24∗∗

−.09

(10)S

train-basedWHI

−.09−.04

.08−.17∗

.29∗∗

.02

.24∗∗

−.11

.69∗∗

(11)T

ime-basedWHI

−.04

.03

.15−.28∗∗

.41∗∗

.16∗

.37∗∗

−.22∗∗

.41∗∗

.51∗∗

(12)P

PD−.02

.01

.01−.22∗∗

.30∗∗

.33∗∗

.10−.25∗∗

.24∗∗

.25∗∗

.30∗∗

(13)A

ge−.07

.09−.05

.17∗

−.15−.05−.15

.06−.10−.03−.16∗

−.11

(14)G

endera

−.05−.18∗

.08−.13

.10.03

.20∗

.02

.05−.08−.02−.12−.21∗∗

(15)M

arita

lstatusb

−.01−.04

.01−.09−.01

.05−.02−.02−.05

.02−.01−.03−.05

.09

(16)C

hildrenin

householdc

.02

.07

.03−.05

.00

.04

.11.14

.06

.06

.10.02−.03−.04

.29∗∗

(17)C

ompensationd

.09−.1

.12−.11−.01

.08

.11.14

.14−.05−.11−.05

.02−.02

.09

.07

(18)Job

demands

−.07−.12

.17∗

−.21∗∗

.21∗∗

.18∗

.10−.05

.21∗∗

.30∗∗

.33∗∗

.21∗∗

.05

.09−.00−.02

.00

(19)A

uton

omy

.10−.07−.03

.22∗∗

−.27∗∗

−.12−.23∗∗

.26∗∗

−.13−.04−.25∗∗

−.15

.23∗∗

−.25∗∗

.11.08

.04−.13

(20)S

ocialSup

port

.15.00

.03

.11−.16−.24∗∗

−.24∗∗

.22∗∗

−.22∗∗

−.22∗∗

−.22∗∗

−.21∗−.01−.01

.14.03

.04−.20∗

.30∗∗

∗

𝑝<.05.∗∗

𝑝<.01.

a 0=male,1=

female.

b 0=sin

gle,1=

cohabitin

gor

marrie

d.c 0

=no

child

renlivingin

theh

ouseho

ld,1

=child

(ren)livingin

theh

ouseho

ld.

d 0=no

compensationforo

n-calldu

ties,1=

compensationforo

n-calldu

ties.


correlation between the number of calls per duty and sat-isfaction with compensation for on-call duties, there wereno significant correlations between on-call exposure andthe experience of being on-call. Also, the on-call exposurevariables showed no significant relationship with any ofthe “outcome” variables. Experienced on-call work demandsshowed a significant positive correlation with PPD and time-based WHI; the experience of restrictions showed a signif-icant positive correlation with fatigue and both time- andstrain-basedWHI.The experience of stress due to the lack ofcontrol correlated positively with all dependent variables, andthe ability to relax during on-call duties correlated negativelywith all four dependent variables. Lastly, satisfaction withcompensation was significantly negatively correlated withstrain-based WHI and perceived performance difficulties.Preliminary multivariate analyses revealed that the controlvariables gender, marital status, and children in the house-hold were not related to the outcome variables (all 𝑝 > 0.05);therefore these variableswere excluded from further analyses.The descriptive statistics of the included control variables canbe found in Table 1.

3.2. On-Call Work Exposure in relation to Fatigue, WHI,and PPD. In order to test the first hypothesis, four two-step hierarchicalmultiple regression analyses were conducted(i.e., one for each of the dependent variables: fatigue, strain-based WHI, time-based WHI, and PPD). The control vari-ables (i.e., age, job demands, autonomy, and social support)were entered in step one of the regression and the exposurevariables (number of hours on-call per month, number ofactive on-call hours per month, and average number of callsper duty) were entered as predictors in step two.

The results showed that the number of hours on-callper month, the number of active on-call hours per month,and average number of calls per duty did not explain anadditional, significant proportion of variance in either fatigue(Δ𝑅2 = 0.01, 𝐹(3, 145) = 0.38, and 𝑝 = 0.770), perceived on-call performance difficulties (Δ𝑅2 = 0.00, 𝐹(3, 145) = 0.15,and 𝑝 = 0.931), strain-based WHI (Δ𝑅2 = 0.01, 𝐹(3, 145) =0.32, and 𝑝 = 0.813), or time-based WHI (Δ𝑅2 = 0.01,𝐹(3, 145) = 0.68, and 𝑝 = 0.567) over and above the controlvariables. As such, Hypothesis 1 was not supported.

3.3. Experience of Being On-Call in relation to Fatigue,WHI, and PPD. Hypothesis 2 was tested with four two-stephierarchicalmultiple regression analyses.The four dependentvariables were fatigue, strain-based WHI, time-based WHI,and PPD. The control variables (i.e., age, compensation foron-call duties, job demands, autonomy, and social support)were entered in step one of the regression; the experiencevariables (i.e., on-call stress due to unpredictability, on-call relaxation, on-call restrictions, on-call work demands,and satisfaction with compensation for on-call duties) wereentered as predictors in step two.

3.3.1. Fatigue. The first analysis revealed that the controlvariables (model 1) explained 11.4% of the variation infatigue and that the experience of being on-call (model 2)

Table 3: Summary of hierarchical regression predicting fatigue fromthe experience of being on-call.

Variable Model 1 Model 2𝐵 SE 𝐵 𝛽 𝐵 SE 𝐵 𝛽

Age −.01 .01 −.11 −.01 .01 −.08Compensation foron-call duties .30 .15 .15 .30 .15 .16∗

Job demands .25 .11 .18∗ .22 .11 .16∗

Autonomy .02 .12 .02 .02 .12 .02Social support −.24 .11 −.18∗ −.24 .11 −.18∗

On-call relaxation −.01 .03 −.04On-call stress .07 .03 .21∗

On-call workdemands −.06 .03 −.20∗

On-call restrictions .02 .03 .09Satisfaction withcompensation foron-call duties

−.01 .02 −.03

𝑅2 .114 .182Δ𝑅2 .114 .068𝐹 for change in 𝑅2 3.797∗∗ 2.361∗∗

𝑝 < .05. ∗∗𝑝 < .01.

explained an additional 6.8%. This change in 𝑅2 was signif-icant (𝐹(5, 142) = 2.36, 𝑝 = 0.043). As can be seen inTable 3, the level of stress experienced during on-call dutiesdue to the lack of control was positively related to fatigue.Contrary to our expectations, the level of experienced workdemands was negatively related to fatigue, and the experienceof restrictions, the ability to relax during on-call duties, andthe satisfactionwith compensation for on-call duties were notsignificantly related to fatigue. As such, Hypothesis (2a) wasonly partly supported.

3.3.2. Work-Home Interference. With regard to strain-basedWHI, the control variables (model 1) explained 12.4% ofthe variation and the experience of being on-call (model 2)explained an additional 9.0%. This change in 𝑅2 was signifi-cant (𝐹(5, 142) = 3.24, 𝑝 = 0.008). As can be seen in Table 4,stress experienced during on-call duties due to the lack ofcontrol was positively related to strain-basedWHI. Contraryto our expectations, satisfaction with compensation for on-call duties, on-call restrictions, and on-call relaxation werenot significantly related to this outcome measure, and thelevel of on-call work demandswas negatively related to strain-based WHI. Consequently, Hypothesis (2b) was only partlyconfirmed.

With regard to time-based WHI, the control variables(model 1) explained 19.7% of the variation, and the experienceof being on-call (model 2) explained an additional 12.2%.Thischange in 𝑅2 was significant (𝐹(5, 142) = 5.07, 𝑝 < 0.001).As can be seen in Table 5, only the experiences of stress andrestrictions during on-call duties were significantly positivelyrelated to time-based WHI. The experienced level of on-callwork demand, the ability to relax during on-call duties, and


Table 4: Summary of hierarchical regression predicting strain-based WHI from the experience of being on-call.


Age −.00 .00 −.06 .00 .00 −.02Compensation foron-call duties −.06 .11 −.04 −.05 .11 −.04

Job demands .29 .08 .28∗∗ .27 .08 .25∗

Autonomy .06 .09 .06 .13 .09 .13Social support −.18 .08 −.19∗ −.17 .08 −.18∗

On-call relaxation .00 .02 −.01On-call stress .06 .02 .25∗

On-call workdemands −.05 .02 −.20∗

On-call restrictions .03 .02 .12Satisfaction withcompensation foron-call duties

−.01 .02 −.03

𝑅2 .124 .214Δ𝑅2 .124 .090𝐹 for change in 𝑅2 4.159∗∗ 3.241∗∗∗∗

𝑝 < .05. ∗∗𝑝 < .01. ∗∗∗𝑝 < .001.

Table 5: Summary of hierarchical regression predicting time-basedWHI from the experience of being on-call.


Age −.01 .00 −.15 −.01 .00 −.09Compensation foron-call duties −.13 .10 −.10 −.16 .10 −.11

Job demands .31 .08 .30∗∗∗ .26 .07 .25∗∗

Autonomy −.14 .08 −.14 −.05 .08 −.05Social support −.12 .07 −.13 −.07 .07 −.07On-call relaxation −.01 .02 −.05On-call stress .05 .02 .20∗

On-call workdemands −.01 .02 −.06

On-call restrictions .04 .02 .20∗

Satisfaction withcompensation foron-call duties

−.02 .02 −.07

𝑅2 .197 .319Δ𝑅2 .197 .122𝐹 for change in 𝑅2 7.213∗∗∗ 5.070∗∗∗∗

𝑝 < .05. ∗∗𝑝 < .01. ∗∗∗𝑝 < .001.

the satisfactionwith compensation for on-call duties were notsignificantly related to time-basedWHI. As such, Hypothesis(2c) was partly confirmed as well.

3.3.3. Perceived On-Call Performance Difficulties. The controlvariables (model 1) accounted for 9.4% of the variation in

Table 6: Summary of hierarchical regression predicting PPD fromthe experience of being on-call.


Age −.02 .01 −.10 −.01 .01 −.08Compensation foron-call duties −.16 .31 −.04 −.14 .30 −.03

Job demands .52 .23 .18∗ .33 .23 .12Autonomy −.16 .24 −.06 .02 .23 .01Social support −.43 .22 −.16∗ −.28 .21 −.11On-call relaxation −.06 .06 −.08On-call stress .09 .07 .14On-call workdemands .14 .05 .22∗

On-call restrictions −.08 .05 −.13Satisfaction withcompensation foron-call duties

−.09 .05 −.16

𝑅2 .094 .214Δ𝑅2 .094 .120𝐹 for change in 𝑅2 3.034∗ 4.344∗∗∗

𝑝 < .05. ∗∗𝑝 < .001.

PPD; the experience of being on-call (model 2) accounted foran additional 12.0% of the variation. This change in 𝑅2 wassignificant (𝐹(5, 142) = 4.34, 𝑝 = 0.001). As can be seenin Table 6, only the level of on-call work demands waspositively related to PPD. Satisfaction with compensationfor on-call duties showed a marginal negative relationshipwith PPD. The experience of restriction, the experience ofstress, and the ability to relax during on-call work were notsignificantly related to PPD. As such, Hypothesis (2d) waspartly supported.

4. Discussion

4.1. Discussion of the Results. Off-site on-call duties are aninteresting yet relatively understudied working time arrange-ment [1, 6].Therefore, we aimed to gain more insight into therelationship between the exposure to off-site on-call dutiesand the experience of being on-call on the one hand andfatigue, strain-based and time-based WHI, and PPD on theother hand.

In the present study sample, there was a large variationin the amount of exposure to on-call work. Contrary to ourfirst hypothesis, differences in exposure to on-call work werenot systematically related to any of the outcome variables.This is not in line with previous research conducted amongphysicians that found frequent on-call duties to be related todistress and turnover intentions [7]. However, physicians’ on-call duties mostly take place during the night, and night shiftshave been shown to be negatively related to employees’ well-being and health, possibly due to sleep deprivation [42]. Thismight explain why the frequency of on-call duties has beenfound to have negative consequences in previous research


but not in the present study, where 93% of the employeesindicated that their on-call duties took place during the day aswell. Another explanation for the nonsignificant results maylie in our exposure measures. Due to the heterogeneity of ourstudy sample, we did not have access to individual on-callschedules and we tried to get insight into the participants’average exposure to on-call duties by means of multipleitems on average exposure. However, when schedules varya lot, it may not be easy for each employee to indicatethe average number of on-call duties a month. Also, theaverage time spent working during one duty and the averagenumber of calls may show large intraindividual variation.In order to be able to draw clear conclusions about therelationship between off-site on-call duty exposure, fatigue,WHI, and PPD, further research with larger samples thatinclude different professions and different types of on-callduties (e.g., with regard to the length and timing) is needed.In addition to self-reports, company registered data of on-call work exposure should be used. Furthermore, a repeatedmeasurements diary study in which participants keep trackof their on-call hours and momentary experiences mightprovide more insight into the relationship between the actualon-call duty exposure and fatigue, work-home interference,and performance. Multiwave diary designs would also allowdisentangling the effects of different types of on-call shifts(e.g., night shifts and day shifts).

The second hypothesis was partly confirmed. Manyemployees experienced their on-call duties as (somewhat)unfavourable. The experience of being on-call in turn wasrelated to fatigue, strain-based and time-based WHI, andon-call performance difficulties, but not all experiencescontributed significantly to the prediction. All in all, on-callstress (i.e., the experience of stress due to the unpredictabilityof being on-call) seemed to be the most important predictoras it was positively related to all “outcome” variables exceptfor performance difficulties. Employees’ satisfaction with thecompensation they received for their on-call duties and theirability to relax during inactive on-call work periods were notrelated to either fatigue, WHI, or PPD when controlling forimportant job characteristics. Feeling restricted during on-call work was related to the most proximal criterion, that is,time-basedWHI, but did not explain any additional variancein strain-basedWHI, PPD, or general fatigue over and abovethe control variables. Contrary to what we expected basedon the effort-recovery model [15], on-call work demandswere negatively related to fatigue and strain-based WHI.This result is theoretically implausible. In post hoc analyses(the results of these analyses can be requested from thefirst author) with (i) fatigue and (ii) strain-based WHI asdependent variables, we compared what happens when on-call work demand is the only predictor besides the controlvariables towhat happenswhen on-call work stress is the onlypredictor. For both dependent variables, the effect sizes of on-call demands and on-call stress are similar, but whereas thesignificant effects of on-call demands only appear when thisvariable is entered into the analysis in combination with theother experiences, this is not the case for on-call stress (whichis a significant predictor when entered alone as well as withthe other predictors). In other words, the effects of on-call

stress remain significant when it is the only predictor besidesthe control variables. Furthermore, the Pearson correlationsbetween on-call work demands and both dependent variableswere not significant. Hence, we interpret the significantresults for on-call demands as an artefact. Further research isneeded to examine the role of on-call work demands in moredetail.

4.2. Strengths and Limitations. The limited previous studieson on-call duties focused on only one profession (mostlymedical staff), thereby limiting the external validity of theresults [1]. Therefore, one asset of the present study is thesample that consisted of a heterogeneous group of employeeswith different professions and from different organizations.Also, previous studies havemainly focused on on-site standbyduties, so another asset is the focus on off-site on-callduties which have been largely neglected so far. Furthermore,to the authors’ knowledge, the present study was the firstto investigate both exposure to on-call work duties andtheir psychological significance (in terms of experiences) inrelation to general “outcomes” such as fatigue and work-home interference and an on-call duty specific performanceindicator, that is, perceived difficulties to perform well whencalled to work. Moreover, our analyses were quite strict. Wecontrolled for important general job characteristics, therebyminimizing the possibility of confounding effects.

Nonetheless, several limitations of the present studywarrant further research. First, the validity of the on-call dutyexposure items is not without problems. Individual scoreswere highly variable and not always plausible (e.g., a smallpercentage reported considerably more on-call hours thanthe legal maximum or more active on-call hours than totalon-call hours a month). Respondents producing such errorswere excluded from the analyses, but it is possible that ourself-developed items were not clear to some participants. Asmentioned above, less valid measures might be a reason forthe null-findings on on-call exposure. Future research amongheterogeneous samples may benefit from more valid mea-sures of on-call duty exposure (e.g., official work schedules).

Second, the present study was cross-sectional, so nocausal relations can be implied. With regard to employees’experience of being on-call, bidirectionality of effects mightbe plausible. For instance, general fatigue and work-homeinterference may affect employees’ experience of being on-call. Further research with longitudinal or experimentaldesigns is needed to investigate the causal direction of theassociations.

4.3. Implications. Since on-call duties are officially consid-ered rest time and take place during time meant for recovery[13], the results of the present study are reason for someconcern. About one-third of the employees experienced theiron-call duties as (reasonably) stressful, which may impairtheir recovery. Recovery, however, is critical for employees’well-being and health [15, 16]. Working time legislationonly takes the length and frequency of on-call duties intoaccount, but the results of the present study suggest thatthe psychological aspect (i.e., employees’ experience) of on-call duties may be more important than the number of


hours spent on-call and that even short or infrequent on-callperiods may interfere with employees’ well-being, therebypresenting a risk for ill health. Employers should thereforepay attention to how employees experience their on-callduties and lighten the on-call burden of employees prone tosuffer from stress in order to prevent negative consequencessuch as fatigue andWHI, which may, in the long run, lead tohealth problems.

5. Conclusions

In sum, the present study showed that employees’ experienceof their on-call duties is related to their general fatigue,work-home interference, and the difficulties they have inperforming well when called to work, even when controllingfor important job characteristics. Our results suggest that itis the experience of being on-call rather than the variationin exposure to on-call duties itself that is associated withnegative outcomes. This means that even a low amount of(active) on-call hours a month and even a low frequency ofbeing called to work can be related to an increase in fatigueand work-home interference, when employees experiencebeing on-call negatively. Therefore, employees’ experience ofon-call duties should be included in future studies on on-callwork. In addition, future research should be conducted togain more insight into potential predictors of the experienceof being on-call (e.g., individual characteristics) and to inves-tigate how the experience of being on-call can be improved,in order to form a basis from which to develop successfulinterventions that decrease the negative consequences ofbeing on-call.



References

[1] A.-M. Nicol and J. S. Botterill, “On-call work and health: areview,” Environmental Health: A Global Access Science Source,vol. 3, no. 1, article 15, 2004.

[2] M. Schult and S. Tobsch, “Leisure stress: when work is con-stantly calling,” in SOEP Papers on Multidisciplinary Panel DataResearch, German Socio-Economic Panel Study (SOEP), Ed.,DIW Berlin, Berlin, Germany, 2012.

[3] J. T. Arnedt, J. Owens,M.Crouch, J. Stahl, andM.A. Carskadon,“Neurobehavioral performance of residents after heavy nightcall vs after alcohol ingestion,” The Journal of the AmericanMedical Association, vol. 294, no. 9, pp. 1025–1033, 2005.

[4] B. M. Emmett, S. M. Dovey, and B. J. Wheeler, “After-hours on-call: the effect on paediatricians’ spouses and families,” Journalof Paediatrics and Child Health, vol. 49, no. 3, pp. 246–250, 2013.

[5] Y. Berger, “Standby periods,”Australian Safety News, vol. 63, pp.3–10, 1999.

[6] E. Bamberg, J. Dettmers,H. Funck, B. Krahe, andT.Vahle-Hinz,“Effects of on-call work on well-being: results of a daily survey,”Applied Psychology: Health andWell-Being, vol. 4, no. 3, pp. 299–320, 2012.

[7] T. Heponiemi, A. Kouvonen, J. Vanska et al., “Effects of activeon-call hours on physicians’ turnover intentions and well-being,” Scandinavian Journal of Work, Environment and Health,vol. 34, no. 5, pp. 356–363, 2008.

[8] P. Tucker, M. Brown, A. Dahlgren et al., “The impact of juniordoctors’ worktime arrangements on their fatigue and well-being,” Scandinavian Journal of Work, Environment and Health,vol. 36, no. 6, pp. 458–465, 2010.

[9] S. Sonnentag, “Psychological detachment from work duringleisure time: the benefits of mentally disengaging from work,”CurrentDirections in Psychological Science, vol. 21, no. 2, pp. 114–118, 2012.

[10] M. L.M. vanHooff, S. A. E. Geurts, D. G. J. Beckers, andM. A. J.Kompier, “Daily recovery fromwork: the role of activities, effortand pleasure,”Work and Stress, vol. 25, no. 1, pp. 55–74, 2011.

[11] E. J. Mojza, C. Lorenz, S. Sonnentag, and C. Binnewies, “Dailyrecovery experiences: the role of volunteer work during leisuretime,” Journal of Occupational Health Psychology, vol. 15, no. 1,pp. 60–74, 2010.

[12] D. G. J. Beckers, D. van der Linden, P. G. W. Smulders, M. A. J.Kompier, T. W. Taris, and S. A. E. Geurts, “Voluntary or invol-untary? Control over overtime and rewards for overtime inrelation to fatigue and work satisfaction,” Work and Stress, vol.22, no. 1, pp. 33–50, 2008.

[13] Ministry of Social Affairs and Employment,The working HoursAct. Information for Employers and Employees, Ministry ofSocial Affairs and Employment, 2010.

[14] S. A. E. Geurts and S. Sonnentag, “Recovery as an explanatorymechanism in the relation between acute stress reactions andchronic health impairment,” Scandinavian Journal of Work,Environment and Health, vol. 32, no. 6, pp. 482–492, 2006.

[15] T. F. Meijman and G. Mulder, “Psychological aspects of work-load,” inHandbook of Work and Organizational Psychology, P. J.D. Drenth, H. Thierry, and C. J. de Wolff, Eds., pp. 5–33, Psy-chlogy Press, Hove, UK, 1998.

[16] B. M. Wen, “Stress, adaptation, and disease allostasis andallostatic load,”Annals of the New York Academy of Sciences, vol.840, pp. 33–44, 1998.

[17] M. Kivimaki, P. Leino-Arjas, L. Kaila-Kangas et al., “Is incom-plete recovery fromwork a riskmarker of cardiovascular death?Prospective evidence from industrial employees,” Psychoso-matic Medicine, vol. 68, no. 3, pp. 402–407, 2006.

[18] J. K. Sluiter,M.H.W. Frings-Dresen, A. J. van der Beek, and T. F.Meijman, “The relation betweenwork-induced neuroendocrinereactivity and recovery, subjective need for recovery, and healthstatus,” Journal of Psychosomatic Research, vol. 50, no. 1, pp. 29–37, 2001.

[19] S. Sonnentag, I. Kuttler, and C. Fritz, “Job stressors, emotionalexhaustion, and need for recovery: a multi-source study onthe benefits of psychological detachment,” Journal of VocationalBehavior, vol. 76, no. 3, pp. 355–365, 2010.

[20] S. Sonnentag and U.-V. Bayer, “Switching off mentally: predic-tors and consequences of psychological detachment from workduring off-job time,” Journal of Occupational Health Psychology,vol. 10, no. 4, pp. 393–414, 2005.

[21] S. Sonnentag, “Recovery from fatigue: the role of psychologicaldetachment,” inCognitive Fatigue:Multidisciplinary Perspectiveson Current Research and Future Applications, P. L. Ackerman,Ed., pp. 253–272, American Psychological Association, Wash-ington, DC, USA, 2011.

[22] T. P. Grantcharov, L. Bardram, P. Funch-Jensen, and J. Rosen-berg, “Laparoscopic performance after one night on call in


a surgical department: prospective study,” British Medical Jour-nal, vol. 323, no. 7323, pp. 1222–1223, 2001.

[23] A. D. Saxena and C. F. P. George, “Sleep andmotor performancein on-call internal medicine residents,” Sleep, vol. 28, no. 11, pp.1386–1391, 2005.

[24] S. Veasey, R. Rosen, B. Barzansky, I. Rosen, and J. Owens, “Sleeploss and fatigue in residency training: a reappraisal,”The Journalof the American Medical Association, vol. 288, no. 9, pp. 1116–1124, 2002.

[25] N. Lamond and D. Dawson, “Quantifying the performanceimpairment associated with fatigue,” Journal of Sleep Research,vol. 8, no. 4, pp. 255–262, 1999.

[26] D. Dawson and K. Reid, “Fatigue, alcohol and performanceimpairment,” Nature, vol. 388, no. 6639, p. 235, 1997.

[27] M. L. M. vanHooff, S. A. E. Geurts, T.W. Taris et al., “Disentan-gling the causal relationships between work-home interferenceand employee health,” Scandinavian Journal of Work, Environ-ment & Health, vol. 31, no. 1, pp. 15–29, 2005.

[28] J. E. McGrath, “Stress and behavior in organizations,” in Hand-book of Industrial and Organizational Psychology, M. D. Dun-nett, Ed., pp. 1351–1395, Rand McNally College Publishing,Chicago, Ill, USA, 1976.

[29] T. F.Meijman,Over vermoeidheid: Arbeidspsychologische studiesnaar de beleving van belastingseffecten [Ph.D. thesis], Universityof Groningen, Groningen, The Netherlands, 1991, Studiecen-trum Arbeid en Gezondheid University of Amsterdam.

[30] J. Siegrist, “Adverse health effects of high-effort/low-rewardconditions,” Journal of Occupational Health Psychology, vol. 1,no. 1, pp. 27–41, 1996.

[31] F. R. H. Zijlstra and S. Sonnentag, “After work is done: psycho-logical perspectives on recovery from work,” European Journalof Work and Organizational Psychology, vol. 15, no. 2, pp. 129–138, 2006.

[32] S. Sonnentag, “Work, recovery activities, and individual well-being: a diary study,” International Journal of Psychology, vol. 35,no. 3-4, p. 399, 2000.

[33] L. L. J. Koppes, E. M. M. de Vroome, M. E. M. Mol, B. J. M.Janssen, and S. N. J. van den Bossche, Netherlands WorkingConditions Survey 2010: Methodology and Overall Results, TNO,Hoofddorp, the Netherlands, 2011.

[34] J. de Bloom, S. A. E. Geurts, T. W. Taris, S. Sonnentag, C. deWeerth, and M. A. J. Kompier, “Effects of vacation from workon health and well-being: lots of fun, quickly gone,” Work andStress, vol. 24, no. 2, pp. 196–216, 2010.

[35] J. de Vries, H.Michielsen, G. L. vanHeck, andM.Drent, “Meas-uring fatigue in sarcoidosis: the fatigue assessment scale (fas),”British Journal of Health Psychology, vol. 9, no. 3, pp. 279–291,2004.

[36] S. A. E. Geurts, T.W. Taris, M. A. J. Kompier, J. S. E. Dikkers, M.L. M. van Hooff, and U.M. Kinnunen, “Work-home interactionfrom a work psychological perspective: development and vali-dation of a new questionnaire, the swing,”Work and Stress, vol.19, no. 4, pp. 319–339, 2005.

[37] M. Van Veldhoven and T. Meijman, Measurement of Psychoso-cial Job Demands with a Questionnaire: The QuestionnaireExperience and Evaluation of Work (VBBA), NIA, Amsterdam,The Netherlands, 1994.

[38] R. A. Karasek, C. Brisson, N. Kawakami, I. Houtman, P.Bongers, and B. Amick, “The Job Content Questionnaire (JCQ):an instrument for internationally comparative assessments ofpsychosocial job characteristics,” Journal of OccupationalHealthPsychology, vol. 3, no. 4, pp. 322–355, 1998.

[39] R. A. Karasek, Job Content Questionnaire and User's Guide(Revision 1.1), University of Massachusetts Lowell, Job ContentQuestionnaire (JCQ) Center, Lowell, Mass, USA, 1985.

[40] P. Smulders, F. Andries, and F. Otten, How Do Dutch PeopleThink about Their Work...? Design, Quality and First Results ofthe Tno Working Situation Survey, TNO Arbeid, Hoofddorp,The Netherlands, 2001.

[41] P. Smulders, Dutch People about Their Work: Methodology andTables of the TNO Working Situation Surveys 2000 and 2002,TNO Arbeid, Hoofddorp, the Netherlands, 2003.

[42] T. Kantermann, M. Juda, C. Vetter, and T. Roenneberg, “Shift-work research: where do we stand, where should we go?” Sleepand Biological Rhythms, vol. 8, no. 2, pp. 95–105, 2010.

Research ArticlePsychosocial Work Factors and MusculoskeletalPain: A Cross-Sectional Study among Swedish FlightBaggage Handlers

Eva L. Bergsten,1,2 S. E. Mathiassen,1 and E. Vingård2

1Centre for Musculoskeletal Research, Department of Occupational and Public Health Sciences, University of Gavle,801 76 Gavle, Sweden2Occupational and Environmental Medicine, Department of Medical Sciences, Uppsala University, 751 85 Uppsala, Sweden

Correspondence should be addressed to Eva L. Bergsten; [email protected]

Received 16 January 2015; Revised 15 April 2015; Accepted 27 April 2015


Copyright © 2015 Eva L. Bergsten et al.This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Objective. Flight baggage handlers sort and load luggage to airplanes. This study aimed at investigating associations betweenpsychosocial exposures and low back and shoulder musculoskeletal disorders (MSDs) among Swedish flight baggage handlers.Methods. A questionnaire addressing MSDs (Standardized Nordic Questionnaire) and psychosocial factors (CopenhagenPsychosocial Questionnaire, COPSOQ) was answered by 525 baggage handlers in six Swedish airports. Results. Low back (LBP)and shoulder pain (SP) were reported by 70% and 60%, respectively. Pain was reported to interfere with work (PIW) by 30% (lowback) and 18% (shoulders), and intense pain (PINT) occurred in 34% and 28% of the population.Quality of leadershipwas the mostdissatisfying psychosocial factor, while the most positive was social community at work. Low ratings in the combined domainWorkorganization and job content were significantly associated with PIW in both low back and shoulders (Adjusted Hazard Ratios 3.65(95% CI 1.67–7.99) and 2.68 (1.09–6.61)) while lower ratings in the domain Interpersonal relations and leadership were associatedwith PIW LBP (HR 2.18 (1.06–4.49)) and PINT LBP and SP (HRs 1.95 (1.05–3.65) and 2.11 (1.08–4.12)). Conclusion. Severity of painamong flight baggage handlers was associated with psychosocial factors at work, suggesting that they may be a relevant target forintervention in this occupation.

1. Background

Worldwide, more than 2000 airlines operated more than23 000 aircrafts in 2006 [1]. These aircrafts made more than28million scheduled departures, carryingmore than 2 billionpassengers. A substantial proportion of these passengerswould be expected to bring baggage that is checked in andthus handled by flight baggage handlers at the airports ofdeparture and arrival. To a major extent, flight baggagehandling services are similar in all larger airports, and soflight baggage handlers perform similar tasks all over theworld.

Workers handling flight baggage are typically engaged inmanual tasks like sorting, loading and unloading baggage,mail and flight cargo to the airplanes, and so-called airportramp service work. The ramp is the area around the aircraft.

In Swedish airports, bags checked in by passengers are placedon a conveyor belt, which transports the bags to a sorting area.In the sorting area, baggage handlers place the bags on a cartor in a Unit Load Device (ULD; i.e., a container that can beloaded on the aircraft), which is eventually transported by atruck to the ramp.There, the bags are loaded into the aircraftbaggage compartment piece by piece, or in one operationfor ULDs. Unloading an aircraft runs in reverse: bags aretransferred from the aircraft compartment to the carts andtransported by trucks to the arrival conveyor belt in thesorting area, which transports the bags to the arrival hall.Sometimes, baggage handlers engage also in communicatingwith air traffic controllers directing air traffic on the groundor engage in towing the aircrafts to and from gates witha pushback vehicle, and they may also serve aircrafts withauxiliary power units, brakes, and light.


http://dx.doi.org/10.1155/2015/798042


Thus, flight baggage handling mainly consists in transfer-ring baggage between carts and aircraft compartments, push-ing and pulling loaded trailers, and stowing bags and freight,often while kneeling or squatting in confined compartments.Thus, baggage handling is associated with several physicalfactors suspected to increase risks for musculoskeletal dis-orders (MSDs), such as heavy manual materials handling,frequent lifting, awkward body postures, and pushing andpulling [2–7].

Scientific reports devoted to the baggage handling occu-pation are rare. One study concluded that the lower back wasexposed to considerable loads when bags of different weightsand destination height were handled in a kneeling posture inthe compartment [8]. Twoother studies investigated opinionsabout suspected causes of MSDs and effective preventionamong baggage handlers and safety professionals [9, 10]. Oneof these studies reported a high prevalence of MSDs in back,knees, and shoulders among the baggage handlers [9].

None of the cited studies on risk factors and disordersamong baggage handlers consider the possible role of thepsychosocial work environment in causing or aggravatingMSDs. It is suggested that psychosocial factors at work areassociated with risks of developing disorders in the lowerback, neck/shoulder region, and upper extremities [10, 11].Evidence for poor social support as a risk factor for lowback pain (LBP) was claimed in one review [11] and in acohort study [12]. Lack of social support was also reportedto be a risk factor for musculoskeletal morbidity, sicknessabsence, restricted activity, and not returning to work [13].High emotional demands, low influence, and pronouncedrole conflicts at work have also been suggested to predictLBP [14]. The influence of psychosocial factors has beenexplained in the context of a biopsychosocial model [15],which emphasizes both mechanical and physiological pro-cesses in the generation and maintenance of pain, as well asthe importance of psychological and social conditions for theresponse to pain and the disability developed by a particularindividual.

Motivated by the lack of literature on psychosocial factorsand their association with back and shoulder MSDs amongflight baggage handlers, the aims of this study were as follows:

(1) To conduct a nationwide Swedish survey of muscu-loskeletal disorders and psychosocial factors in theflight baggage handling occupation.

(2) To determine the extent to which these psychosocialfactors are associated with pain intensity and withpain interfering with work. We hypothesized lessfavourable psychosocial conditions to be associatedwith an increased likelihood of experiencing painamong the flight baggage handlers, as documented inseveral other occupational settings.

This study was part of a two-year work environment projectamong flight baggage handlers in Sweden, conducted during2010–2012 under the auspices of the Vocational Training andWorking Environment Council (TYA), a council formed bythe Swedish aviation industry employers’ association andthe transportation worker union. TYA covers about 60% of

all Swedish flight baggage handlers. The main goal of theoverall project was to document physical and psychosocialwork environment conditions, as a basis for developing inter-ventions to improve health.

2. Methods

2.1. Airports and Baggage Handlers. Sweden has 41 airportswith a total of about 1 400 baggage handlers employed forramp service, either by a handling company or, at smallerairports, directly by the airport. All six handling companiesaffiliated to TYA in the three largest public Swedish airports(Stockholm-Arlanda, Goteborg-Landvetter, Malmo-Sturup)agreed to participate in the study; these companies areresponsible for 75%of the yearly traffic at the three airports. Inaddition, all handlers working in three small private airports(Arvidsjaur, Smaland and Skavsta)were approached. Baggagehandlers working less than half-time were excluded, as wellas handlers on vacation, parental leave, or sick leave bythe time of data collection (December 2010 to April 2011).Altogether, 806 of the about 1400 baggage handlers workingin Sweden were eligible for the study, which was approved bythe regional Ethical Review Board in Uppsala.

2.2. Procedure. All companies and airports were visited by amember of the research team for a sufficient period of timeto ensure that all handlers could be approached in person andinformed about the study.The researchers handed out a ques-tionnaire to each handler, which was, by most respondents,answered in about 25–30 minutes during working hours.Themember of the research team collected questionnaires at thatsame occasion, but the handler also had the choice of sendingthe questionnaire later on to the researchers in a sealedenvelope. Handlers who did not submit their questionnairesin the first place were approached again, in person, by mem-ber(s) of the research team. Completion and submission ofthe questionnaire were enthusiastically encouraged by localteam managers and safety officers, but telephone numbers oraddresses were not available for reminders.

2.3. Questionnaire. The questionnaire contained questionsin six different areas: demographic factors, psychosocialfactors, physical workload in different tasks, musculoskeletaldisorders, general health, and fatigue.

2.4. Demographic Factors. Thedemographic factors includedage, gender, height, weight, and years of experience as abaggage handler.

2.5. Psychosocial Factors. Psychosocial factors at work wereassessed using themedium-length Copenhagen PsychosocialQuestionnaire (COPSOQ) [16] in its latest edition, COP-SOQ II [17]. In the present study, 13 scales with a totalof 42 questions were used. The scales represent two maindomains, that is, Work organization and job content (fivescales) and Interpersonal relations and leadership (eightscales). Questions in six scales (influence at work, variation,commitment to the workplace, social support from colleagues,


Table 1: General health by age group (𝑛 with percent of column totals).

General Health Age (years)All (𝑛 = 525) <34 (𝑛 = 256) 35–49 (𝑛 = 177) >50 (𝑛 = 71) Age missing (𝑛 = 21)

Excellent/very good 265 (50) 163 (64) 74 (42) 20 (28) 8 (38)Good 189 (36) 78 (30) 71 (40) 32 (45) 8 (38)Somewhat bad/bad 66 (13) 10 (4) 32 (18) 19 (27) 5 (24)Health rating missing 5 (1) 5 (2)

social support from supervisors, and social community atwork) were answered using a five-step response ranging from“always” to “never/hardly ever.” In seven scales (possibilitiesfor development, meaning of work, predictability, recognition,role clarity, role conflicts, and quality of leadership) questionswere answered in five steps from “to a large extent” to“to a very small extent.” For all questions, the answer wastransformed into a number between 0 and 100 (i.e., 0, 25,50, 75, and 100) for the five response steps, and an overallscale score was computed as the mean score across questionscontained in each of the 13 scales. A higher score indicateda more positive work environment, except for role conflictwhere a high score indicates more conflicts. For reasons ofcomparability questions included in the different scales werecopied in their original form from the second version of theCOPSOQ questionnaire, as reported by Pejtersen et al. 2010[17] (see the appendix), and the procedure for calculatingscale scores was also adopted from Pejtersen et al.

2.6. Physical Work Load. The perceived physical work loadwas rated for low back and shoulder separately in a number oftasks reported by the handlers to occur frequently in the job,using the question “how do you perceive the physical load intask xx” with answers on a six-grade scale from “not at all” to“to a large extent.” For 31 baggage handlers, the occurrences ofthese tasks were determined from video recordings collectedby amember of the research team for half a work shift. Acrossthese 31 baggage handlers, the mean time proportions of thejob spent in loading/unloading outside, loading inside air-craft compartment, and pushing/pulling carts were 5%, 5%,and 2%, respectively. We used this information to calculate a“physical load index” for each worker, as the average rating ofperceived load in both the low back and the shoulders acrossthese tasks, weighted by their occurrence.

2.7. General Health and Musculoskeletal Disorders. Generalhealth was reported using one question, that is, “In general,how would you rate your health.” The one-year prevalence oflow back pain (LBP), shoulder pain (SP), and pain interferingwith work (PIW) was retrieved using the StandardizedNordic Questionnaire [18, 19]. The intensity of pain (PINT)during the preceding 12 months was reported on a 10-gradescale from “no pain” to “very very high (almost maximal).”

2.8. Data Analysis. Ratings of psychosocial factors in thetwo domainsWork organization and job content (five factors)and Interpersonal relations and leadership (eight factors) wereanalyzed both for each factor separately and after combining

factor ratings within each of the two domains. In this process,ratings of role conflict were reversed. For each of the resulting15 psychosocial variables, scores were divided into populationquartiles and the lower and upper quartile populations wereused for comparisons in logistic regression (see below).

Four dichotomized outcomes were addressed in particu-lar, that is, pain interfering with work (PIW) and high painintensity (PINT) during the preceding year, separately for thelow back (LBP) and for the shoulder (SP). Pain interferingwith work (PIW) was classified as “yes” or “no.” “High painintensity” (PINT)was defined as the subject rating 5 or higheron the pain intensity scale for either low back or shoulder.This definition was based on the finding by Andersen et al.[20] in which subjects reporting a pain intensity of 5 or largerare more at risk of eventually suffering long-term sicknessabsence than those reporting less than 5. A case of “No pain”was registered when the subject reported “no pain” for allbody regions.

For all psychosocial factors, Hazard Ratios (HR) with95% confidence intervals (95% CI) were estimated using Coxproportional hazards regression for the group with upperquartile ratings to have more severe outcomes (PIW andPINT) than the corresponding lower quartile group. HR wasadjusted for age, BMI, general health, and physical work load,while we did not adjust for current fatigue. All analyses weredone in SAS 9.3 (SAS Institute Inc.).

3. Results

Of 806 eligible baggage handlers, 525 (98% males and 2%females) answered the questionnaire, that is, a 65% responserate. General health by age is shown in Table 1. Stratifiedinformation was not available for twenty-six subjects.

The one-year prevalence of pain in the low back andshoulders was 70% and 60%, respectively, among thoseworkers answering to the question on pain (missing 𝑛 = 44for low back and 𝑛 = 53 for shoulders). The different combi-nations of low back and shoulder pain (Table 2) showed thatalmost half (45%) of the total population of handlers reportedto have both low back and shoulder pain. Among workersreporting lowback pain (LBP+/SP+ andLBP+/SP−; 𝑛 = 339),30% (𝑛 = 101) reported low back pain only and 70% (𝑛 =238) also reported shoulder pain (Table 2). Sixteen percentof workers with shoulder pain reported pain in that regiononly (LBP−/SP+; 𝑛 = 47), while 84% had also low back pain(LBP+/SP+; 𝑛 = 238). Of the 339 and 285 workers reportingpain in the low back and shoulders, respectively, 328 and 265proceeded, as intended, to rate whether that pain interfered


Table 2: Number of workers (percent of total study population, 𝑛 = 525) reporting any pain (LBP, SP), pain interfering with work (PIW) andhigh pain intensity (PINT).

Yes (+) LBP+ LBP+ LBP− LBP− Missing answers on pain LBP SP LBP SPNo (−) SP+ SP− SP+ SP− PIW+ PIW+ PINT+ PINT+Number of workers 𝑛 (%) 238 (45) 101 (19) 47 (9) 79 (15) 60 (12) 156 (30) 96 (18) 180 (34) 147 (28)

with work (PIW). Pain intensity (PINT) was rated by moreworkers than those reporting pain; 506 and 508 workersrated intensity for low back and shoulder pain, respectively.However, 101 and 137 of these workers rated the intensityas 0 or 0.5, that is, effectively an absence of pain. Low backpain more often inhibited work than shoulder pain (Table 2),with 46% (𝑛 = 156) of those reporting any pain in the lowback (𝑛 = 339) being inhibited by that pain, while pain wasinhibiting for only 34% (𝑛 = 96) of those reporting shoulderpain (𝑛 = 285).

3.1. Psychosocial Work Factors and Means for the OutcomeGroups. Mean values for the 13 psychosocial factors and forthe combined domains Work organization and job contentand Interpersonal relations and leadership are presented inTable 3 for each of the outcome groups.

In all outcome groups, scores were lowest on qualityof leadership and influence at work and highest on socialcommunity at work. Overall, values were lower, indicatingmore dissatisfaction, for theWork organization domain thanfor the Interpersonal relations domain.

For all psychosocial factors, baggage handlers with nopain reported better psychosocial working conditions thanany of the four pain groups.

3.2. Associations between Psychosocial Factors and Pain.Lower ratings in the domain Interpersonal relations and, inparticular, in the domain Work organization and job contentwere significantly associated with increased occurrence ofpain inhibiting work, PIW, both in the low back and in theshoulder region (Table 5). Several separate psychosocial fac-tors in the two domains contributed to this overall association(Table 5). For intense pain, PINT, only role clarity showeda significant association with LBP, and this contributed toan overall association with pain for ratings in the domainInterpersonal relations (Table 6). For the shoulder, socialcommunity at work was associated with intense pain, andInterpersonal relation was again the only domain showinga significant association (Table 6). Workers being more dis-satisfied with Interpersonal relations were, thus, more likelyto show both pain interfering with work and intense pain,while workers with a more negative opinion on the Workorganization reported, to a larger extent, that their work wasinhibited by the pain, but not that the pain was intense.

4. Discussion

4.1. Summary. In this large population of flight baggagehandlers, intense low back and shoulder pain and paininterfering with work were both significantly associated with

some psychosocial factors at work, as measured by separatescales within the two domains Work organization and jobcontent and Interpersonal relations and leadership from theCOPSOQ II questionnaire [17]. These associations appearedafter adjusting for perceived physical load. When scales werecombined into the domainWork organization and job contentwe found significant associations with pain inhibiting work,but not with intense pain, while Interpersonal relations andleadership were strongly associated with both expressions ofpain. Social community at workwas the strongest single factorexplaining intense pain, while pain interfering with workshowed a particularly strong association with possibilitiesfor development for the low back, and with social supportfrom colleagues for the shoulder. The results suggest thatpsychosocial factors may be important to development andpersistence of pain even in occupations characterized byconsiderable physical loads.

4.2. Comparison with Previous Studies. The one-year preva-lence of LBP (64%) was similar to or even higher than thatfound in other occupations requiring heavy manual han-dling, for example, scaffolders (60%) [21], ambulanceworkers(60%) [22], and industrial workers (52%) [23]. The one-yearprevalence of LBP in the general population varies in theliterature between 10% and 56% [24]. The global prevalenceof LBP, irrespective of time window, was reported in asystematic review to be 31%, and the prevalence of activitylimiting LBP was 17% [25]. Thus, flight baggage handlersappear to have larger low back pain prevalence than thegeneral population. The prevalence of SP among the flightbaggage handlers (55%) was similar to previous reports fromscaffolders (50%) [21] and ambulance workers (46%) [22]. Inthe general population, the one-year SP prevalence reportedin the literature varies between studies from 5% to 47% [26].

Eighteen percent of the baggage handlers reported shoul-der pain interfering with work, which is considerably morethan among ambulance workers (7%) [22]. Explanationsmay be that it is easier to compensate for shoulder pain inambulance work by engaging in other tasks than lifting andcarrying equipment and patients to and from the ambulanceand that there may be better possibilities to reschedule workwith support from colleagues.

Almost one-third (30%) of the baggage handlers reportedlow back pain interfering with work, which is somewhatmore than the prevalence of activity limitation due to LBPreported by scaffolders (21%) and ambulance workers (23%).Low back pain with perceived disability has been shown torelate significantly to awkward arm postures [21], and a largeroccurrence of extreme arm postures in baggage handling, forinstance when operating in narrow aircraft compartments,than in ambulance work may explain some of the larger


Table 3: Mean values of ratings of 13 psychosocial work factors and the combined domains Work organization and job content and Inter-personal relations and leadership in each of the outcome groups; that is, no pain, pain interfering with work (PIW) and high pain intensity(PINT) for low back (LBP) and shoulder (SP).

All No pain LBP PIW LBP PINT SP PIW SP PINT𝑛 = 501 𝑛 = 79 𝑛 = 156 𝑛 = 180 𝑛 = 96 𝑛 = 147

m (SD) m (SD) m (SD) m (SD) m (SD) m (SD)Work organization, job content

Influence at work 38 (17) 39 (16) 35 (16) 36 (18) 32 (18) 35 (19)Possibilities for development 46 (17) 48 (17) 43 (16) 45 (15) 42 (16) 45 (15)Variation 44 (18) 48 (18) 40 (16) 43 (16) 40 (17) 43 (17)Meaning of work 58 (19) 62 (17) 54 (19) 56 (18) 51 (20) 54 (18)Commitment to the workplace 47 (29) 53 (22) 40 (18) 43 (20) 39 (20) 42 (19)

Interpersonal relationsPredictability 44 (20) 47 (21) 40 (18) 41 (19) 38 (20) 40 (19)Recognition 50 (22) 55 (21) 44 (22) 46 (22) 38 (21) 43 (22)Role clarity 67 (18) 69 (17) 63 (18) 63 (18) 63 (19) 64 (18)Role conflicts 46 (17) 44 (16) 50 (16) 49 (16) 53 (16) 50 (16)Quality of leadership 38 (22) 45 (22) 31 (20) 34 (20) 29 (20) 32 (20)Social support from colleagues 57 (18) 59 (18) 55 (17) 56 (16) 51 (19) 53 (18)Social support from supervisors 46 (24) 50 (25) 40 (23) 42 (23) 37 (25) 40 (24)Social community at work 79 (15) 80 (15) 79 (14) 78 (15) 76 (16) 76 (15)

Work organization 46 (13) 50 (12) 42 (12) 44 (12) 42 (14) 43 (13)Interpersonal relations 53 (12) 56 (12) 50 (13) 51 (13) 47 (14) 50 (12)

Table 4: Mean values of ratings of 13 psychosocial work factors and the combined domains Work organization and job content and Inter-personal relations and leadership in each of the six studied airports.

All Airport 1 Airport 2 Airport 3 Airport 4 Airport 5 Airport 6𝑛 = 501 𝑛 = 330–334 𝑛 = 100-101 𝑛 = 34 𝑛 = 26-27 𝑛 = 13-14 𝑛 = 11

m (SD) m (SD) m (SD) m (SD) m (SD) m (SD) m (SD)Work organization, job content

Influence at work 38 (17) 37 (18) 36 (16) 38 (19) 47 (17) 41 (16) 36 (16)Possibilities for development 46 (17) 44 (17) 49 (16) 49 (15) 55 (17) 53 (11) 52 (11)Variation 44 (18) 41 (18) 48 (16) 53 (16) 50 (18) 60 (9) 54 (12)Meaning of work 58 (19) 56 (20) 62 (16) 59 (18) 65 (18) 60 (15) 63 (21)Commitment to the workplace 47 (29) 45 (17) 54 (19) 48 (21) 52 (22) 51 (18) 45 (17)

Interpersonal relationsPredictability 44 (20) 45 (19) 42 (21) 41 (19) 57 (17) 28 (15) 22 (15)Recognition 50 (22) 51 (21) 52 (21) 39 (21) 67 (23) 42 (15) 33 (19)Role clarity 67 (18) 68 (18) 66 (18) 63 (18) 77 (11) 57 (18) 68 (18)Role conflicts 46 (17) 46 (17) 46 (17) 51 (12) 42 (15) 54 (14) 48 (6)Quality of leadership 38 (22) 41 (21) 35 (20) 19 (16) 53 (18) 28 (25) 11 (12)Social support from colleagues 57 (18) 56 (18) 57 (18) 65 (16) 62 (17) 69 (13) 57 (17)Social support from supervisors 46 (24) 48 (24) 40 (23) 33 (20) 65 (24) 41 (27) 21 (16)Social community at work 79 (15) 79 (16) 77 (13) 86 (11) 82 (16) 84 (9) 76 (12)

Work organization 46 (13) 44 (14) 50 (12) 49 (13) 53 (14) 53 (10) 50 (12)Interpersonal relations 53 (12) 55 (13) 53 (13) 49 (10) 63 (15) 49 (12) 42 (9)


Table 5: Hazard ratios (HR) with 95% confidence intervals for associations of psychosocial factors with pain interfering with work (PIW) inthe low back (LBP) and shoulder (SP) during the preceding year. All analyses were adjusted for age, BMI, general health and physical workload. Significant HRs marked in boldface.

LBP PIW SP PIWHR 95% CI HR 95% CI

Work organization, job contentInfluence at work 1.60 0.83–3.09 2.12 0.98–4.57Possibilities for development 2.86 1.32–6.18 2.63 1.06–6.51Variation 2.31 1.08–4.94 1.33 0.55–3.20Meaning of work 2.76 1.35–5.61 2.06 0.86–4.91Commitment to the workplace 2.39 1.15–4.95 1.44 0.63–3.29

Interpersonal relationsPredictability 1.94 0.94–3.98 1.44 0.62–3.37Recognition 2.67 1.33–5.35 2.57 1.11–5.95Role clarity 1.61 0.76–3.40 1.05 0.45–2.46Role conflicts 1.25 0.58–2.72 2.08 0.81–5.30Quality of leadership 1.77 0.82–3.82 1.22 0.51–2.94Social support from colleagues 2.48 1.16–5.29 4.06 1.55–10.65Social support from supervisors 2.22 1.08–4.58 1.33 0.60–2.95Social community at work 0.85 0.42–1.73 1.47 0.67–3.25

Work organization 3.65 1.67–7.99 2.68 1.09–6.61Interpersonal relations 2.18 1.06–4.49 2.09 0.88–4.96

Table 6: Hazard ratios (HR) with 95% CI for associations of psychosocial factors with high intensity pain (PINT) in the low back (LBP) andshoulder (SP) during the preceding year. All analyses were adjusted for age, BMI, general health and physical work load. Significant HRsmarked in boldface.

LBP PINT SP PINTHR 95% CI HR 95% CI

Work organization, job contentInfluence at work 1.46 0.86–2.46 1.43 0.83–2.46Possibilities for development 0.99 0.53–1.84 1.07 0.54–2.11Variation 0.97 0.53–1.79 0.89 0.47–1.69Meaning of work 1.02 0.56–1.86 1.57 0.83–2.97Commitment to the workplace 1.17 0.65–2.10 1.55 0.84–2.88

Interpersonal relationsPredictability 1.59 0.88–2.88 1.70 0.91–3.17Recognition 1.58 0.88–2.48 1.83 0.98–3.41Role clarity 2.07 1.08–3.95 1.81 0.94–3.50Role conflicts 1.17 0.61–2.24 1.76 0.88–3.53Quality of leadership 1.76 0.91–3.42 0.98 0.50–1.95Social support from colleagues 1.08 0.57–2.03 1.79 0.92–3.49Social support from supervisors 1.24 0.67–2.28 0.96 0.51–1.80Social community at work 1.61 0.89–2.93 2.21 1.18–4.13

Work organization 1.22 0.66–2.24 1.30 0.69–2.44Interpersonal relations 1.95 1.05–3.65 2.11 1.08–4.12

prevalence of LBP. While comparisons of prevalence databetween studies must be made with caution because ofpossible differences in definitions of LBP and SP, the citedstudies above of ambulance workers, scaffolders, and indus-trial workers all used the standardized Nordic Questionnaireto investigate the one-year retrospective prevalence of painand pain interfering with work. The scaffolder study used a

modified expression of pain and pain intensity during thepast 12 months, and so quantitative comparisons with ourresults need to be done with caution.

Higher mean values on the psychosocial rating scalesindicate more satisfaction (except for role conflicts). Meanscores for the 13 scales varied between 38 and 79 in theentire population of baggage handlers. Factorswith the lowest


mean scores were influence at work, quality of leadership,and predictability (mean = 37, 38, and 44, resp.), which issomewhat lower than among Danish construction workers(mean = 50, 54, and 54, resp.) (TheDanish National ResearchCenter for theWorking Environment 2014) [27].Mean scoreswere largest for social community at work, followed by roleclarity and meaning of work for both baggage handlers andconstruction workers. These two occupational groups, simi-lar in their rating of psychosocial factors, are both dominatedby a male workforce and perform heavy manual handlingtasks.Themore positive values for role clarity andmeaning ofworkmay be a sign of a vocational pride among the workers.

Several reviews have concluded that lack of social supportis a risk factor for pain in low back and upper extremities[11, 13, 28], while others do not claim an association [29]. Inour study, limited social support was clearly associated withlow back and shoulder pain interfering with work. This isconsistent with two cross-sectional studies, one of Swedishmale ambulance workers [22] where lack of social supportwas also associated with low back complains and activitylimitation and another of the general Canadian workingpopulation where low social support was associated withrestricted activity due to low back pain [30]. One hypothesismay be that poor social support is a contributing factor tothe onset or aggravation of MSDs through a stress response.Lack of support may cause increased muscle tension, asfacilitated by stress hormones, which eventually may leadto pain [31]. Social support from colleagues may also be animportant factor for coping with pain and staying in thebaggage handling job, since it may support, for example,work rotation and task variation. This theory correspondswith Woods [13], who claimed evidence, even if limited, forlack of social support being associated with absenteeism andrestricted activity due to MSD. Woods also viewed goodsocial support as an important promoting factor for workerswith MSDs to be able to return to work.

4.3. Methodological and Theoretical Considerations. Thisstudy used a cross-sectional design, and so the observedassociations between psychosocial factors and pain can beinterpreted to show causal relationships only with greatcaution. In another cross-sectional study, Davies and Heaney[32] showed that associations between psychosocial workcharacteristics and pain were stronger when using self-reported pain than when pain was diagnosed in a physicalexamination. The authors interpreted this to show thatpsychosocial stressors influence reporting of pain ratherthan physiologic responses associated with pain and alsoconsidered their results to reflect the influence of low backpain on the reporting of poor psychosocial conditions. Thus,cross-sectional relationships between factors at work andpain may, to some extent, be spurious. To this end, it is wellknown that subjective opinions about the work environmentmay be influenced by several factors in addition to thehealth status of the respondent, such as the context of where,how, and when exposure and outcome data were collected.Thus, workers being observant of a relationship betweenoccupational factors andMSDsmay both consider exposures

to be larger and attribute their possible pain to their work[32]. This possible attribution bias of pain ratings may beparticularly pronounced if the workers are required to answerquestions while at work, as compared to outside work [33].Thus, in our study, the administration of questionnaires tobe answered during working hours and in the context of aproject addressing the work environment may have led to anoverestimation of the prevalence and intensity of pain and theextent to which that pain interfered with work.

For reasons of feasibility we assessed pain by self-report,using the NMQ questionnaire, which has been used in aplethora of previous studies since its publication in 1987.Several studies have shown that the pain prevalence obtainedwhen using NMQ is larger than that “confirmed” by clinicalexamination (e.g., [34]) and thus that specificity might be anissue when using NMQ. However, pain ratings obtained withNMQ have also been shown to have a good predictabilitywith respect to secondary outcomes, such as sickness absencefrom work [19], and we utilized this property by categorizingworkers according to their self-reported pain intensity, usinga discrimination limit which is predictive of long-termsickness absence [20].

Many previous studies of psychosocial factors at workhave only addressed the standard demand-control-supportmodel or have focused on the Job content questionnaire.We used the validated COPSOQ method, which encom-passes several additional dimensions of the psychosocialwork environment. COPSOQ is currently a well-establishedquestionnaire for workplace investigations. By using COP-SOQ, our study can elucidate even positive aspects of thebaggage handlers’ working conditions, for example, socialcommunity at work, possibilities for development, andmeaningof work. In addition to the 13 separate factors included inthe two psychosocial domains documented in our study,that is, Work organization and job content and Interpersonalrelations and leadership, we also, to our knowledge forthe first time, used these combined domains as indepen-dent variables in an analysis for associations with MSDs.Using combined domains renders the study less sensitive toredundant findings on separate factors, caused by the sameworkers reporting low scores on several factors at the sametime.

Mechanisms for causation of MSDs by psychosocialexposures in the presence of physical workloads are notfully clarified. Several theories have been presented [35],one example being the biopsychosocial model, suggestingpsychosocial risk factors to exacerbate the effects of physicalexposure on the risk for developing MSDs. As an example,according to thismodel, sociocultural factors such as demandor pressure from colleagues and supervisors and attitudesand behaviors at the workplace may act together with biome-chanical and biological factors, such as personal capacities ininfluencing the work, and modify the risk for MSDs. Mentalworkloads may even increase muscle tension, which can thenlead to biomechanical stress [36, 37] leading to increasedmuscle metabolites, inflammatory changes, and muscle pain[38].This influence of mental loads onmuscle activationmayparticularly affect certain low threshold motor units, which


may then be prone to develop chronic conditions, accordingto the “Cinderella” recruitment hypothesis [36].

4.4. Generalizability and Implications. Our study is a nation-wide investigation of a homogenous occupational group offlight baggage handlers working at the ramp or in sorting,and in spite of some limitations, for instance the somewhatmeager response rate, we believe that the results can be gen-eralized to the general population of baggage handlers on anational level and that theymay have some portability even toother countries, noting that flight baggage handlers probablyhave very similar jobs in all major airports worldwide (seebelow).

Only workers in ramp and sorting areas were includedin this study and we therefore believe that the group isreasonably homogeneous with respect to physical workload.However, tasks and work organization may differ, in partic-ular between workers in different airports. We were able toadjust for the general effect of perceived physical load, butdata did not permit a nested analysis of associations betweenpsychosocial factors and pain in each of the six airportsdue to the limited number of workers in some of them(cf. Table 4). Since the psychosocial conditions appearedto differ between airports, we cannot rule out that someof the observed associations are confounded by airport-specific factors that were not recorded and analyzed, suchas variation in physical work load. We chose not to adjustfor seniority at work, since it correlates highly with age,which was adjusted for. We did not have access to data onpossible confounders describing current acute or systemicdisease.

In our data collection, pain could have been a motivatingfactor for participation, resulting in a study population withan overrepresentation of workers with pain. Other studieshave, indeed, shown nonresponders to have less pain thanresponders [39]. If so, our resultsmay overestimate the preva-lence of pain among baggage handlers. In spite of extensiveand repeated efforts in retrieving questionnaires from all806 eligible baggage handlers, we only got a response rateof 65%. While a formal analysis of nonrespondents was notpossible, we have the impression that workers on night shiftsresponded less than day-shift workers, which would limit therepresentativeness of our results to mainly day shift baggagehandling, and also that some team managers were less activein encouraging their team members into participating. Itis possible that such nonresponding teams would have adifferent experience of their psychosocial conditions thanteams motivated to participate, but we could, for obviousreasons, not explore that. Differential nonresponding mayalso have occurred due to workers with a deviating job strainbeing less inclined to participate [40].

Workers developing MSDs at work are more likely toleave their job, which may lead to an underestimated riskof exposures causing MSDs (the so-called healthy workereffect). However, data provided by the participating handlingcompanies showed that the annual workforce turnover wasless than five percent. Thus, despite a considerable preva-lence of pain and negative opinions on psychosocial factors,

workers stayed at the job. In addition to the harsh conditionsin the current labor market in Sweden, the small turn-overmay be a result of the valued social community at work.Social community at work was scored as the most positivepsychosocial factor in our investigation, and the impressionof a good social community was confirmed by several infor-mal conversations with the baggage handlers and their unionrepresentatives.

Studies have shown that if low back pain becomeschronic, workers may not necessarily be on sick leave [41],but the pain may influence productivity and company costsin a negative way [42]. This agrees with our results of a con-siderable proportion of workers reporting that their paininhibited work (PIW). Thus, Heuvel et al. (2010) found psy-chosocial factors to be more strongly associated with a lowperformance at work than with sickness absence in a nationalcross-sectional study of the general Dutch working popula-tion [43]. Favorable psychosocial work conditionsmay there-fore have a decisive role in securing that productivity goalsare met, such as, for baggage handling, average time spentloading or unloading an aircraft, frequency of departures ontime, proportion of baggage being delivered undamaged, andproportion of baggage going to the correct destination.

Airport baggage handling is a world-wide occupationwith, to a large extent, similar working conditions, as setout by the standardized construction of airplanes and ramps,and so we believe that our study is of interest even outsideSweden, at least in large- and medium-sized airports. How-ever, we also emphasize that psychosocial conditions may,to a considerable extent, be specific to individual handlingcompanies and that our quantitative results may thereforebe difficult to transfer directly to other companies thanthose investigated. This said, our study revealed associationsbetween psychosocial factors and MSDs, which may be usedas a general inspiration for identifying targets for interventionin baggage handling, in addition to possible interventions onthe physical workloads.

5. Conclusions

We conducted a nationwide study of psychosocial work con-ditions and musculoskeletal health among baggage handlerswithin the aviation industry in Sweden. We found the one-year prevalence of low back and shoulder pain to be in paritywith other heavy manual occupations. We found significantassociations between, on one hand, the psychosocial domainsWork organization and job content and Interpersonal relationsand leadership, and, on the other hand, intense pain andpain interfering with work. Thus, while being cross-sectionaland therefore only tentatively interpretable in terms of causalrelationships, our study suggests that psychosocial factorsmay be involved in explaining the occurrence of pain inflight baggage handling, in spite of this job also presentingconsiderable physical loads. Our results also suggest that thepsychosocial work environment may be a relevant target forintervention in this occupation.


Appendix

A. Questions on PsychosocialFactors as Appearing in the Second,Updated Version of the StandardizedCOPSOQ Questionnaire [17]

A.1. Work Organization and Job Contents (Five Scales)

Influence at Work (Four Questions). Do you have a largedegree of influence concerning your work? Do you have asay in choosing who you work with? Can you influence theamount of work assigned to you? Do you have any influenceon what you do at work?

Possibilities for Development (Four Questions). Does yourwork require you to take the initiative? Do you have thepossibility of learning new things through your work? Canyou use your skills or expertise in your work?Does your workgive you the opportunity to develop your skills?

Meaning ofWork (ThreeQuestions). Is yourworkmeaningful?Do you feel that the work you do is important? Do you feelmotivated and involved in your work?

Variation (Two Questions). Is your work varied? Do you haveto do the same thing over and over again?

Commitment to theWorkplace (FourQuestions).Do you enjoytelling others about your place of work? Do you feel thatyour place of work is of great importance to you? Wouldyou recommend a good friend to apply for a position atyour workplace? How often do you consider looking for workelsewhere?

A.2. Interpersonal Relations and Leadership (Eight Scales)

Predictability (Two Questions). At your place of work, are youinformedwell in advance concerning, for example, importantdecisions, changes, or plans for the future? Do you receive allthe information you need in order to do your work well?

Recognition (Three Questions). Is your work recognized andappreciated by the management? Does the management atyour workplace respect you? Are you treated fairly at yourworkplace?

Role Clarity (Three Questions). Does your work have clearobjectives? Do you know exactly which areas are yourresponsibility? Do you know exactly what is expected of youat work?

Role Conflicts (Four Questions). Do you do things at work,which are accepted by some people but not by others? Arecontradictory demands placed on you at work? Do yousometimes have to do things which ought to have been donein a different way?Do you sometimes have to do things whichseem to be unnecessary?

Quality of Leadership (Four Questions). To what extent wouldyou say that your immediate superior: makes sure that theindividual member of staff has good development opportu-nities? gives high priority to job satisfaction? is good at workplanning? is good at solving conflicts?

Social Support from Colleagues (Three Questions). How oftendo you get help and support from your colleagues? How oftenare your colleagues willing to listen to your problems at work?How often do your colleagues talk with you about how wellyou carry out your work?

Social Support from Supervisors (Three Questions).How oftenis your nearest superior willing to listen to your problemsat work? How often do you get help and support from yournearest superior? How often does your nearest superior talkwith you about how well you carry out your work?

Social Community at Work (Three Questions). Is there a goodatmosphere between you and your colleagues? Is there goodcooperation between the colleagues at work? Do you feel partof a community at your place of work?

Abbreviations

ULD: Unit load deviceMSD: Musculoskeletal disordersLBP: Low back painTYA: The Vocational Training and Working

Environment CouncilCOPSOQ: Copenhagen Psychosocial QuestionnaireSP: Shoulder painPIW: Pain interfering with workPINT: Pain of high intensity.


The authors, Eva L. Bergsten, S. E. Mathiassen, and E.Vingard, declare no conflict of interests regarding the studyand this paper.

Authors’ Contribution

Eva L. Bergsten contributed to the design of the study, carriedout the data collection, participated in data analysis, and wasmainly responsible for drafting the paper. S. E. Mathiassencontributed to the design of the study, participated in dataanalysis and data interpretation, and contributed in draftingand revising the manuscript. E. Vingard participated in anal-ysis and interpretation of data and in drafting and revising thepaper. All authors read and approved the final paper.

Acknowledgments

The present study was financially supported by grants fromAFA Insurance (Dnr 2010/358) and the Swedish ResearchCouncil for Health, Working Life, and Welfare (Forte Dnr.2009-1761). The authors gratefully acknowledge the Voca-tional Training & Working Environment Council (TYA) for


collaboration.They would also express their gratitude to Rei-dar Pettersson, Erik Alphonse, and Dan Holmberg for theircontributions to the data collection and to Tobias Nordquistfor support in data analyses. Finally, they would particularlylike to thank the handling companies and baggage handlerswho participated in this study.

References

[1] Global Airline Industry, 2014, http://web.mit.edu/airlines/anal-ysis/analysis airline industry.html.

[2] S. Gallagher, “Physical limitations and musculoskeletal com-plaints associated with work in unusual or restricted postures:a literature review,” Journal of Safety Research, vol. 36, no. 1, pp.51–61, 2005.

[3] S. Gallagher and J. R. Heberger, “Examining the interaction offorce and repetition onmusculoskeletal disorder risk: a system-atic literature review,”Human Factors, vol. 55, no. 1, pp. 108–124,2013.

[4] R. M. Van Rijn, B. M. Huisstede, B. W. Koes, and A. Burdorf,“Associations between work-related factors and specific disor-ders of the shoulder—a systematic review of the literature,”Scandinavian Journal of Work, Environment & Health, vol. 36,no. 3, pp. 189–201, 2010.

[5] B. Larsson, K. Søgaard, and L. Rosendal, “Work related neck–shoulder pain: a review onmagnitude, risk factors, biochemicalcharacteristics, clinical picture and preventive interventions,”Best Practice & Research: Clinical Rheumatology, vol. 21, no. 3,pp. 447–463, 2007.

[6] E. W. P. Bakker, A. P. Verhagen, E. Van Trijffel, C. Lucas, and B.W. Koes, “Spinal mechanical load as a risk factor for low backpain: a systematic review of prospective cohort studies,” Spine,vol. 34, no. 8, pp. E281–E293, 2009.

[7] J. I. Kuiper, A. Burdorf, J. H. A. M. Verbeek, M. H. W. Frings-Dresen, A. J. van der Beek, and E. R. A. Viikari-Juntura, “Epi-demiologic evidence on manual materials handling as a riskfactor for back disorders:a systematic review,” InternationalJournal of Industrial Ergonomics, vol. 24, no. 4, pp. 389–404,1999.

[8] R. E. Splittstoesser, G. Yang, G. G. Knapik et al., “Spinal loadingduring manual materials handling in a kneeling posture,”Journal of Electromyography and Kinesiology, vol. 17, no. 1, pp.25–34, 2007.

[9] H. R. Stalhammar, T. P. J. Leskinen, I. A. A. Kuorinka, M. H. J.Gautreau, and J. D. G. Troup, “Postural, epidemiological andbiomechanical analysis of luggage handling in an aircraft lug-gage compartment,” Applied Ergonomics, vol. 17, no. 3, pp. 177–183, 1986.

[10] P. M. Bongers, A. M. Kremer, and J. T. Laak, “Are psychosocialfactors, risk factors for symptoms and signs of the shoulder,elbow, or hand/wrist?: a review of the epidemiological litera-ture,” American Journal of Industrial Medicine, vol. 41, no. 5, pp.315–342, 2002.

[11] A. Hauke, J. Flintrop, E. Brun, and R. Rugulies, “The impactof work-related psychosocial stressors on the onset of muscu-loskeletal disorders in specific body regions: A review andmeta-analysis of 54 longitudinal studies,”Work & Stress, vol. 25, no. 3,pp. 243–256, 2011.

[12] W. E. Hoogendoorn, P. M. Bongers, H. C.W. De Vet et al., “Psy-chosocial work characteristics and psychological strain in rela-tion to low-back pain,” Scandinavian Journal of Work, Envi-ronment and Health, vol. 27, no. 4, pp. 258–267, 2001.

[13] V. Woods, “Work-related musculoskeletal health and socialsupport,” Occupational Medicine, vol. 55, no. 3, pp. 177–189,2005.

[14] T. Clausen, L. L. Andersen, A.Holtermann,A. F. B. Jorgensen, B.Aust, and R. Rugulies, “Do self-reported psychosocial workingconditions predict low back pain after adjustment for bothphysical work load and depressive symptoms? A prospectivestudy among female eldercare workers,” Occupational andEnvironmental Medicine, vol. 70, no. 8, pp. 538–544, 2013.

[15] D. C. Turk and A. Okifuji, “Psychological factors in chronicpain: evolution and revolution,” Journal of Consulting and Clin-ical Psychology, vol. 70, no. 3, pp. 678–690, 2002.

[16] T. S. Kristensen, H. Hannerz, A. Høgh, and V. Borg, “TheCopenhagen Psychosocial Questionnaire—a tool for the assess-ment and improvement of the psychosocial work environment,”Scandinavian Journal of Work, Environment and Health, vol. 31,no. 6, pp. 438–449, 2005.

[17] J. H. Pejtersen, T. S. Kristensen, V. Borg, and J. B. Bjorner, “Thesecond version of the Copenhagen Psychosocial Question-naire,” Scandinavian Journal of Public Health, vol. 38, no. 3, pp.8–24, 2010.

[18] I. Kuorinka, B. Jonsson, A. Kilbom et al., “Standardised Nordicquestionnaires for the analysis of musculoskeletal symptoms,”Applied Ergonomics, vol. 18, no. 3, pp. 233–237, 1987.

[19] A. Descatha, Y. Roquelaure, J. F. Chastang et al., “Validity ofNordic-style questionnaires in the surveillance of upper-limbwork-related musculoskeletal disorders,” Scandinavian Journalof Work, Environment & Health, vol. 33, no. 1, pp. 58–65, 2007.

[20] L. L. Andersen, T. Clausen, H. Burr, and A. Holtermann,“Threshold of musculoskeletal pain intensity for increased riskof long-term sickness absence among female healthcareworkersin eldercare,” PLoS ONE, vol. 7, no. 7, Article ID e41287, 2012.

[21] L. A. M. Elders and A. Burdorf, “Interrelations of risk factorsand low back pain in scaffolders,” Occupational and Environ-mental Medicine, vol. 58, no. 9, pp. 597–603, 2001.

[22] U.Aasa,M. Barnekow-Bergkvist, K.-A. Angquist, andC. Brulin,“Relationships between work-related factors and disorders inthe neck-shoulder and low-back region among female andmaleambulance personnel,” Journal of Occupational Health, vol. 47,no. 6, pp. 481–489, 2005.

[23] W. Ijzelenberg andA. Burdorf, “Risk factors formusculoskeletalsymptoms and ensuing health care use and sick leave,” Spine,vol. 30, no. 13, pp. 1550–1556, 2005.

[24] D. Hoy, P. Brooks, F. Blyth, and R. Buchbinder, “The epidemi-ology of low back pain,” Best Practice & Research: ClinicalRheumatology, vol. 24, no. 6, pp. 769–781, 2010.

[25] D. Hoy, C. Bain, G. Williams et al., “A systematic review of theglobal prevalence of low back pain,” Arthritis & Rheumatism,vol. 64, no. 6, pp. 2028–2037, 2012.

[26] J. J. Luime, B.W.Koes, I. J.M.Hendriksen et al., “Prevalence andincidence of shoulder pain in the general population; a system-atic review,” Scandinavian Journal of Rheumatology, vol. 33, no.2, pp. 73–81, 2004.

[27] NRCWE, April 2014, http://olddata.arbejdsmiljoforskning.dk/Nationale%20Data/3DII.aspx?lang=da.

[28] W. E. Hoogendoorn, M. N. M. van Poppel, P. M. Bongers, B. W.Koes, L. M. Bouter, and L. Hoogendoorn, “Systematic review ofpsychosocial factors at work and private life as risk factors forback pain,” Spine, vol. 25, no. 16, pp. 2114–2125, 2000.

[29] J. Hartvigsen, S. Lings, C. Leboeuf-Yde, and L. Bakketeig, “Psy-chosocial factors at work in relation to low back pain and


consequences of low back pain; a systematic, critical reviewof prospective cohort studies,” Occupational & EnvironmentalMedicine, vol. 61, no. 1, p. e2, 2004.

[30] D. C. Cole, S. A. Ibrahim, H. S. Shannon, F. Scott, and J. Eyles,“Work correlates of back problems and activity restrictiondue to musculoskeletal disorders in the Canadian nationalpopulation health survey (NPHS) 1994-5 data,” Occupationaland Environmental Medicine, vol. 58, no. 11, pp. 728–734, 2001.

[31] U. Lundberg, “Psychophysiology of work: stress, gender, endo-crine response, and work-related upper extremity disorders,”American Journal of Industrial Medicine, vol. 41, no. 5, pp. 383–392, 2002.

[32] K. G. Davis and C. A. Heaney, “The relationship between psy-chosocial work characteristics and low back pain: underlyingmethodological issues,” Clinical Biomechanics, vol. 15, no. 6, pp.389–406, 2000.

[33] S. Baron, T. Hales, and J. Hurrell, “Evaluation of symptom sur-veys for occupational musculoskeletal disorders,” AmericanJournal of Industrial Medicine, vol. 29, no. 6, pp. 609–617, 1996.

[34] D. Jonker, E. Gustafsson, B. Rolander, I. Arvidsson, and C.Nordander, “Health surveillance under adverse ergonomicsconditions—validity of a screening method adapted for theoccupational health service,” Ergonomics, pp. 1–10, 2015.

[35] C. Deeney and L. O’Sullivan, “Work related psychosocial risksand musculoskeletal disorders: potential risk factors, causationand evaluation methods,” Work, vol. 34, no. 2, pp. 239–248,2009.

[36] B. Melin and U. Lundberg, “A biopsychosocial approach towork-stress and musculoskeletal disorders,” Journal of Psy-chophysiology, vol. 11, no. 3, pp. 238–247, 1997.

[37] P. M. Bongers, S. Ijmker, S. van den Heuvel, and B. M. Blatter,“Epidemiology of work related neck and upper limb problems:psychosocial and personal risk factors (part I) and effectiveinterventions from a bio behavioural perspective (part II),”Journal of Occupational Rehabilitation, vol. 16, no. 3, pp. 279–302, 2006.

[38] T. Theorell, H.-M. Hasselhorn, and Music Norrtalje StudyGroup, “Endocrinological and immunological variables sensi-tive to psychosocial factors of possible relevance towork-relatedmusculoskeletal disorders,”Work&Stress, vol. 16, no. 2, pp. 154–165, 2002.

[39] A. C. Papageorgiou, P. R. Croft, S. Ferry, M. I. V. Jayson, andA. J. Silman, “Estimating the prevalence of low back pain in thegeneral population: evidence from the South Manchester BackPain Survey,” Spine, vol. 20, no. 17, pp. 1889–1894, 1995.

[40] M. Cifuentes, J. Boyer, R. Gore et al., “Job strain predicts surveyresponse in healthcare industry workers,”TheAmerican Journalof Industrial Medicine, vol. 51, no. 4, pp. 281–289, 2008.

[41] E. Vingard, M. Mortimer, C. Wiktorin et al., “Seeking care forlow back pain in the general population. A two-year follow-upstudy: results from the MUSIC-Norrtalje study,” Spine, vol. 27,no. 19, pp. 2159–2165, 2002.

[42] W. F. Stewart, J. A. Ricci, E. Chee,D.Morganstein, andR. Lipton,“LOst productive time and cost due to common pain conditionsin the US workforce,” The Journal of the American MedicalAssociation, vol. 290, no. 18, pp. 2443–2454, 2003.

[43] S. G. van den Heuvel, G. A. Geuskens, W. E. Hooftman, L. L. J.Koppes, and S. N. J. van den Bossche, “Productivity loss at work;health-related and work-related factors,” Journal of Occupa-tional Rehabilitation, vol. 20, no. 3, pp. 331–339, 2010.

Research ArticleThe Association between Job Strain and Atrial Fibrillation:Results from the Swedish WOLF Study

Eleonor I. Fransson,1,2 Magdalena Stadin,1 Maria Nordin,3,4 Dan Malm,1,5

Anders Knutsson,6 Lars Alfredsson,2 and Peter J. M. Westerholm7

1School of Health Sciences, Jonkoping University, 551 11 Jonkoping, Sweden2Institute of Environmental Medicine, Karolinska Institutet, 171 77 Stockholm, Sweden3Department of Psychology, Umea University, 901 87 Umea, Sweden4Stress Research Institute, Stockholm University, 106 91 Stockholm, Sweden5Department of Internal Medicine, County Hospital Ryhov, 551 85 Jonkoping, Sweden6Department of Health Sciences, Mid Sweden University, 851 70 Sundsvall, Sweden7Occupational and Environmental Medicine, Uppsala University, 751 85 Uppsala, Sweden

Correspondence should be addressed to Eleonor I. Fransson; [email protected]

Received 16 January 2015; Accepted 23 March 2015

Academic Editor: Giancarlo Cesana

Copyright © 2015 Eleonor I. Fransson et al. This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.

Introduction. Atrial fibrillation (AF) is a common heart rhythm disorder. Several life-style factors have been identified as risk factorsfor AF, but less is known about the impact of work-related stress. This study aims to evaluate the association between work-relatedstress, defined as job strain, and risk of AF. Methods. Data from the Swedish WOLF study was used, comprising 10,121 workingmen and women. Job strain was measured by the demand-control model. Information on incident AF was derived from nationalregisters. Cox proportional hazard regression was used to estimate hazard ratios (HR) and 95% confidence intervals (CI) for theassociation between job strain and AF risk. Results. In total, 253 incident AF cases were identified during a total follow-up time of132,387 person-years. Job strain was associated with AF risk in a time-dependent manner, with stronger association after 10.7 yearsof follow-up (HR 1.93, 95% CI 1.10–3.36 after 10.7 years, versus HR 1.11, 95% CI 0.67–1.83 before 10.7 years). The results pointedtowards a dose-response relationship when taking accumulated exposure to job strain over time into account. Conclusion. Thisstudy provides support to the hypothesis that work-related stress defined as job strain is linked to an increased risk of AF.

1. Introduction

Cardiovascular disease (CVD) is the leading cause of deathin a global perspective. According to the World HealthOrganization (WHO) 17.3 million people died from CVD in2008, which represents 30% of the global deaths [1]. Atrialfibrillation (AF) is the most common cardiac arrhythmia andis also a well-confirmed risk factor of stroke [2–4]. Symptomsof atrial fibrillation include palpitations, shortness of breath,fatigue, chest pain, dizziness, and reduced physical capacity.Atrial fibrillation often affects the patients, as well as theirfamilymembers, with distress and reducedwell-being in theirdaily life [5–7]. In 2010, the estimated global age-adjustedprevalence of AF in the population of 35 years and older was

1368.5 per 100 000 inmen and 856.8 in women [8]. In Europe,3.7–4.2% of those aged 60–70 and 10–17% of those 80 yearsor older suffer from AF [9]. The prevalence of AF in Swedenis estimated to be 2.9% [10]. The incidence and prevalenceof AF increase with age [11–14], and AF is more commonin men than women [8, 11]. During the recent decades, theincidence of AF has increased, and this tendency is presumedto maintain [8].

It is not unusual that AF occurs in conjunction withother CVD (e.g., heart failure and heart valve problems) andhypertension [15]. However, AF may also occur without theimpact of those factors and hereditary and life style factorsare likely to play a part in the pathogenesis of AF [15–18]. Obesity, sleep apnea, heavy alcohol consumption, and


http://dx.doi.org/10.1155/2015/371905


prolonged physical exertion are examples of life-style factorsthat have been proposed as risk factors of AF [15, 18–22].

Mental stress is often reported by AF patients as atriggering factor of AF attacks [23], but few studies haveevaluated the association between psychosocial stress andAF. However, in a case-control study it was found that acutelife stress was related to AF risk [24], and in one recentlypublished prospective study, an association between work-related stress and increased risk of AF was observed [25].

Although AF is common among the general populationand considered as a public health disease, the knowledgeabout different risk factors and AF is still insufficient. Theaim of this study is to provide additional knowledge aboutthe relation betweenwork-related stress, defined as job strain,and the onset of AF.

2. Material and Methods

Data were obtained from the Work, Lipids, and Fibrinogen(WOLF) study, a longitudinal occupational cohort studyconducted in Sweden [26]. The original aim with the WOLFstudy was to investigate the associations between psychoso-cial work environment and cardiovascular risk factors.

2.1. Procedure and Participation. The baseline data collectionin WOLF was carried out in the county of Stockholmduring 1992–1995 (WOLF Stockholm, 𝑛 = 5698). In 1996–1998, the data collection was extended to northern Sweden,in the counties of Vasternorrland and Jamtland (WOLFNorrland, 𝑛 = 4718). WOLF Norrland was partly estab-lished in order to recruit more blue-collar workers into theproject. Altogether 36 occupational health service units inthe counties of Stockholm, Vasternorrland, and Jamtlandwere invited to participate. Of those, 33 occupational healthservice units accepted to participate.The occupational healthservice units represented approximately 60 companies indifferent branches comprising various occupations. However,including all employees at the 60 companies was not pos-sible. Instead, all employees representing certain workplaces(e.g., a department, garage, institution, laboratory, and salesorganization) were asked to participate. This selection wasbasically due to practical reasons from the perspective ofthe occupational health service units. Employees who wereon more or less permanent leave from the workplace, forexample, those stationed abroad or chronically ill, were notincluded in the study population. The participation rateat baseline was 82%, with higher participation rate in thenorthern part in Sweden.

At baseline, the participants filled in an extensive ques-tionnaire covering different occupational aspects (e.g., worktasks, work hours, and work environment), sociodemo-graphic aspects (e.g., education level), and lifestyle habits(e.g., smoking and physical exercise) as well as differentaspects of health. In addition, a minor clinical examinationwas conducted by specially trained nurses at the occupationalhealth service units. The clinical examination included mea-surements of height, weight, waist and hip circumference, andblood pressure. Blood samples were also collected.

A follow-up study in WOLF Norrland was conducted in2000–2003. In total, 3633 participants fromWOLF Norrlandprovided repeated measurements on work and life-stylefactors by taking part in the follow-up study.

2.2. Analytical Sample. In total, 10 416 working men andwomen participated by answering the questionnaire andtaking part of the clinical examination at baseline. For thepresent study, we excluded participants who reported thatthey had experienced a myocardial infarction or heart failureprior to baseline. We also excluded those with a recordeddiagnosis of AF in national hospital discharge and outpatientregisters prior to baseline, leaving 10 121 participants (6971men and 3150 women) as our analytical sample. In theanalyses with repeated measurements, 3123 participants wereincluded.

2.3. Work-Related Stress according to the Demand-ControlModel. Work-related stress was defined according toKarsek’sjob demand-controlmodel [27, 28], which is characterized bythe combination of psychological job demands and controlover the work situation. The model proposes that thoseexperiencing high job demands in combination with lowcontrol (i.e., high strain or job strain) are in a stressful worksituation and are at higher risk for developing ill health. In thisstudy we used the Swedish demand-control questionnairecontaining five job demands items and six control items tomeasure job strain [29]. Cronbach’s alpha for the job demandsand job control subscales was 0.72 and 0.74, respectively.Mean response scores for the job demands and for the jobcontrol items were computed for each participant. We usedthe median scores as cut-points for high and low demands(“high demands” being defined as scores strictly above thestudy-specific median score) and job control (“low control”being defined as scores strictly below the study-specificmedian score). In the analyses we used both a dichotomizedmeasure of job strain (high strain versus all others) andfour categories based on the combination of job demandsand control: low strain jobs (low demands, high control);passive jobs (low demands, low control); active jobs (highdemands, high control); and high strain jobs (high demands,low control).

2.4. Atrial Fibrillation. Information on incident AF, or flutter,was derived from the Swedish national hospital discharge,outpatient, andmortality registers by using the following ICDcodes: ICD-10 code I48; ICD-9 code 427D; and ICD-8 code427.92.

2.5. Potential Confounding and Mediating Factors. In addi-tion to age in years (continuous), sex, and part of study(Stockholm, Norrland), we considered the following fac-tors as potential confounding or mediating factors: socioe-conomic status (manual workers, lower level/intermediatenonmanual employees, and professionals), exercise (seldom,sometimes, and regularly), smoking (never (neither currentnor ex-smoker), ex-smoker (has previously smoked for atleast one year but is not a current smoker), and current


smoker), alcohol consumption (none (0 units/week), mod-erate (1–14 units/week for women, 1–21 units/week for men),intermediate (15–20 units/week forwomen, 22–27 units/weekfor men), and heavy (21 units/week or more for women, 28units/week or more for men)), waist circumference (<94 cm(men) or <80 cm (women), 94–101.99 cm (men) or 80–88(women), and >102 cm (men) or >88 cm (women)), andhypertension (systolic blood pressure ≥140mmHg, or dias-tolic blood pressure ≥90mmHg, or self-reported treatmentwith antihypertensive drugs).

2.6. Statistical Analyses. The participants were followed upfrom their baseline assessment of job strain to the firstregistered AF event, migration out of Sweden, death, or endof follow-up, whichever came first. Independent 𝑡-tests andChi2-tests for bivariate analyses were conducted in order totest potential differences in baseline characteristics betweenparticipants with and without AF. Cox proportional hazardregression was used to estimate hazard ratios (HR) and 95%confidence intervals (CI) to quantify the relationship betweenjob strain and risk of atrial fibrillation. All analyses wereadjusted for age, sex, and part of study. Other potentialconfounding and mediating factors were added one by oneto the age, sex, and part of study adjusted Cox proportionalhazard regression model. Only those factors changing esti-mates of job strain versus others with more than 10% wereto be included in subsequent models [30]. To evaluate apotential effectmodification by sex, a stratified analysis by sexwas carried out, as well as including a statistical interactionterm between job strain and sex in the Cox proportionalhazard regression model. Kaplan-Meier curves were used toinspect the proportionality of hazards over time. Analysesof accumulated exposure to job strain and the risk of AFwere carried out in a subsample of theWOLFNorrland studypopulation for whom repeated measures of job strain wereavailable. In the analyses with repeated measures, the startof follow-up time was set at the date of the second datacollection. A 𝑃 value for trend was derived by including thevariable on accumulated job strain as a continuous variablewith three levels in the Cox proportional hazard regressionmodel. All data analyses were carried out using SAS version9.2.

2.7. Ethics. All participants gave informed consent to partici-pate in the study. TheWOLF study has been approved by theEthics Committee at Karolinska Institutet, Stockholm (# 92-198), and the Regional Ethical Review Board in Stockholm(# 2006/257-31, # 2008/1638-31/5).

3. Results

Characteristics of the study sample are presented in Table 1.During a total follow-up time of 132,387 person-years(median follow-up time 13.6 years), 253 incident AF eventswere recorded. Compared to participants without AF, partic-ipants with AFweremore likely to bemale, older, and currentor ex-smokers and have higher waist circumference andmorelikely to suffer from hypertension.

Table 1: Baseline characteristics in the total study sample and amongparticipants with and without atrial fibrillation (AF), the WOLFstudy, Sweden.

Characteristics Totaln = 10121

Not AFn = 9868

AFn = 253 𝑃 value

∗

Age, mean (sd) 42.5 (10.7) 42.3 (10.7) 51.3 (8.3) <0.001Sex, 𝑛 (%)Men 6971 (69) 6757 (68) 214 (85)

<0.001Women 3150 (31) 3111 (32) 39 (15)Study part, 𝑛 (%)Stockholm 5518 (55) 5368 (54) 150 (59) 0.12Norrland 4603 (45) 4500 (46) 103 (41)

Demand-control, 𝑛(%)No strain 8960 (89) 8739 (89) 221 (87) 0.55Job strain 1161 (11) 1129 (11) 32 (13)Low strain 3192 (32) 3112 (32) 80 (32)

0.64Passive 3259 (32) 3186 (32) 73 (29)Active 2509 (25) 2441 (25) 68 (27)High strain 1161 (11) 1129 (11) 32 (13)

SES, 𝑛 (%)Manual workers 4422 (44) 4310 (44) 112 (44)Lower level/intermediatenonmanualemployees

4341 (43) 4242 (44) 99 (39) 0.09

Professionals 1234 (12) 1192 (12) 42 (17)Physical exercise, 𝑛(%)Seldom 2488 (25) 2420 (25) 68 (27)

0.64Sometimes 3876 (38) 3781 (38) 95 (38)Regularly 3727 (37) 3639 (37) 88 (35)

Smoking, 𝑛 (%)Never smokers 4739 (48) 4664 (48) 75 (30)

<0.001Ex-smokers 2893 (29) 2788 (29) 105 (42)Current smokers 2250 (23) 2182 (23) 68 (27)

Alcoholconsumption, 𝑛 (%)Non 487 (5) 470 (5) 17 (7)

0.37Moderate 8688 (88) 8476 (88) 212 (87)Intermediate 318 (3) 311 (3) 7 (3)Heavy 401 (4) 394 (4) 7 (3)

Waist circumference≤94.0 (M); ≤80.0(W) 6322 (63) 6211 (63) 111 (44)

<0.00194–101.99 (M);80–88 (W) 2241 (22) 2174 (22) 67 (26)

>102 (M); >88 (W) 1511 (15) 1436 (15) 75 (30)Hypertension, 𝑛 (%)No 8043 (80) 7891 (80) 152 (60)

<0.001Yes 2062 (20) 1961 (20) 101 (40)∗Chi2-tests for comparison of proportions, 𝑡-test for comparisons of contin-uous variable.

In the age, sex, and part of study adjusted Cox propor-tional hazard regressionmodel, job strainwas associatedwitha 38% increased risk of AF when compared to all others,although the association was not statistically significant (HR1.38, 95% CI 0.95–2.00) (Table 2). None of the investi-gated potential confounding or mediating factors changed


Table 2: The estimated association between job strain and the risk of atrial fibrillation. Hazard ratios (HR) with 95% confidence intervals(95% CI).

Work-related stress HR (95% CI)∗ HR (95% CI)∗ HR (95% CI)∗

Complete follow-up, 253 events First 10.7 years of follow-up, 165 events Follow-up after 10.7 years, 88 eventsNo strain 1 (ref) 1 (ref) 1 (ref)Job strain 1.38 (0.95–2.00) 1.11 (0.67–1.83) 1.93 (1.10–3.36)Low strain 1 (ref) 1 (ref) 1 (ref)Passive 1.08 (0.79–1.49) 1.05 (0.71–1.54) 1.16 (0.66–2.03)Active 1.21 (0.87–1.67) 1.22 (0.82–1.83) 1.20 (0.67–2.09)High strain 1.50 (0.99–2.27) 1.19 (0.69–2.06) 2.13 (1.13–4.04)∗Adjusted for age, sex, and part of study.

Prob

abili

ty o

f bei

ng A

F-fre

e

Job strain

Follow-up time (days)

1.00

0.99

0.98

0.97

0.96

0 1000 2000 3000 4000 5000 6000 7000

YesNo

Figure 1: Unadjusted Kaplan-Meier plot, job strain versus no strain.

the estimated HRwith more than 10% and were therefore notincluded in the regressionmodel. No clear effectmodificationby sex was observed (HR 1.42, 95% CI 0.94–2.14 for men,HR 1.22, 95% CI 0.51–2.92 for women, 𝑃 value for interaction= 0.75). When using the four demand-control categories,high strain was associated with a 50% increased risk ofAF compared with the low strain group, the result beingborderline significant (HR 1.50, 95% CI 0.99–2.27) (Table 2).

When inspecting the crude Kaplan-Meier plot, it wasobserved that those with job strain had a slightly betterprobability of beingAF-free during the first years of follow-upas compared with the nonstrain group, but the curves crossedat approximately 10.7 years after baseline (Figure 1). This ledus to do stratified analysis, splitting the follow-up period at10.7 years after baseline. The seemingly lower AF risk in thejob strain group during the first follow-up period was mainlyexplained by higher prevalence of job strain among womenthan men (14% versus 10%) and that the job strain group

Table 3: The estimated association between accumulated exposureto job strain and the risk of atrial fibrillation. Hazard ratios (HR)with 95% confidence intervals (95% CI), based on a subsample fromthe WOLF Norrland study population with baseline measure in1996–1998 (t1) and repeated measure in 2000–2003 (t2), 𝑛 = 3123.

HR(95% CI)∗

P valuefor trend

Subsample with repeatedmeasurements, 47 eventsUnexposed to job strain atboth t1 and t2 1 (ref) 0.06

Job strain at either t1 or t2 1.68 (0.83–3.40)Job strain at both t1 and t2 2.28 (0.70–7.44)

∗Adjusted for age and sex.

tended to be younger than the nonstrain group (mean age41.2 versus 42.7 years). After adjusting for age, sex, and studypart, the HR for job strain versus others was 1.11 (95% CI0.67–1.83) during the first part of the follow-up period. Inthe analysis based on the follow-up period after 10.7 years, itwas observed that job strain versus all others was significantlyassociated with the risk of AF in the age, sex, and study partadjusted models (HR 1.93 95% CI 1.10–3.36) (Table 2). Thesame pattern was seen when using the four demand-controlcategories, where the HR for the high strain group comparedwith low strain was 2.13 (95% CI 1.13–4.04) (Table 2).

For a subsample of theWOLFNorrland study populationwe had repeated measurements of job strain, measured atbaseline (1996–1998) and follow-up (2000–2003). Taking intoaccount the exposure to job strain at none (𝑛 = 2472, AF cases= 34), one (𝑛 = 527, AF cases = 10), or both measurementoccasions (𝑛 = 124, AF cases = 3), we observed an associationbetween job strain and AF risk in a dose-response manner(Table 3). However, as the number of participants and AFcases exposed at both occasions was few, the estimates wereimprecise.

4. Discussion

In this study, we observed an association between work-related stress, defined as job strain, and the risk of atrialfibrillation. The association was time-dependent and morepronounced at the end of the follow-up period. The risk of


AFwas approximately two times higher among those exposedto job strain compared to those unexposed during the latterpart of the follow-up period. In analyses taking repeatedmeasurements into account, our results were suggestive of adose-response relationship between accumulated exposure tojob strain and AF risk.

Published studies on the association between psychoso-cial stress in general and work-related stress in particular andAF are scarce. We are only aware of one previously publishedstudy on job strain and AF risk [25]. That recently publishedstudy by Toren et al. was also based on a Swedish sample butonly included men and used a job exposure matrix based onoccupation at baseline to measure job strain. They found a32% increased risk associatedwith being exposed to job strainversus all others (HR 1.32, 95%CI 1.003–1.75) [25], which is inaccordance with our overall HR estimate of 1.38. In anotherstudy, mental stress in terms of acute life stress was found tobe related to AF risk [24], and in a study conducted amongAF patients, mental stress was the most frequent reportedtriggering factor of AF attacks [23]. Some case reports linkingemotional stress to AF have also been published [31, 32]. Inaddition, different aspects of work-related stress, includingjob strain, effort-reward imbalance, and job insecurity, havebeen linked to increased risk of coronary heart disease [33–35].

The biological pathway between work-related stress andAF is not clear. Ectopic foci in pulmonary veins are rec-ognized as triggers of AF, and the processes leading to theonset of AF include atrial fibrosis, structural remodeling ofthe heart tissue, and inflammation [36]. Altered sympatheticand parasympathetic balance and neurohormonal activationhave also been proposed to play key roles in the developmentof AF [36, 37]. In a study on 77 AF patients, Bettoni andZimmerman showed an increase in adrenergic tone followedby a marked shift towards vagal predominance immediatelybefore the onset of paroxysmal AF [38], and Patterson etal. showed in an experimental study on dogs that both theparasympathetic and sympathetic nervous system have a rolein initiating and triggering pulmonary vein activity [39].Reactions to stress include several physiological responsesinvolving both the hypothalamic-pituitary-adrenal axis andthe autonomous nervous system [40, 41]. Responses includeincreased release of glucocorticoid hormones, such as corti-sol, and increased sympathetic activity, with increased releaseof adrenaline and noradrenaline. An effect on inflammationhas also been observed [40]. These factors are making a linkbetween psychosocial stress, including work-related stress,and AF plausible. Atrial fibrosis and structural remodelingdevelop over time andmay be asymptomatic for several years.This may explain our finding with a stronger associationbetween job strain and AF observed at the later part ofthe follow-up period and that accumulated exposure to jobstrain over time seems to be associated with higher risk ascompared to shorter episodes of exposure, although the exactmechanism behind this observation is not clear.

Our study has several strengths, including the prospectivedesign, being based on a large sample of working people,and including both men and women. A high participationrate and low internal dropout are further strengths. We used

a well-established measure on work-related stress, based onthe demand-control or job strain model, frequently used instudies on work-related stress and health related outcomes.Job strain was measured by self-report through a validatedquestionnaire [29, 42]. The outcome was defined throughnational registers with high quality and coverage [43]. Onemajor advantage with our study is that we had access torepeated measures of job strain for a subset of our studysample. We also had access to several potential confoundingfactors, which we could take into account in the analyses.Indeed, participants diagnosed with AF during the follow-upperiod were to a higher degreemale, older, smoker, and obeseand were more likely to suffer from hypertension, which isin accordance with previous studies [8, 11, 15, 18]. However,after adjusting for age, sex, and study part, taking into accountlife-style factors, obesity or hypertension did not change theestimated association in any substantial way. However, it isimportant to acknowledge that an unfavorable work situationmay affect life-style factors in a longitudinal perspective [44],potentially contributing to the association between long-termexposure to job strain and ill health.

Our study also has some limitations. Despite the largestudy sample and a rather long follow-up period (medianfollow-up time: 13.6 years), the number of AF cases wasrelatively low, limiting the power of the study. This isespecially true for the longitudinal analyses taking repeatedmeasures into account. Also, the majority of the incidentAF cases in our study had an unspecified AF diagnosis,preventing more detailed analyses of AF subtypes such asparoxysmal, persistent, and chronic AF. Furthermore, theexperience and perception of stress at work is a complex issue,and there are several ways of operationalizing work-relatedstress. Here, we used the most frequently utilized model, thejob strain model. However, there are several other modelsand operationalizations available covering other aspects ofwork-related stress, such as the effort-reward imbalance, jobinsecurity, and organisational injustice, which is not coveredin the present study. Evaluating other aspects of work-relatedstress in relation to AF in addition to the job strain modelwill yield a more complete picture of the association betweenwork-related stress and AF.

5. Conclusion

Our study lends some support to the hypothesis that work-related stress, defined as job strain, is related to the develop-ment of AF over time. Our results suggest that the associationmay be time-dependent and that long-term exposure to jobstrain may be more strongly associated with AF risk thanshorter bouts of exposure.


The authors declare no conflict of interests.

AcknowledgmentsThe study has received funding from the Swedish Heart andLung Association and Futurum, the Academy for Healthcare,Jonkoping County Council.


References

[1] WorldHealthOrganization,Global Atlas onCardiovascularDis-ease Prevention and Control, edited by S. Mendis, P. Puska andB. Norrving, World Health Organization, Geneva, Switzerland,2011.

[2] A. J. Camm, G. Y. H. Lip, R. de Caterina, and et al, “2012focused update of the ESC Guidelines for the management ofatrial fibrillation: an update of the 2010 ESC Guidelines for themanagement of atrial fibrillation—developed with the specialcontribution of the European Heart Rhythm Association,”Europace, vol. 14, no. 10, pp. 1385–1413, 2012.

[3] E. Z. Soliman, M. M. Safford, P. Muntner et al., “Atrial fibril-lation and the risk of myocardial infarction,” JAMA InternalMedicine, vol. 174, no. 1, pp. 107–114, 2014.

[4] D. A. Dulli, H. Stanko, and R. L. Levine, “Atrial fibrillation isassociated with severe acute ischemic stroke,” Neuroepidemiol-ogy, vol. 22, no. 2, pp. 118–123, 2003.

[5] T. Dalteg, E. Benzein, A. Sandgren, B. Fridlund, and D. Malm,“Managing uncertainty in couples living with atrial fibrillation,”The Journal of Cardiovascular Nursing, vol. 29, no. 3, pp. E1–E10,2014.

[6] H. Ekblad, H. Ronning, B. Fridlund, and D. Malm, “Patients’well-being: experience and actions in their preventing and han-dling of atrial fibrillation,” European Journal of CardiovascularNursing, vol. 12, no. 2, pp. 132–139, 2013.

[7] P. J. McCabe and S. A. Barnason, “Illness perceptions, copingstrategies, and symptoms contribute to psychological distress inpatients with recurrent symptomatic atrial fibrillation,” Journalof Cardiovascular Nursing, vol. 27, no. 5, pp. 431–444, 2012.

[8] S. S. Chugh, R. Havmoeller, K. Narayanan et al., “Worldwideepidemiology of atrial fibrillation: a global burden of disease2010 study,” Circulation, vol. 129, no. 8, pp. 837–847, 2014.

[9] M. Zoni-Berisso, F. Lercari, T. Carazza, and S. Domenicucci,“Epidemiology of atrial fibrillation: European perspective,”Clinical Epidemiology, vol. 6, pp. 213–220, 2014.

[10] U. Friberg, P. J. Svensson, and M. von Arbin, Atrial fibrillation.Prevalence and Risk for Stroke, SBU, The Swedish Council onHealth Technology Assessment, Stockholm, Sweden, 2013.

[11] A. S. Go, E. M. Hylek, K. A. Phillips et al., “Prevalence ofdiagnosed atrial fibrillation in adults: National implications forrhythm management and stroke prevention: the anticoagula-tion and risk factors in atrial fibrillation (ATRIA) study,” Journalof the American Medical Association, vol. 285, no. 18, pp. 2370–2375, 2001.

[12] J. Heeringa, D. A. M. van der Kuip, A. Hofman et al.,“Prevalence, incidence and lifetime risk of atrial fibrillation: theRotterdam study,” European Heart Journal, vol. 27, no. 8, pp.949–953, 2006.

[13] W. B. Kannel, P. A. Wolf, E. J. Benjamin et al., “Prevalence,incidence, prognosis, and predisposing conditions for atrialfibrillation: population-based estimates,”The American Journalof Cardiology, vol. 82, no. 8, pp. 2N–9N, 1998.

[14] S. Rietbrock, E. Heeley, J. Plumb, and T. van Staa, “Chronicatrial fibrillation: incidence, prevalence, and prediction ofstroke using the congestive heart failure, hypertension, age >75,diabetes mellitus, and prior Stroke or transient ischemic attack(CHADS2) risk stratification scheme,” The American HeartJournal, vol. 156, no. 1, pp. 57–64, 2008.

[15] B. A. Schoonderwoerd, M. D. Smit, L. Pen, and I. C. van Gelder,“New risk factors for atrial fibrillation: causes of ‘not-so-loneatrial fibrillation’,” Europace, vol. 10, no. 6, pp. 668–673, 2008.

[16] M. F. Sinner, N. R. Tucker, K. L. Lunetta et al., “Integratinggenetic, transcriptional, and functional analyses to identify 5novel genes for atrial fibrillation,” Circulation, vol. 130, no. 15,pp. 1225–1235, 2014.

[17] B. Zoller, H. Ohlsson, J. Sundquist, and K. Sundquist, “Highfamilial risk of atrial fibrillation/atrial flutter in multiplexfamilies: a nationwide family study in Sweden,” Journal of theAmerican Heart Association, vol. 2, no. 1, Article ID e003384,2013.

[18] M. Rosiak, M. Dziuba, M. Chudzik et al., “Risk factors foratrial fibrillation: not always severe heart disease, not always so‘lonely’,” Cardiology Journal, vol. 17, no. 5, pp. 437–442, 2010.

[19] D. Conen and C. M. Albert, “Alcohol consumption and riskof atrial fibrillation: how much is too much?” Journal of theAmerican College of Cardiology, vol. 64, no. 3, pp. 290–292, 2014.

[20] S. C. Larsson, N. Drca, and A.Wolk, “Alcohol consumption andrisk of atrial fibrillation: a prospective study and dose-responsemeta-analysis,” Journal of the American College of Cardiology,vol. 64, no. 3, pp. 281–289, 2014.

[21] D. Karasoy, T. Bo Jensen, M. L. Hansen et al., “Obesity is arisk factor for atrial fibrillation among fertile young women: anationwide cohort study,” Europace, vol. 15, no. 6, pp. 781–786,2013.

[22] A. Rosengren, P. J.Hauptman,G. Lappas, L.Olsson, L.Wilhelm-sen, and K. Swedberg, “Big men and atrial fibrillation: effects ofbody size and weight gain on risk of atrial fibrillation in men,”European Heart Journal, vol. 30, no. 9, pp. 1113–1120, 2009.

[23] A. Hansson, B. Madsen-Hardig, and S. B. Olsson, “Arrhythmia-provoking factors and symptoms at the onset of paroxysmalatrial fibrillation: a study based on interviews with 100 patientsseeking hospital assistance,”BMCCardiovascular Disorders, vol.4, article 13, 2004.

[24] A. V. Mattioli, S. Bonatti, M. Zennaro, R. Melotti, and G.Mattioli, “Effect of coffee consumption, lifestyle and acute lifestress in the development of acute lone atrial fibrillation,”Journal of Cardiovascular Medicine, vol. 9, no. 8, pp. 794–798,2008.

[25] K. Toren, L. Schioler, M. Soderberg, K. W. Giang, and A.Rosengren, “The association between job strain and atrialfibrillation in Swedish men,” Occupational and EnvironmentalMedicine, vol. 72, no. 3, pp. 177–180, 2015.

[26] L. Alfredsson, N. Hammar, E. Fransson et al., “Job strain andmajor risk factors for coronary heart disease among employedmales and famales in a Swedish study on work, lipids andfibrinogen,” Scandinavian Journal of Work, Environment andHealth, vol. 28, no. 4, pp. 238–248, 2002.


[28] R. Karasek and T. Theorell, Healthy Work: Stress, Productivityand the Reconstruction of Working Life, Basic Books, New York,NY, USA, 1990.

[29] T. Theorell, “The demand-control-support model for studyinghealth in relation to the work environment—an interactivemodel,” in Behavioral Medicine Approaches to CardiovascularDisease Prevention, K.Orth-Gomer andN. Schneiderman, Eds.,pp. 69–85, Lawrence Erlbaum Associates, Mahwah, NJ, USA,1996.

[30] K. J. Rothman, S. Greenland, and T. L. Lash,Modern Epidemiol-ogy, Lippincott Williams &Wilkins, Philadelphia, Pa, USA, 3rdedition, 2008.


[31] R. C. Ziegelstein, “Acute emotional stress and cardiac arrhyth-mias,”The Journal of the AmericanMedical Association, vol. 298,no. 3, pp. 324–329, 2007.

[32] D. Legallois, S. Gomes, A. Pellissier, and P. Milliez, “Medicalemotional stress-induced atrial fibrillation: my own personalexperience,” International Journal of Cardiology, vol. 167, no. 6,pp. e182–e183, 2013.

[33] M. Kivimaki, M. Virtanen, M. Elovainio, A. Kouvonen, A.Vaananen, and J. Vahtera, “Work stress in the etiology ofcoronary heart disease—ameta-analysis,” Scandinavian Journalof Work, Environment and Health, vol. 32, no. 6, pp. 431–442,2006.

[34] M. Kivimaki, S. T. Nyberg, G. D. Batty et al., “Job strain as a riskfactor for coronary heart disease: a collaborative meta-analysisof individual participant data,” The Lancet, vol. 380, no. 9852,pp. 1491–1497, 2012.

[35] M. Virtanen, S. T. Nyberg, G. D. Batty et al., “Perceived jobinsecurity as a risk factor for incident coronary heart disease:systematic review and meta-analysis,” British Medical Journal,vol. 347, Article ID f4746, 2013.

[36] B. Aldhoon, V. Melenovsky, P. Peichl, and J. Kautzner, “Newinsights into mechanisms of atrial fibrillation,” PhysiologicalResearch, vol. 59, no. 1, pp. 1–12, 2010.

[37] P. Taggart, M. R. Boyett, S. Logantha, and P. D. Lambiase,“Anger, emotion, and arrhythmias: from brain to heart,” Fron-tiers in Physiology, vol. 2, article 67, 2011.

[38] M. Bettoni and M. Zimmermann, “Autonomic tone variationsbefore the onset of paroxysmal atrial fibrillation,” Circulation,vol. 105, no. 23, pp. 2753–2759, 2002.

[39] E. Patterson, S. S. Po, B. J. Scherlag, and R. Lazzara, “Triggeredfiring in pulmonary veins initiated by in vitro autonomic nervestimulation,” Heart Rhythm, vol. 2, no. 6, pp. 624–631, 2005.

[40] A. Hansel, S. Hong, R. J. A. Camara, and R. von Kanel,“Inflammation as a psychophysiological biomarker in chronicpsychosocial stress,” Neuroscience and Biobehavioral Reviews,vol. 35, no. 1, pp. 115–121, 2010.

[41] A. Steptoe andM.Kivimaki, “Stress and cardiovascular disease,”Nature Reviews Cardiology, vol. 9, no. 6, pp. 360–370, 2012.

[42] H. S. Chungkham, M. Ingre, R. Karasek, H. Westerlund, andT. Theorell, “Factor structure and longitudinal measurementinvariance of the demand control support model: an evidencefrom the Swedish Longitudinal Occupational Survey of Health(SLOSH),” PLoS ONE, vol. 8, no. 8, Article ID e70541, 2013.

[43] J. F. Ludvigsson, E. Andersson, A. Ekbomet al., “External reviewand validation of the Swedish national inpatient register,” BMCPublic Health, vol. 11, article 450, 2011.

[44] E. I. Fransson, K. Heikkila, and S. T. Nyberg, “Job strain as arisk factor for leisure-time physical inactivity: an individual-participant meta-analysis of up to 170,000 men and women: theIPD-work consortium,” American Journal of Epidemiology, vol.176, no. 12, pp. 1078–1089, 2012.

Research ArticleBurnout Is Associated with Reduced Parasympathetic Activityand Reduced HPA Axis Responsiveness, Predominantly in Males

Wieke de Vente,1,2 Jan G. C. van Amsterdam,3 Miranda Olff,4

Jan H. Kamphuis,1 and Paul M. G. Emmelkamp1,5,6

1Department of Clinical Psychology, University of Amsterdam, Weesperplein 4, 1018 XA Amsterdam, Netherlands2Research Institute Child Development and Education, University of Amsterdam, P.O. Box 15776,1001 NG Amsterdam, Netherlands3Amsterdam Institute for Addiction Research, Academic Medical Center, P.O. Box 75867,1070 AW Amsterdam, Netherlands4Center for Psychological Trauma, Department of Psychiatry, Academic Medical Center, P.O. Box 22660,1100 DD Amsterdam, Netherlands5Netherlands Institute for Advanced Study, Meijboomlaan 1, 2242 PR Wassenaar, Netherlands6The Center for Social and Humanities Research, King Abdulaziz University, P.O. Box 80 202,Jeddah 21589, Saudi Arabia

Correspondence should be addressed to Wieke de Vente; [email protected]

Received 15 January 2015; Accepted 27 May 2015

Academic Editor: Maureen F. Dollard

Copyright © 2015 Wieke de Vente et al.This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

There is mounting evidence that burnout is a risk factor for cardiovascular disease (CVD). Stress-related dysregulation of thesympathetic and parasympathetic system and the hypothalamic pituitary adrenal (HPA) axis may explain the enhanced risk forCVD. To test this hypothesis, 55 patients (34 males and 21 females) with burnout on sickness absence and 40 healthy participants(16 males and 24 females) were exposed to a psychosocial stressor consisting of mental arithmetic and public speech. Physiologicalvariables (i.e., blood pressure, heart rate, cardiac output, vascular resistance, cortisol, and alpha-amylase) were measured. Basallevels, reactivity, and recovery were compared between groups. In male patients, baseline systolic blood pressure was higher,whereas basal alpha-amylase and cortisol reactivity were lower than in healthy males. In female patients, a tendency for lowerbasal cortisol was found as compared to healthy females. Furthermore, reduced basal heart rate variability and a trend for elevatedbasal cardiac output were observed in both male and female patients. Burnout is characterised by dysregulation of the sympatheticand parasympathetic system and the HPA axis, which was more pronounced in males than in females. This study further supportsburnout as being a risk factor for CVD through dysregulation of the sympathetic and parasympathetic system and the HPA axis.

1. Introduction

Burnout is a state that results from prolonged exposure towork-related stressors and goes along with health complaints(see Lindblom et al. [1], for an overview; see [2, 3]). Burnoutcomplaints include emotional exhaustion, negative attitudestowards work, and a sense of diminished competence to fulfilthe demands posed by the job [2, 3]. Burnout is accompaniedby distress including affective (e.g., depressed mood), phys-ical (e.g., fatigue), cognitive (e.g., concentration problems),

and behavioural (e.g., sleeping problems) symptoms [3].Burnout, and more generally work-related stress, is a riskfactor for cardiovascular disease (CVD; see Melamed et al.[4] and Belkic et al. [5], for reviews; see [6]). In this studywe aimed to identify physiological stress mechanisms thatmay explain this relationship by assessing indices of mainphysiological stress systems, that is, the sympathetic system,the parasympathetic system, and the hypothalamic pituitaryadrenal (HPA) axis [7, 8], associated with a clinical levelof burnout. Assessing indices of various physiological stress


http://dx.doi.org/10.1155/2015/431725


systems allows for an initial step in the development of anintegrative view on physiological stress adaptation that maymediate the burnout-CVD association.

Physiological stress mechanisms that are hypothesised tomediate the association between stress and CVD by promot-ing adverse health processes associated with CVD, includ-ing the metabolic syndrome and atherosclerosis, are sus-tained enhanced sympathetic activity, reduced parasympa-thetic vagal activity, enhanced sympathetic reactivity and/ordelayed sympathetic recovery after stressor exposure, anddys-regulation of the HPA axis [4, 5, 9–14]. Support for an associ-ation between these physiological mechanisms in associationwith CVD has indeed been obtained. For example, in theirreview, Palatini and Julius [13] report support for sustainedenhanced sympathetic activity in the development of CVD.Furthermore, evidence for a role of reduced parasympatheticvagal activity in CVD development has been found in a largepopulation study (𝑁 = 14.672) [15]. In addition, two reviews[16, 17] report support for a relationship between enhancedsympathetic reactivity and development of CVD, includingthe metabolic syndrome and atherosclerosis. Finally, a bodyof evidence associates dysregulation of the HPA axis withCVD (see Melamed et al. [4], for a review).

However, results on sympathetic and parasympatheticactivity and HPA axis activity in association with work-related stress have been mixed (e.g., see Danhof-Pont etal. [18] for a review; see [4, 19–26]). The null-findings thatwere found in several studies including relatively healthysamples or employees that were still able to carry out theirwork despite the presence of stress complaints may be partlydue to the healthy worker effect or restriction of range. Toprevent selection bias due to inclusion of solely healthy, hardy,individuals and to ensure that exposure to stressors hadresulted in a serious state of distress, we chose to includea clinical burnout sample consisting of employees that hadcalled themselves sick because of severe work-related stresscomplaints and compared their physiological profile withthat of a healthy reference sample. Selection of a clinicalsample also ensured that stressor exposure duration had beensufficient to expect physiological stress adaptation. Anothersource of mixed results may be inconsistent consideration ofgender differences in physiological stress responses. There isevidence to suggest that cardiovascular and neuroendocrineresponding while under stress differs between the sexes dueto biological (e.g., hormonal) and psychological differences(e.g., appraisal and coping) [27–29]. The generally lowerphysiological stress-reactivity in women between pubertyand menopause in comparison to men may conceal physio-logical adaptation in combined samples. Moreover, as genderdifferences have been previously demonstrated in the relationbetween job strain and CVD risk [5], physiological profilesmay differ between males and females with burnout. Wetherefore chose to assess gender differences in physiologicaladaptation associated with burnout.

To our knowledge, our previous study [30] on physio-logical adaptation in burnout was the first that combinedautonomic and neuroendocrine stress indices in a clinicalsample. In this study, we found support for a hyperactivephysiological state as evidenced by elevated heart rate in

rest and during a psychosocial stressor and elevated cortisollevels at the moment of awakening in the patient group withburnout complaints. Our results regarding stress-reactivityduring a psychosocial stressor were inconclusive though, andour sample size did not allow for subgroup-analyses based ongender. The current study replicates and extends our earlierwork [30] by including a larger sample enabling the assess-ment of gender differences in stress-responsiveness. Further-more, with the aim to further elucidate the physiologicalmechanisms that may explain the association between work-related stress and CVD, we included additional measuresrelevant for CVD-promoting processes, that is: (a) cardiacoutput, (b) vascular resistance, (c) heart rate variability, and(d) alpha-amylase. Cardiac output and vascular resistancereflect well the haemodynamic function and they are themain determinants of blood pressure. More importantly,cardiac output and vascular resistance enable detectionof physiological adaptation related to the development ofhypertension that may not be detected by measuring bloodpressure alone [31]. Rapid, beat-to-beat, heart rate variability,an accepted measure of cardiac parasympathetic or vagalactivation [32, 33], was added to obtain a relatively puremeasure of parasympathetic activation. Alpha-amylase wasadded as an additional measure of change in the sympa-thetic tone when studied during a psychosocial stressor[34].

In sum, our aim was to examine physiological mecha-nisms thatmaymediate the association betweenwork-relatedstress and CVD. Initial support for a mediating role of physi-ological mechanisms would be to demonstrate the presenceof adverse cardiovascular and neuroendocrine profiles inpatients with burnout, which we addressed in this study.We hypothesised that burnout would be associated withsympathetic predominance in the sympathetic vagal balanceand dysregulation of the HPA axis. Since Kudielka et al. [21]stated that ongoing chronic stress is generally associated witha hyperactive HPA axis and our sample was recruited withinweeks to a few months after having called themselves sick,we expected to find support for a hyperactive rather than ahypoactive HPA axis. To test our hypotheses, we compareda clinical sample with burnout with a healthy referencesample on indices of sympathetic activity, parasympatheticactivity, and HPA axis activity. Since burnout is not includedas diagnostic classification category in the DSM-IV [35] orICD-10 [36], the criteria for undifferentiated somatoformdisorder (DSM-IV) or neurasthenia (ICD-10) are commonlyused, with the added criterion that the main cause of thesymptoms is work-related [37]. Comparisons were madeduring rest situations aswell as during a psychosocial stressor.For reasons of ecological validity, we chose a psychosocialstressor consisting of a task that required mental effort(mental arithmetic) and social performance (speech task).We predicted higher levels of basal heart rate, blood pressure,alpha-amylase, and cortisol and a lower heart rate variabilityin burnout patients compared to healthy individuals. Second,we predicted larger reactivity and less recovery of these mea-sures in patients compared to healthy individuals.Third, sincemales were expected to demonstrate larger physiological


stress-reactivity as indicated by cardiovascular and HPA axisindices [27–29], we predicted to find more between-groupdifferences for males than for females.

2. Methods

2.1. Participants. Eighty-five patients were recruited throughoccupational physicians, general practitioners, and by self-referral to participate in a RCT about treatment efficacyof stress-management training (e.g., de Vente et al. [38]).Twenty-seven patients did not fulfil the inclusion criteria(e.g., having major depression as a primary diagnosis) orrefused to take part in the study. Of the remaining 58 patients,43were recruited through occupational physicians, 3 throughgeneral practitioners, and 12 by self-referral to our study.Eligibility was based on a telephone screening interview bya clinical psychologist that assessed the presence of work-related burnout complaints. The screening interview was fol-lowed by an intake interview, in which a semistructured diag-nostic interview was administered face-to-face by a clinicalpsychologist and the patient completed the Beck DepressionInventory (BDI) [39]. During the semistructured interviewthe complaint history was assessed and the Composite Inter-national Diagnostic Interview (CIDI) [40] was administered.Inclusion criteria were (1) endorsement of the symptomsof neurasthenia, that is, continuous mental and/or physicalfatigue and increased fatigability, and at least two otherstress complaints out of the following: dizziness, dyspepsia,muscular aches or pains, tension headaches, inability to relax,irritability, and sleep disturbance; (2) a primary role of (a)work-related stressor(s) in the development of complaints asjudged by the patient, the referring clinician, and the clinicalpsychologist; and (3) presence of impaired daily functioningas indicated by (partial) sickness absence which had lasted atleast two weeks but less than six months. Exclusion criteriawere (1) a primary diagnosis of major depression, socialphobia, panic disorder, somatoform disorder other thanundifferentiated, posttraumatic stress disorder, obsessive-compulsive disorder, hypomania, or psychotic disorders,all as assessed with the CIDI [40]; (2) severe depressivecomplaints (i.e., conservatively defined as ≥25 on the BDI);(3) a traumatic event in the past six months; and (4) amedical condition that could better account for the fatigue(e.g., diabetes); (5) excessive alcohol and/or drug use; and (6)pregnancy. The current physiological study was a part of acomprehensive project about psychological and physiologicalaspects of work-related stress (e.g., de Vente et al. [38]). Forthis study, patients were refunded for their travel expensesand received a printed report of their baseline blood pressureand heart rate.

As a reference group, forty healthy individuals wererecruited by flyers in public places (e.g., libraries and super-markets; 𝑛 = 29) and among part-time working psychologystudents (𝑛 = 11). They were screened by telephone.Participants in good physical health and working for at least16 hours a week were included in the study. Exclusion criteriawere (1) psychiatric illness as determined by the CIDI [40];(2) currently taking sick leave; (3) a traumatic event in thepast six months; (4) a history of immune, diabetic, or other

medical disease causing fatigue; (5) excessive alcohol and/ordrug use; and (6) pregnancy. Healthy participants were paid15 euro and received a printed report of their baseline bloodpressure and heart rate values after attending the laboratorysession and completion of the questionnaires.

2.2. Materials2.2.1. Acute Psychosocial Stressor. To study physiological reac-tivity and recovery, participants were exposed to an acutepsychosocial stressor consisting of a speech preparation task,a mental arithmetic task, and a speech task (see Figure 1 forthe complete procedure). The speech preparation task con-sisted of preparing a story about a dramatic social situationin which the participant was unfairly accused of causingdamage to the property of others. The mental arithmetic taskentailed continuous attention-demanding addition, subtrac-tion, multiplication, and division. For the speech task, theprepared story had to be told in front of the camera andthe participant was told that the tape would be analysed.Psychosocial stress procedures have been demonstrated toenhance perceived stress and result in cardiovascular andneuroendocrine reactions (e.g., [30, 41–44]).

2.2.2. Cardiovascular Assessment. Heart rate (HR) and bloodpressure (BP) were measured by continuous measurement offinger BP using a Finapres (Ohmeda Finapres type 2300E,Blood Pressure Monitor) and the software Vsrrp98 [45].Systolic blood pressure (SBP), diastolic blood pressure (DBP),HR, cardiac output (CO), and total peripheral resistance(TPR) were calculated using the software Beatscope (version1.1 [46]). Heart rate variability was calculated using the rootmean square of successive differences (RMSSD) of interbeatintervals (IBIs): √(1/𝑛Σ(IBIi − IBIi−1)

2), reflecting mainlyhigh frequency power, and is therefore an adequate measureof the cardiac vagal tone [32, 33]. IBIs were defined as thenumber ofmilliseconds between peaks of subsequent systolesin the photoplethysmographic signal, analysed with Vsrrp98(version 5.4b). The RMSSD based on the IBIs determined inthe photoplethysmographic signal was called estimated heartrate variability (EHRV). EHRV based on the photoplethys-mographic signal appears to be a valid measure of HRV[47]. Before analyses, the photoplethysmographic signal wasinspected visually and artefacts (e.g., movement) and ectopicbeats were removed. Mean values of cardiovascular measureswere calculated per five minutes. Mean values during the lastfive minutes of the prestressor baseline phase (see Figure 1)were used as baseline values, indicative of basal functioning.For reactivity, mean values during the stress-inducing taskswere related to the prestressor baseline. For recovery, meanvalues of the first to third recovery phase (i.e., 6–10min., 11–15min., and 21–25min. poststressor) were also related to theprestressor baseline.

2.2.3. Neuroendocrine Measures and Protocols. Alpha-am-ylase and cortisol were determined in saliva collectedas described by Navazesh [48]. Accordingly, participantsrefrained from swallowing for a period of four minutes,allowing the saliva to accumulate in the floor of the mouth.


Rest[baseline]

Speechpreparation

Mentalarithmetic

Speechtask

Rest[recovery]

1 2 3 4 5

1st 3rdCVMSC and

Minutes

0 20 55

2nd

MQ

−15

Figure 1: Time diagram of the psychosocial stress procedure. Note: CVM: cardiovascular measurements and SC and MQ: saliva collectionfor endocrine measures and mood questionnaire.

The saliva is spitted out into a cup every 60 s. The collectionstarts with the instruction to void the mouth of saliva byswallowing. Fifteen minutes before the first saliva collection,participants rinse their mouths with water.

Saliva samples were stored on ice until the end of theexperiment. Immediately after the session, that is, within90min. after collection, saliva was homogenised using avortex mixer and clarified by centrifugation (10,000×g. for4min). Aliquots (0.5mL) with clear supernatant were storedat −20∘C until analysis.

Alpha-amylase activity was assayed photometrically(Roche, Almere, Netherlands) after 500-fold dilution using5 ethylidene-G

7

PNP as substrate. The lower detection limitfor amylase was 3U/L.The amount of free cortisol was deter-mined using enzyme-immunoassay (EIA). Kits were pur-chased from Diagnostic System Laboratories (DSL, Veghel,Netherlands). The sensitivity of cortisol assay was 1 ng/mL.All samples were assayed in duplo. Intra-assay variabilityof alpha-amylase and cortisol was 0.4%–2% and 2%–10%,respectively.

Means of the first (−4min. in relation to the start of thestressor) and second (+5min.) saliva samples were used asresting values, indicative of basal functioning (see Figure 1).For alpha-amylase, the third (+20min., i.e., immediately aftercessation of the stressor) saliva sample indicated reactivityand the fourth (+35min.) and fifth (+50min.) saliva sam-ples recovery. For cortisol, the third (+20min.) and fourth(+35min.) saliva samples indicated reactivity, and the fifth(+50min.) indicated sample recovery.

2.2.4. Psychological Measures and Background Variables.Burnout complaints were measured with the MaslachBurnout Inventory-General Survey (MBI-GS) [49], whichconsists of three subscales: emotional exhaustion (5 items),depersonalisation (4 items), and professional competence (6items). Items were scored on 7-point Likert scales (0 = neverto 6= always/daily) andmean subscale scoreswere calculated.Higher scores reflect higher levels of emotional exhaustion,distant/cynical attitudes towardswork, and professional com-petence. Cronbach’s alpha was 0.85 for emotional exhaus-tion, 0.81 for depersonalisation, and 0.75 for professionalcompetence in the patient sample. For healthy participants,Cronbach’s alpha was 0.83, 0.73, and 0.68, respectively.

Distress complaints were defined as fatigue, depression,anxiety, and stress complaints. Fatigue wasmeasuredwith theChecklist Individual Strength (CIS) [50].TheCIS consisted of20, whichwere scored on a 7-point Likert scale (1 = false to 7 =

true). Lower scores indicate lower levels of fatigue. Cronbach’salpha in the current sample was 0.90 in both the patientgroup and healthy group. Depression, anxiety, and stress weremeasured with the Depression, Anxiety, and Stress Scales(DASS) [51]. The DASS comprises three subscales of 14 itemseach, referring to depressive, anxiety, and stress complaints.Severity of complaints during the past week is rated on 4-point Likert scales (0 = not at all/never applicable to 3 = verymuch/most of the time applicable). Higher scores representhigher levels of complaints. Cronbach’s alpha was 0.93 fordepression, 0.80 for anxiety, and 0.92 for stress in the patientsample, and 0.83 for depression, 0.72 for anxiety, and 0.94 forstress in the healthy sample.

Mood during the psychosocial stress procedure wasmeasured by the Profile of Mood Scale (POMS) [52]. TheTension (6 items) andAnger (7 items) subscales were selectedto measure aspects of negative affect, indicative of subjectivestress induction. Items were scored on a five-point scale(0 = not at all to 4 = very much). Higher scores on theTension and Anger subscales are indicative of more ten-sion and anger, respectively. Cronbach’s alpha at the firstadministration during the psychosocial stress procedure (i.e.,MQ1; see Figure 1) was 0.86 for Anger and 0.82 for Tensionin patients, and 0.68 for Anger and 0.74 for Tension inhealthy participants.Moodquestionnaireswere administeredfive times along with saliva collections (see Figure 1). Thesemood-dimensions have been previously used to measure thesubjective response during stress-inducing tasks [30, 44] andwere used as a manipulation check in the present study.

The relevant control variables for cardiovascular andneuroendocrine parameters such as smoking, hours of sleep,and body mass index (BMI) were assessed by questionnaire.Womenwere also asked to report onmenstrual phase, the useof oral contraceptives, and pre-/postmenopausal status.

2.3. Procedure. The ethics committee of the Department ofPsychology, University of Amsterdam, approved the researchprotocol and all participants gave written informed con-sent. Questionnaires regarding biographical information andburnout and distress complaints were completed at homeduring the week before the psychosocial stress procedurein the laboratory. To control for time of the day effects,all psychosocial stress procedures took place between 14.00and 16.30 hr. Participants were asked to refrain from eating,smoking, and coffee and tea consumption for at least one hourbefore the start of the experiment.Thebloodpressure cuffwasattached to the nondominant arm and the arm remained at


Table 1: Characteristics of patients and healthy participants listed for males and females separately [M (SD)/frequency (%)].

Males FemalesPatient (𝑛 = 34) Healthy (𝑛 = 16) Patient (𝑛 = 21) Healthy (𝑛 = 24)

Age (years) 42.79 (9.70) 38.00 (9.44) 37.95 (9.13) 37.42 (10.09)Education (1 = primary school, 6 = university) 3.06 (1.39)a 4.25 (1.18)a 4.29 (1.45) 4.08 (1.38)Employment (hrs/wk) 38.71 (2.98)a 31.63 (9.89)a 32.52 (5.75) 28.00 (8.09)Smoker (yes/no) 7/27 (21/79) 4/12 (25/75) 4/17 (19/81) 4/18 (18/82)Sleep duration (hours) 7.59 (1.12) 7.93 (0.47) 7.35 (1.58) 7.87 (1.47)Body mass index (kg/m2) 25.92 (3.27)a 23.55 (3.17)a 23.79 (5.09) 23.19 (2.74)Emotional exhaustion (MBI-GS, range: 0–6)b 4.44 (1.17) 1.25 (0.84) 4.09 (1.43) 1.32 (0.82)Depersonalisation (MBI-GS, range: 0–6)b 2.82 (1.56) 1.23 (0.77) 3.21 (1.28) 1.57 (1.02)Professional competence (MBI-GS, range: 0–6) 3.81 (1.09) 3.98 (0.89) 3.67 (0.92) 3.76 (0.92)Fatigue (CIS, range: 20–140)b 111.00 (16.79) 43.19 (17.41) 98.83 (19.66) 54.75 (20.75)Anxiety (DASS, range: 0–42)b 8.04 (6.34) 2.19 (2.40) 7.05 (4.06) 2.63 (3.10)Depression (DASS, range: 0–42)b 15.03 (8.31) 3.19 (3.29) 11.21 (6.69) 4.46 (3.78)Stress (DASS, range: 0–42)b 20.40 (9.16) 4.75 (4.36) 17.56 (6.70) 8.71 (8.57)Note: MBI-GS: Maslach Burnout Inventory—General Survey; CIS: Checklist Individual Strength; DASS: Depression, Anxiety, and Stress Scales. a𝑝 < 0.05;bbetween-group differences stratified by gender: 𝑝 < 0.001.

approximately heart level throughout the session. To preventpulse dampening, the Finapres was switched off for threeminutes during the first and fourth saliva collection. Dur-ing the whole experimental session, participants remainedseated. The questionnaire regarding control variables such assmoking was completed at the start of the session.

2.4. Statistical Analyses. To test between-group differencesin resting values indicative of basal levels, between-groupdifferences in baseline values of cardiovascular and neu-roendocrine measures were assessed by analyses of vari-ance (ANOVA), using a one between-subject (group) factordesign.

To test between-group differences in reactivity-recoveryand in overall mean activity during the psychosocial stressprocedure, mood, cardiovascular, and neuroendocrine reac-tivity and recovery during the psychosocial stress procedurewere examined with ANOVA for repeated measures, usinga one within- (time), one between- (group) subject factordesign. When the assumption of sphericity was violated,Greenhouse-Geisser-corrected results were reported, result-ing in somewhat more conservative testing. When time-group interactions were statistically significant, simple con-trasts were employed to explore differences in reactivity andrecovery as compared to baseline.

As gender differences in cardiovascular, cortisol, andalpha-amylase activity have been reported previously (e.g.,see Kajantie and Phillips [27], for a review; see [53–55]),effect modification of gender was investigated. When effectmodification was found, stratified results were presented.As age and BMI are known to be related to physiologicaloutcomes, they were added as covariates to all analysesconcerning cardiovascular and neuroendocrine outcomes.In addition, menstrual phase, oral contraceptive use, andmenopausal status are known to be related to cortisol [27,28] and were therefore added as covariates when analysingcortisol in women.

Because of positively skewed data for Anger, EHRV, andalpha-amylase, square root transformed data were analysed,resulting in approximately normal distributions of the data,and skewness and kurtosis were < |2|, except for the kurtosisof Anger which was <3. Two-sided test was performed,applying a significance level of 0.05. All analyses werecarried out using SPSS 20. For (square rooted) cardiovascularand (square rooted) neuroendocrine measures, outliers (i.e.,values ± >3 SDs of the mean) were removed; the numberof outliers was <5%. Some physiological data were missingdue to equipment problems (CO, TPR, EHRV; <1%) orinsufficient saliva (alpha-amylase, cortisol; 3%–8.5%).

3. Results

3.1. Sample Characteristics. Sample characteristics of thepatient and healthy group are presented in Table 1. Groupsdiffered on gender distribution, 𝜒2(1, 𝑛 = 95) = 4.42,𝑝 = 0.035. Furthermore, male patients were somewhatlower educated, 𝑡(49) = −3.06, 𝑝 = 0.004, had somewhathigher BMI, 𝑡(49) = 2.45, 𝑝 = 0.018, and were 7.1 hoursmore employed, 𝑡(16.2) = 2.80, 𝑝 = 0.013, than healthymales. Education did not appear to be a confounder inthe analyses of group differences in neuroendocrine andimmune measures. Hence, the presented results were notadjusted for education. Mean duration of sickness absencein patients was 8.58 (SD = 7.39) weeks. None of the healthyparticipants was on sickness leave. Three patients (5%) thatwere using beta-blocker antihypertensive medication wereexcluded from further analyses. Two patients (4%) usedantidepressive medication and four patients (7%) used ananxiolyticum. Five female patients (24%) were using oralcontraceptives. Healthy participants, except for eight women(33%) who used oral contraceptives, were medication-free.Three women in the patient group (14%) reported to be inthe menstrual phase (days 1–6), three (14%) in the follicular


phase (days 7–14), and 10 (48%) in the luteal phase (days 15–28). In the healthy group (missing: 𝑛 = 1), the numberswere eight (33%), four (17%), and seven (29%), respectively.Five patients (24%) and four healthy women (17%) reportedhaving passed their menopause. No statistically significantdifferences were found in menstrual phase distribution orpre-postmenopausal distribution.

Patients had significantly higher mean scores on allcomplaints than healthy participants; effect sizes (i.e., Cohen’s𝑑) were between 1.20 and 2.73; all 𝑝 values < 0.001. Anexception was noted for professional competence, for whichthe between-group effect was small (Cohen’s 𝑑= 0.10) and notstatistically significant. All patients scored above the validatedcut-off score for the emotional exhaustion subscale of theMBI-GS (i.e., >2.20, [49]), indicating severe exhaustion,and/or the validated cut-off score for CIS (i.e., >76, [56]),indicating severe fatigue. Male and female patients did notdiffer significantly on any of the complaints (all Cohen’s 𝑑’s< 0.50; all 𝑝 values > 0.10).

3.2. Mood during the Psychosocial Stress Procedure. Descrip-tive information on the subjective stress response during thepsychosocial stress procedure is provided in Figures 2(a)-2(b). Anger and Tension changed over time (𝐹-values > 10,𝑝 values < 0.001); they increased during the stressors (simplecontrasts Anger: MQ2 versus MQ1: 𝑝 < 0.001, and MQ3versusMQ1:𝑝 < 0.001; TensionMQ2 versusMQ1:𝑝 < 0.001,and MQ3 versus MQ1: 𝑝 < 0.001), supporting subjectivestress-induction by the acute psychosocial stress tasks.

3.3. Sympathetic, Parasympathetic, andNeuroendocrine Activ-ity, Reactivity, and Recovery. Figures 3(a)–3(g) show meansand standard errors of cardiovascular variables, and Figures4(a)–4(d) demonstrate means and standard errors of neu-roendocrine variables during the psychosocial stress proce-dure. Consistent with the literature, effect modification ofgender was found for SBP, alpha-amylase, and cortisol, forwhich stratified results are presented.

In Table 2, test results of baseline values and reactivity andrecovery during the psychosocial stress procedure relative tobaseline values are listed. At baseline, patients demonstratedhigher SBP (males only), lower EHRV, and lower alpha-amylase (males only) than healthy individuals. In addition,a trend was found for lower cortisol in female patients ascompared to healthy females.

All physiological measures changed over time during thepsychosocial stress procedure (F-values > 3.40, p values <0.05). The observed patterns were consistent with expectedactivation and recovery due to stress induction. An exceptionwas the observed pattern in cortisol in healthy females,which demonstrated a reduction instead of an increase afterbaseline. Similar to differences in baseline values, groupdifferences of mean values during the complete psychosocialstress procedure were found for SBP in males, EHRV, andalpha-amylase for males. A trend was found for CO, sug-gesting higher CO in patients. Since significant time × groupinteraction effects were absent, these main effects of groupsupport differences in basal activation, independent of acutestress-induction.

Differences in SBP-dynamics and the trends for DBP-dynamics and alpha-amylase-dynamics (females) during thepsychosocial stress procedure could not be attributed toeither reactivity or recovery (i.e., none of the simple con-trasts was statistically significant). For cortisol, healthy malesshowed earlier and stronger cortisol reactivity immediatelyafter cessation of the stressor (+20min.;𝑝 = 0.008) thanmalepatients. Mean reactivity immediately after the stressor ofhealthy males was 0.91 ng/mL (2.51 nmol/L), which is almostequal to the operational guideline for cortisol reactivity of1 ng/mL (2.76 nmol/L) [57]. A cortisol reaction could not beobserved in male patients at this moment. A trend was foundfor a similar pattern at the fourth measurement (+35min.;𝑝 = 0.080). Mean cortisol reactivity for healthy males at thispoint was 1.18 ng/mL (3.26 nmol/L), which clearly indicates acortisol secretory response. Mean cortisol reactivity for malepatients was 0.42 ng/mL (1.16 nmol/L), which does not crossthe secretory threshold. In addition, at the fifthmeasurement(+50min.), cortisol in healthy males had not returned tothe baseline level, in contrast to cortisol in male patients(𝑝 = 0.039). The trend for different cortisol-dynamics(females) during the psychosocial stress procedure could notbe attributed to either reactivity or recovery (i.e., none of thesimple contrasts was statistically significant).

4. Discussion

This study assessed whether burnout is characterised bydysregulation of the sympathetic vagal balance and theHPA axis that may explain the association between burnoutand CVD, taking into account gender differences in thesemechanisms. Support for predominance of the sympatheticsystem in burnout was obtained as indicated by elevatedbasal systolic blood pressure (males only), reduced basalheart rate variability, and a trend for elevated cardiac outputin the burnout group as compared to the healthy referencegroup.The reduction in basal alpha-amylase in male patientsis apparently inconsistent with the predicted sympatheticpredominance. The latter result is discussed in further detailbelow. In contrast to prediction, reduced cortisol reactivity toan acute psychosocial stressor was observed in male patients,which suggests hyporeactivity of the HPA axis, rather thanhyperactivity. Although the simultaneous predominance ofthe sympathetic system and hyporeactivity of the HPA axiswas not predicted, this pattern has been found previously inthe context of chronic stress, burnout, and vital exhaustionand gives rise to unfavourable alterations in immune func-tioning fostering risk for CVD [4].

Our study further highlights gender differences in car-diovascular functioning and in cortisol reactivity to a psy-chosocial stressor and in basal alpha-amylase in the con-text of burnout. Given nonsignificant gender differencesin levels of complaints, males seem to develop a moreadverse physiological profile as compared to females whenthey experience work-related stress. More specifically, in theadult life phase roughly between 25 and 60 years, which forfemales largely covers the premenopausal phase, enhancedcardiac sympathetic activation and hyporeactivity of theHPA-axis aremore evident amongmales than among females


−1

0

1

2

MQ2 MQ3 MQ4 MQ5

PatientHealthy

Ang

er

(a)

−2

−1

0

1

2

3

MQ2 MQ3 MQ4 MQ5

PatientHealthy

Tens

ion

(b)

Figure 2: Mean change scores and standard errors of negative affect during the psychosocial stress procedure using the score on MQ1, thatis, the mood questionnaire administered during rest, as a reference.

Table 2: Test results comparing prestressor resting values (ANOVA), mean values during the psychosocial stress procedure, and reactivityand recovery (ANOVA for repeated measures) between the patient and healthy group.

Restinga (group) Mean during sessiona (group) Reactivity and recoverya (group × phase)df ’s 𝐹 𝑝 df ’s 𝐹 𝑝 df ’s 𝐹 𝑝

SBPM 1,46 6.57 0.014 1,46 4.87 0.032 3.3, 153.6 3.11 0.024F 1,41 0.18 0.676 1,41 0.23 0.636 3.7, 151.4 0.68 0.596

DBP 1,90 0.44 0.507 1,90 1.13 0.291 3.5, 310.6 2.08 0.094HR 1,90 0.97 0.328 1,90 1.39 0.242 3.0, 268.8 0.44 0.723CO 1,89 1.71 0.195 1,89 3.81 0.054 3.7, 327.9 0.61 0.644TPR 1,89 0.85 0.360 1,89 1.31 0.256 3.4, 298.2 1.16 0.327EHRV 1,89 4.60 0.035 1,89 5.83 0.018 3.7, 332.8 0.50 0.726AA

M 1,41 7.14 0.011 1,41 6.60 0.014 2.0, 83.0 0.21 0.813F 1,39 0.23 0.634 1,39 <0.01 0.998 2.5, 98.8 2.47 0.077

CORTM 1,45 1.38 0.246 1,45 <0.01 0.997 2.0, 89.5 3.32 0.041F 1,34 2.92 0.097 1,34 1.22 0.278 2.1, 72.5 2.61 0.077

Note: Group:mean difference between the patient and the healthy group; group∗ phase: interaction effect of group× phase of the psychosocial stress procedure;SBP: systolic blood pressure; DBP: diastolic blood pressure; HR: heart rate; CO: cardiac output; TPR: total peripheral resistance; EHRV: estimated heart ratevariability; AA: alpha-amylase; CORT: cortisol; M: males; F: females. aAll analyses were adjusted for the covariates age, BMI, and gender. Cortisol analyses forfemales were also adjusted for menstrual phase, oral contraceptive use, and menopausal status. Statistically significant differences are presented in bold and arealso indicated in Figures 3 and 4, using superscripts.

with burnout. These gender-specific profiles associated withburnout in this life phase suggest that different mechanismsare at work in men and women with regard to cardiovascularrisks. More prominent adverse physiological profiles in menare in line with the findings of Belkic et al. [5], who showedmore consistent associations between job strain and CVD inmen than in women. Future research may enlighten whetherthe same adverse profile also emerges in postmenopausalfemales.

The unexpected finding of reduced basal alpha-amylase-level in male patients deserves further discussion, as it seemsto be inconsistent with the other indicators of a basal sym-pathetic predominance in the male patient group. One mayargue that alpha-amylase values were confounded by salivary

flow rate. To rule out this possibility of confounding,we testedwhether flow rate differed between groups and we reanalysedthe data by adjusting for flow rate. Indeed, in line with otherstudies (e.g., Bosch et al. [58]; Rohleder et al. [59]), no supportfor confounding was obtained as no differences in flow ratebetween healthy males and male patients were observed(data not shown), and similar outcomes were obtained whenanalyses were adjusted for flow-rate (results not shown). Amore plausible explanation is that negative affect may haveaffected basal alpha-amylase levels, as negative associationshave been found between resting alpha-amylase and negativeaffect as well as with avoidance behaviour (e.g., Fortunato etal. [60]). Hence, the group difference in alpha-amylase thatwe observed may reflect a group difference in a potentially


Patient MHealthy M

110

120

130

140

150

160

170

a

REST

PREP MA

SPCH

REC1

REC2

REC3

SBP

(mm

Hg)

(mal

es)

(a)

110

120

130

140

150

160

170

REST

PREP MA

SPCH

REC1

REC2

REC3

Patient FHealthy F

SBP

(mm

Hg)

(fem

ales

)

(b)

REST

PREP MA

SPCH

REC1

REC2

REC3

60

70

80

90

100

PatientHealthy

DBP

(mm

Hg)

(c)

60

65

70

75

80

85

REST

PREP MA

SPCH

REC1

REC2

REC3

PatientHealthy

HR

(bts/

min

)

(d)

REST

PREP MA

SPCH

REC1

REC2

REC3

4

5

6

7

PatientHealthy

CO (L

/min

)

(e)

1200

1300

1400

1500

1600

1700

1800

REST

PREP MA

SPCH

REC1

REC2

REC3

PatientHealthy

TPR

(dyn

·s/

cm5

)

(f)

Figure 3: Continued.


PatientHealthy

10

20

30

40

50

REST

PREP MA

SPCH

REC1

REC2

REC3

a

EHRV

(mse

c)

(g)

Figure 3:Means and standard errors of cardiovascular measures during the psychosocial stress procedure. Note: SBP: systolic blood pressure;DBP: diastolic blood pressure; HR: heart rate; CO: cardiac output; TPR: total peripheral resistance; EHRV: estimated heart rate variability;REST: baseline rest phase; PREP: speech preparation; MA: mental arithmetic; SPCH: speech task; REC: recovery phase; M: males; F: females.For EHRV the group difference remained statistically significant (p values < 0.05) throughout the experiment with no interaction effect ofgroup by time. The group difference during the experiment in CO was marginally significant (𝑝 = 0.054), with no interaction effect of groupby time. a𝑝 < 0.05 for group differences in prestressor levels (REST); see also Table 2, results in bold.

disease-state related basal affective state. It is also possiblethat, in a resting condition, alpha-amylase reflects a differentbalance between sympathetic and parasympathetic activationthan during a stress reaction. Alpha-amylase in resting con-ditions may be more strongly influenced by parasympatheticactivation than by sympathetic activation. Indeed, animalresearch has demonstrated that alpha-amylase also increasesunder parasympathetic activation (see Nater et al. [61], foran overview), which makes sense when one considers thatalpha-amylase was first identified as a digestive stimulatinghormone (e.g., Ramasubbu et al. [62]). Hence, the higheralpha-amylase level during rest in healthy males may be areflection of their higher basal parasympathetic activation.This reasoning then sustains that the observation of simul-taneous elevated basal cardiovascular activity, reduced basalheart rate variability, and reduced basal alpha-amylase inpatients is a result of reduced parasympathetic activity. Thestressor, in its turn then, evokes an additional sympatheticresponse that results in a similar alpha-amylase reaction inboth groups. Whether reduced alpha-amylase in a restingcondition should be interpreted more as an indication ofreduced parasympathetic activation rather than reducedsympathetic activation remains to be elucidated in futureresearch. It is interesting to note, though, that chronic stress inchildren has also been associatedwith reduced alpha-amylasein rest [63].

If a relation between burnout and adverse physiologicalprofiles is replicated, these results give rise to several clinicalimplications. First, in order to keep employees healthy,general awareness of adverse consequences of prolonged jobstress and signs of burnout may be increased by educationalinterventions in the workplace, for both employers andemployees. These types of interventions may also focus on

solutions to reduce stress. Second, monitoring job stressand other life stress may be added to regular physicalhealth checks in organizations. Our results provide an initialsuggestion to carefully monitor signs of (pre)hypertensionparticularly in male patients reporting work-related stress,as we found elevated SBP among males and a tendencyfor elevated CO in the group as a whole. Moreover, DBPwas 80mmHg in male patients (results not shown), whichis nowadays considered “prehypertension” [64]. Before pro-viding actual guidelines regarding preventive monitoring,further information has to be obtained about the chronicityof these adverse profiles. If changes turn out to be last-ing, or worse progressive, early intervention is required, ashypertension poses a main risk factor for cardiovasculardisease and mortality (e.g., Yusuf et al. [65]). Third, incase of sickness absence due to burnout, coworkers may beencouraged to keep in touch and support the sick colleague,since coworker support is suggested to promote reductionof burnout complaints [66]. Furthermore, a lack of socialsupport, that is, social isolation, has been associated withmore adverse physiological profiles [67] and enhanced riskof CVD [68].

Treatment implications for reduced cortisol responsive-ness (in contrast to reduced basal values) are not available. Inparticular with respect to burnout, insufficient informationabout the defects in cortisol responsiveness and their poten-tial spontaneous normalisation is available. However, thereis initial evidence that psychotherapy results in increases ofcortisol in burnout patients [69] and in the stress-relatedcondition PTSD [70]. Future studies may investigate whetherpsychotherapy or pharmacotherapy such as a low dose ofhydrocortisone normalizes cortisol responsiveness in burnoutas well.


75

125

175

225

275

325

375

SC1 SC2 SC3 SC4 SC5

a

Patient MHealthy M

Alp

ha-a

myl

ase (

U/m

L) (m

ales

)

(a)

75

125

175

225

275

325

375

SC1 SC2 SC3 SC4 SC5

Patient FHealthy F

Alp

ha-a

myl

ase (

U/m

L) (f

emal

es)

(b)

4

5

6

7

8

9

SC1 SC2 SC3 SC4 SC5

a

b

Patient MHealthy M

Cor

tisol

(ng/

mL)

(mal

es)

(c)

4

5

6

7

8

9

SC1 SC2 SC3 SC4 SC5

Patient FHealthy F

Cor

tisol

(ng/

mL)

(fem

ales

)

(d)

Figure 4: Means and standard errors of neuroendocrine measures during the psychosocial stress procedure. Note: SC: saliva collection; M:males; F: females. To convert salivary cortisol (ng/mL) to System International Units (nmol/L), multiply by 2.76. For alpha-amylase (males),the group difference remained statistically significant (p values < 0.05) throughout the experiment with no interaction effect of group bytime. a𝑝 < 0.05 for group differences in alpha-amylase prestressor levels (REST) or cortisol recovery (SC5; see also Table 2, results in bold).b𝑝 < 0.01, for group differences in cortisol reactivity (SC3; see also Table 2, result in bold).

Some methodological issues and limitations of this studydeserve consideration. Firstly, we did not recruit our patientsample and healthy sample in a similar way and did notmatchour samples on background variables. However, to compen-sate for suboptimal matching, we adjusted the analyses forrelevant covariates such as age, gender, body mass index,menstrual phase, oral contraceptive use, andmenopausal sta-tus, thus ruling out biased outcomes due to sample differenceson these variables. Secondly, since the majority of the patientsample consisted of employees working in small- tomedium-size enterprises, the reported levels of burnout complaintscannot be generalized to the total Dutch population. Thirdly,despite the fact that we found no statistically significantgender differences on complaints in the patient sample, themale patients seemed to overall report somewhat higher

levels of general distress complaints, such as depressive com-plaints. Consequently, it cannot be ruled out that the genderdifferences in physiological profiles reflect this tendency forhigher general distress complaints in males, rather thanreflecting essentially different physiological mechanisms inmales and females (e.g., those that are generally linked tohormonal differences [27]). Fourthly, as discussed above, thestress response of healthy females was below the criterion forHPA axis activation, which may have hindered detection ofdeviant values in women with burnout. Finally, we studiedthe acute physiological stress response in a laboratory setting,which may not necessarily generalize to a real life work-setting. In support of external validity of our manipulationof the acute physiological stress response is the fact thatwe used tasks that resemble daily challenges at work. Whileseparate central neuroendocrine circuits have been suggested


for different stressors (e.g., Pacak and Palkovits, [71]), themost important distinctions in physiological stress responsesare considered to occur between physical and psychologicalstressors [72–74], supporting the external validity of ouracute stress manipulation and its relevance for studyingphysiological adaptations associated with burnout.

Future research may focus on gender differences inthe longitudinal course of sympathetic and parasympatheticadaptation andHPA axis adaptation associated with burnout.Furthermore, longitudinal studies may also focus on explain-ing inconsistent findings with regard to cardiovascular andneuroendocrine changes associated with burnout. Factorslike complaints severity and duration and presence of absenceof the stressor seem to be relevant in this respect [18,21, 37, 72]. Third, future research may aim to clarify theadverse physiological profiles associated with specific aspectsof burnout (see, e.g., Marchand et al. [75]) or general distresscomplaints, such as depressive complaints, that often cooccurwith burnout.

In summary, this study indicates that a clinical level ofburnout is associated with adverse physiological changes thatmost likely increase the risk for CVD. More specifically, wefound reduced parasympathetic activity and a tendency forelevated cardiac output, which points towards predominanceof sympathetic activity in the sympathetic vagal balance.As basal systolic blood pressure was also elevated in males,further support was found for a sympathetic predominancespecifically in males. In addition, in males, evidence forhyporeactivity of the HPA axis was found. Consequently,the present study provides support for gender-specific car-diovascular and neuroendocrine profiles associated withburnout. Further longitudinal research is needed in orderto assess gender differences in the developmental trajectoryof sympathetic and parasympathetic changes and HPA axishypoactivity associated with burnout and the associationswith cardiovascular diseases.



Acknowledgments

This study was funded by the Netherlands Organization forScientific Research (NWO; Grant no. 580-02.403) and TheNetherlands Organization for Health Research and Develop-ment (ZON; Grant no. 2200.0005).This study could not havebeen realised without the contributions of the occupationalhealth services AGW (Hoorn, Netherlands) and AMD-UvA(Amsterdam, Netherlands) and of various general practition-ers in and around Amsterdam who informed their patientsabout our study. B. Cupido, E. Driessen, L. van der Ham,N. Heerooms, B. Janssen, M. Kwakman, and M. Rechesare gratefully acknowledged for their aid during the datacollection phase. The authors wish to thank P. K. Beekhof, R.van Loenen, A. Verlaan, I. Zutt, and DSLabs for biochemicalanalyses and Dr. J. Houtveen for advice.

References

[1] K. M. Lindblom, S. J. Linton, C. Fedeli, and I.-L. Bryngelsson,“Burnout in the working population: relations to psychosocialwork factors,” International Journal of Behavioral Medicine, vol.13, no. 1, pp. 51–59, 2006.

[2] C. Maslach, W. B. Schaufeli, and M. P. Leiter, “Job burnout,”Annual Review of Psychology, vol. 52, pp. 397–422, 2001.

[3] W. B. Schaufeli and B. P. Buunk, “Burnout: an overview of 25years of research and theorizing,” inTheHandbook of Work andHealth Psychology, M. J. Schabracq, J. A.M.Winnubst, andC. C.Cooper, Eds., pp. 383–429,Wiley,West Sussex, UK, 2nd edition,2003.

[4] S. Melamed, A. Shirom, S. Toker, S. Berliner, and I. Shapira,“Burnout and risk of cardiovascular disease: evidence, possiblecausal paths, and promising research directions,” PsychologicalBulletin, vol. 132, no. 3, pp. 327–353, 2006.

[5] K. L. Belkic, P. A. Landsbergis, P. L. Schnall, and D. Baker,“Is job strain a major source of cardiovascular disease risk?”Scandinavian Journal of Work, Environment and Health, vol. 30,no. 2, pp. 85–128, 2004.

[6] T. Chandola, A. Britton, E. Brunner et al., “Work stress andcoronary heart disease: what are the mechanisms?” EuropeanHeart Journal, vol. 29, no. 5, pp. 640–648, 2008.

[7] G. P. Chrousos and P. W. Gold, “The concepts of stress andstress system disorders: overview of physical and behavioralhomeostasis,” Journal of the American Medical Association, vol.267, no. 9, pp. 1244–1252, 1992.

[8] G. A. Kaltsas and G. O. Chrousos, “The neuroendocrinology ofstress,” in Handbook of Psychophysiology, J. T. Cacioppo, L. G.Tassinary, and G. G. Berntson, Eds., pp. 303–318, CambridgeUniversity Press, New York, NY, USA, 3rd edition, 2007.

[9] G. Grassi, F. Arenare, F. Quarti-Trevano, G. Seravalle, and G.Mancia, “Heart rate, sympathetic cardiovascular influences, andthe metabolic syndrome,” Progress in Cardiovascular Diseases,vol. 52, no. 1, pp. 31–37, 2009.

[10] S. B. Manuck, “Cardiovascular reactivity in cardiovasculardisease: ‘Once more unto the breach’,” International Journal ofBehavioral Medicine, vol. 1, no. 1, pp. 4–31, 1994.

[11] S. B. Manuck, A. L. Kasprowicz, and M. F. Muldoon,“Behaviorally-evoked cardiovascular reactivity and hyperten-sion: conceptual issues and potential associations,” Annals ofBehavioral Medicine, vol. 12, no. 1, pp. 17–29, 1990.

[12] K. J. Osterziel and R. Dietz, “Improvement of vagal tone byACE inhibition: a mechanism of cardioprotection in patientswith mild-to-moderate heart failure,” Journal of CardiovascularPharmacology, vol. 27, no. 2, pp. S25–S30, 1996.

[13] P. Palatini and S. Julius, “Elevated heart rate: a major riskfactor for cardiovascular disease,” Clinical and ExperimentalHypertension, vol. 26, no. 7-8, pp. 637–644, 2004.

[14] A. R. Schwartz, W. Gerin, K. W. Davidson et al., “Towarda causal model of cardiovascular responses to stress andthe development of cardiovascular disease,” PsychosomaticMedicine, vol. 65, no. 1, pp. 22–35, 2003.

[15] J. M. Dekker, R. S. Crow, A. R. Folsom et al., “Low heart ratevariability in a 2-minute rhythm strip predicts risk of coronaryheart disease and mortality from several causes: the ARICstudy,” Circulation, vol. 102, no. 11, pp. 1239–1244, 2000.

[16] D. Gasperin, G. Netuveli, J. S. Dias-da-Costa, andM. P. Pattussi,“Effect of psychological stress on blood pressure increase: ameta-analysis of cohort studies,”Cadernos de Saude Publica, vol.25, no. 4, pp. 715–726, 2009.


[17] F. A. Treiber, T. Kamarck, N. Schneiderman, D. Sheffield, G.Kapuku, and T. Taylor, “Cardiovascular reactivity and develop-ment of preclinical and clinical disease states,” PsychosomaticMedicine, vol. 65, no. 1, pp. 46–62, 2003.

[18] M. B. Danhof-Pont, T. van Veen, and F. G. Zitman, “Biomarkersin burnout: a systematic review,” Journal of PsychosomaticResearch, vol. 70, no. 6, pp. 505–524, 2011.

[19] S. Bellingrath, T. Weigl, and B. M. Kudielka, “Chronic workstress and exhaustion is associated with higher allostastic loadin female school teachers,” Stress, vol. 12, no. 1, pp. 37–48, 2009.

[20] B. B. Gump and K. A. Matthews, “Do background stressorsinfluence reactivity to and recovery from acute stressors?”Journal of Applied Social Psychology, vol. 29, no. 3, pp. 469–494,1999.

[21] B.M. Kudielka, S. Bellingrath, andD. H. Hellhammer, “Cortisolin burnout and vital exhaustion: an overview,”Giornale Italianodi Medicina del Lavoro ed Ergonomia, vol. 28, supplement 1, no.1, pp. 34–42, 2006.

[22] R.W.May,M.A. Sanchez-Gonzalez, and F.D. Fincham, “Schoolburnout: increased sympathetic vasomotor tone and attenuatedambulatory diurnal blood pressure variability in young adultwomen,” Stress, vol. 18, no. 1, pp. 11–19, 2015.

[23] R. W. May, M. A. Sanchez-Gonzalez, P. C. Brown, A. P. Kout-nik, and F. D. Fincham, “School burnout and cardiovascularfunctioning in young adultmales: a hemodynamic perspective,”Stress, vol. 17, no. 1, pp. 79–87, 2014.

[24] P. M. C. Mommersteeg, C. J. Heijnen, M. J. P. M. Verbraak,and L. J. P. van Doornen, “Clinical burnout is not reflected inthe cortisol awakening response, the day-curve or the responseto a low-dose dexamethasone suppression test,” Psychoneuroen-docrinology, vol. 31, no. 2, pp. 216–225, 2006.

[25] M. Sonnenschein, P. M. C. Mommersteeg, J. H. Houtveen, M. J.Sorbi,W. B. Schaufeli, and L. J. P. vanDoornen, “Exhaustion andendocrine functioning in clinical burnout: an in-depth studyusing the experience sampling method,” Biological Psychology,vol. 75, no. 2, pp. 176–184, 2007.

[26] L. J. P. van Doornen, J. H. Houtveen, S. Langelaan, A. B. Bakker,W. van Rhenen, and W. B. Schaufeli, “Burnout versus workengagement in their effects on 24-hour ambulatory monitoredcardiac autonomic function,” Stress and Health, vol. 25, no. 4,pp. 323–331, 2009.

[27] E. Kajantie and D. I. W. Phillips, “The effects of sex andhormonal status on the physiological response to acute psy-chosocial stress,” Psychoneuroendocrinology, vol. 31, no. 2, pp.151–178, 2006.

[28] B. M. Kudielka and C. Kirschbaum, “Sex-differences in HPAaxis responses to stress: a review,” Biological Psychology, vol. 69,no. 1, pp. 113–132, 2005.

[29] N. Rohleder, J. M. Wolf, and C. Kirschbaum, “Glucocorticoidsensitivity in humans-interindividual differences and acutestress effects,” Stress, vol. 6, no. 3, pp. 207–222, 2003.

[30] W. de Vente, M. Olff, J. G. C. van Amsterdam, J. H. Kamphuis,and P. M. G. Emmelkamp, “Physiological differences betweenburnout patients and healthy controls: blood pressure, heartrate, and cortisol responses,” Occupational and EnvironmentalMedicine, vol. 60, no. 1, pp. i54–i61, 2003.

[31] P. Palatini and S. Julius, “The role of cardiac autonomic functionin hypertension and cardiovascular disease,” Current Hyperten-sion Reports, vol. 11, no. 3, pp. 199–205, 2009.

[32] J. Penttila, A. Helminen, T. Jartti et al., “Time domain, geomet-rical and frequency domain analysis of cardiac vagal outflow:

effects of various respiratory patterns,” Clinical Physiology, vol.21, no. 3, pp. 365–376, 2001.

[33] Task Force of the European Society of Cardiology and theNorth American Society of Pacing and Electrophysiology,“Heart rate variability: standards ofmeasurement, physiologicalinterpretation and clinical use,” Circulation, vol. 93, no. 5, pp.1043–1065, 1996.

[34] U.M.Nater, R. laMarca, L. Florin et al., “Stress-induced changesin human salivary alpha-amylase activity: associations withadrenergic activity,” Psychoneuroendocrinology, vol. 31, no. 1, pp.49–58, 2006.

[35] American Psychiatric Association, Diagnostic and StatisticalManual of Mental Disorders, (DSM-IV), American PsychiatricAssociation, Washington, DC, USA, 4th edition, 1994.

[36] World Health Organization, International Classification of Dis-eases, (ICD-10), World Health Organization, Geneva, Switzer-land, 10th edition, 1992.

[37] B. G. Oosterholt, J. H. Maes, D. van der Linden, M. J. Verbraak,andM. A. Kompier, “Burnout and cortisol: evidence for a lowercortisol awakening response in both clinical and non-clinicalburnout,” Journal of Psychosomatic Research, vol. 78, no. 5, pp.445–451, 2015.

[38] W. de Vente, J. H. Kamphuis, P. M. G. Emmelkamp, and R. W.B. Blonk, “Individual and group cognitive-behavioral treatmentfor work-related stress complaints and sickness absence: arandomized controlled trial,” Journal of Occupational HealthPsychology, vol. 13, no. 3, pp. 214–231, 2008.

[39] A. T. Beck and R. A. Steer, Beck Depression Inventory, Manual,Harcourt Brace, New York, NY, USA, 1987.

[40] WorldHealthOrganization,Composite International DiagnosticInterviewManual, (CIDI), Version 2.1, World Health Organiza-tion, Geneva, Switzerland, 1997.

[41] E. Childs and H. de Wit, “Hormonal, cardiovascular, andsubjective responses to acute stress in smokers,” Psychopharma-cology, vol. 203, no. 1, pp. 1–12, 2009.

[42] N. A. Nicolson and R. van Diest, “Salivary cortisol patterns invital exhaustion,” Journal of Psychosomatic Research, vol. 49, no.5, pp. 335–342, 2000.

[43] M. Olff, L. J. M. Mulder, T. H. The, L. de Leij, and P.M. G. Emmelkamp, “The psychobiological stress response indepressed patients compared to healthy controls,” inProceedingsof the 29th Annual Congress of the EABCT, p. 345, Dresden,Germany, September 1999.

[44] M. P. Roy, C. Kirschbaum, and A. Steptoe, “Psychological,cardiovascular, and metabolic correlates of individual differ-ences in cortisol stress recovery in young men,” Psychoneuroen-docrinology, vol. 26, no. 4, pp. 375–391, 2001.

[45] B. Molenkamp, VSRRP98 Manual, Version 5.4, University ofAmsterdam, Amsterdam, The Netherlands, 2005.

[46] Finapres Medical Systems BV, Beatscope 1.1; Users’s Guide,Finapres Medical Systems BV, Arnhem,The Netherlands, 2002.

[47] G. Lu, F. Yang, J. A. Taylor, and J. F. Stein, “A comparison ofphotoplethysmography andECG recording to analyse heart ratevariability in healthy subjects,” Journal of Medical Engineeringand Technology, vol. 33, no. 8, pp. 634–641, 2009.

[48] M. Navazesh, “Methods for collecting saliva,” Annals of the NewYork Academy of Sciences, vol. 694, pp. 72–77, 1993.

[49] W. Schaufeli andD. vanDierendonck,UBOS, Utrechtse BurnoutSchaal Manual, Swets & Zeitlinger B.V., Lisse, The Netherlands,2000.


[50] A. J. H. M. Beurskens, U. Bultmann, I. Kant, J. H. M. M.Vercoulen, G. Bleijenberg, and G. M. H. Swaen, “Fatigueamong working people: validity of a questionnaire measure,”Occupational and Environmental Medicine, vol. 57, no. 5, pp.353–357, 2000.

[51] E. de Beurs, R. van Dyck, L. A. Marquenie, A. Lange, and R.W. B. Blonk, “The DASS: a questionnaire for the measurementof depression, anxiety, and stress,” Gedragstherapie, vol. 34, pp.35–53, 2001.

[52] M. D. McNair, M. Lorr, and L. F. Droppleman, Manual forthe Profile of Mood States, Educational and Industrial TestingService, San Diego, Calif, USA, 1971.

[53] K. A. Matthews, B. B. Gump, and J. F. Owens, “Chronicstress influences cardiovascular and neuroendocrine responsesduring acute stress and recovery, especially in men,” HealthPsychology, vol. 20, no. 6, pp. 403–410, 2001.

[54] A. van Stegeren, N. Rohleder, W. Everaerd, and O. T. Wolf,“Salivary alpha amylase asmarker for adrenergic activity duringstress: effect of betablockade,” Psychoneuroendocrinology, vol.31, no. 1, pp. 137–141, 2006.

[55] A. Steptoe, J. Siegrist, C. Kirschbaum, and M. Marmot, “Effort-reward imbalance, overcommitment, and measures of cortisoland blood pressure over the working day,” PsychosomaticMedicine, vol. 66, no. 3, pp. 323–329, 2004.

[56] U. Bultmann, I. Kant, S. V. Kasl, A. J. H. M. Beurskens, and P.A. van den Brandt, “Fatigue and psychological distress in theworking population psychometrics, prevalence, and correlates,”Journal of Psychosomatic Research, vol. 52, no. 6, pp. 445–452,2002.

[57] C. Kirschbaum and D. H. Hellhammer, “Salivary cortisol inpsychobiological research: an overview,” Neuropsychobiology,vol. 22, no. 3, pp. 150–169, 1989.

[58] J. A. Bosch, H. S. Brand, T. J. M. Ligtenberg, B. Bermond,J. Hoogstraten, and A. V. Nieuw Amerongen, “Psychologicalstress as a determinant of protein levels and salivary- inducedaggregation of Streptococcus gordonii in human whole saliva,”Psychosomatic Medicine, vol. 58, no. 4, pp. 374–382, 1996.

[59] N. Rohleder, J. M. Wolf, E. F. Maldonado, and C. Kirschbaum,“The psychosocial stress-induced increase in salivary alpha-amylase is independent of saliva flow rate,” Psychophysiology,vol. 43, no. 6, pp. 645–652, 2006.

[60] C. K. Fortunato, A. E. Dribin, D. A. Granger, and K. A. Buss,“Salivary alpha-amylase and cortisol in toddlers: differentialrelations to affective behavior,” Developmental Psychobiology,vol. 50, no. 8, pp. 807–818, 2008.

[61] U. M. Nater, N. Rohleder, J. Gaab et al., “Human salivaryalpha-amylase reactivity in a psychosocial stress paradigm,”International Journal of Psychophysiology, vol. 55, no. 3, pp. 333–342, 2005.

[62] N. Ramasubbu, V. Paloth, Y. Luo, G. D. Brayer, and M.J. Levine, “Structure of human salivary 𝛼-amylase at 1.6 Aresolution: implications for its role in the oral cavity,” ActaCrystallographica, Section D: Biological Crystallography, vol. 52,no. 3, pp. 435–446, 1996.

[63] J. M. Wolf, E. Nicholls, and E. Chen, “Chronic stress, salivarycortisol, and 𝛼-amylase in children with asthma and healthychildren,” Biological Psychology, vol. 78, no. 1, pp. 20–28, 2008.

[64] American Heart Association, 2004 Heart and Stroke StatisticalUpdate, American Heart Association, Dallas, Tex, USA, 2004.

[65] S. Yusuf, S. Reddy, S. Ounpuu, and S. Anand, “Global burdenof cardiovascular diseases part II: variations in cardiovascular

disease by specific ethnic groups and geographic regions andprevention strategies,” Circulation, vol. 104, no. 23, pp. 2855–2864, 2001.

[66] W. de Vente, J. H. Kamphuis, R. W. Blonk, and P. M.Emmelkamp, “Recovery of work-related stress: complaintreduction and work-resumption are relatively independentprocesses,” Journal of Occupational Rehabilitation, 2015.

[67] M. P. Roy,A. Steptoe, andC.Kirschbaum, “Life events and socialsupport as moderators of individual differences in cardiovas-cular and cortisol reactivity,” Journal of Personality and SocialPsychology, vol. 75, no. 5, pp. 1273–1281, 1998.

[68] A. Steptoe andM.Kivimaki, “Stress and cardiovascular disease,”Nature Reviews Cardiology, vol. 9, no. 6, pp. 360–370, 2012.

[69] P. M. C. Mommersteeg, C. J. Heijnen, G. P. J. Keijsers, M. J.P. M. Verbraak, and L. J. P. Van Doornen, “Cortisol deviationsin people with burnout before and after psychotherapy: a pilotstudy,” Health Psychology, vol. 25, no. 2, pp. 243–248, 2006.

[70] M. Olff, G.-J. de Vries, Y. Guzelcan, J. Assies, and B. P. R.Gersons, “Changes in cortisol and DHEA plasma levels afterpsychotherapy for PTSD,” Psychoneuroendocrinology, vol. 32,no. 6, pp. 619–626, 2007.

[71] K. Pacak and M. Palkovits, “Stressor specificity of centralneuroendocrine responses: implications for stress-related dis-orders,” Endocrine Reviews, vol. 22, no. 4, pp. 502–548, 2001.

[72] M. M. Ostrander, Y. M. Ulrich-Lai, D. C. Choi, N. M. Richtand,and J. P. Herman, “Hypoactivity of the hypothalamo-pituitary-adrenocortical axis during recovery from chronic variablestress,” Endocrinology, vol. 147, no. 4, pp. 2008–2017, 2006.

[73] R. M. Sapolsky, L. M. Romero, and A. U. Munck, “How do glu-cocorticoids influence stress responses? Integrating permissive,suppressive, stimulatory, and preparative actions,” EndocrineReviews, vol. 21, no. 1, pp. 55–89, 2000.

[74] A. Singh, J. S. Petrides, P. W. Gold, G. P. Chrousos, and P.A. Deuster, “Differential hypothalamic-pituitary-adrenal axisreactivity to psychological and physical stress,” Journal ofClinical Endocrinology andMetabolism, vol. 84, no. 6, pp. 1944–1948, 1999.

[75] A. Marchand, R.-P. Juster, P. Durand, and S. J. Lupien, “Burnoutsymptom sub-types and cortisol profiles: what’s burningmost?”Psychoneuroendocrinology, vol. 40, no. 1, pp. 27–36, 2014.

Research ArticleJob Strain and Self-Reported Insomnia Symptomsamong Nurses: What about the Influence of EmotionalDemands and Social Support?

Luciana Fernandes Portela,1 Caroline Kröning Luna,2 Lúcia Rotenberg,2

Aline Silva-Costa,2 Susanna Toivanen,3 Tania Araújo,4 and Rosane Härter Griep2

1National School of Public Health (ENSP/Fiocruz), Avenida Brasil 4365, 21040-360 Rio de Janeiro, RJ, Brazil2Health and Environmental Education Laboratory, Oswaldo Cruz Institute (IOC/Fiocruz), Avenida Brasil 4365,21040360 Rio de Janeiro, RJ, Brazil3Centre for Health Equity Studies (CHESS), Stockholm University and Karolinska Institute, Sveaplan, Sveavagen 160,Floor 5, 106-91 Stockholm, Sweden4Department of Health, State University of Feira de Santana, R. Claudio Manoel da Costa 74/1401, Canela,40110-180 Salvador, BA, Brazil

Correspondence should be addressed to Luciana Fernandes Portela; [email protected]

Received 16 January 2015; Revised 8 April 2015; Accepted 8 May 2015


Copyright © 2015 Luciana Fernandes Portela et al. This is an open access article distributed under the Creative CommonsAttribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work isproperly cited.

Job strain, derived from high psychological demands and low job control, is associated with insomnia, but information on therole of emotional demands and social support in this relationship is scarce. The aims of this study were (i) to test the associationbetween job strain and self-reported insomnia symptoms, (ii) to evaluate the combination of emotional demands and job controlregarding insomnia symptoms, and (iii) to analyze the influence of social support in these relationships. This cross-sectional studyrefers to a sample of nurses (N = 3,013 andN = 3,035 for Job Strain and Emotional demand-control model, resp.) working at publichospitals in Rio de Janeiro, Brazil. Data were collected through a self-report questionnaire. The prevalence of insomnia symptomswas 34.3%. Job strain was associated with increased odds for insomnia symptoms (OR: 2.20); the same result was observed with thecombination of emotional demands and low job control (OR: 1.99). In both models, the inclusion of low social support combinedwith high demands and low job control led to increased odds for insomnia symptoms, compared to groups with high social supportfrom coworkers and supervisors. Besides job strain, the study of emotional demands and social support are promising with regardsto insomnia symptoms, particularly among nurses.

1. Introduction

Insomnia is known as a relevant public health problem, witha complex etiology [1, 2]. Sleep complaints can have negativeeffects on the immune system and metabolism [3] as wellas various health issues such as depression [4], hyperten-sion [5], and coronary heart diseases [6]. Employees’ sleepdisturbances can have significant effects on organizations’performances due to impairments in concentration, commu-nication skills, decision-making, and flexible thinking. Sleep

disturbances may also lead to reduced job motivation andpoor leadership qualities [7].

Previous researches identified several risk factors forinsomnia. On one hand, there were individual factors such asfemale gender, increasing age, low body mass index (BMI),low socioeconomic status, marital status, and presence ofphysical or mental illness [8, 9]. On the other hand, personalbehaviours, such as frequent use of alcohol [10] and dietarypatterns [11], were also documented in the literature. Inaddition, work-related factors including psychosocial effects


http://dx.doi.org/10.1155/2015/820610


of the work environment (such as job satisfaction, overcom-mitment, effort-reward-imbalance, hectic work, physicallystrenuous work, shift work, job insecurity, organizationalinjustice, low employment opportunities, and job strain) canalso be counted as risk factors for insomnia [7, 12–17].

The job strainmodel by Karasek assesses psychosocial jobstress by taking psychological demands and the job controlinto account. Psychological demands arise particularly fromtime pressure, amount of work, and conflicting work instruc-tions and job control or decision latitude describes the level ofworkers’ control over the performance of their jobs. Job strainresults from the interaction of high psychological demandsand low job control [18, 19].

In a further approach to psychosocial stress at workby Karasek, emotional demands [20] and social supportfrom supervisors and coworkers [18] were evaluated in thejob strain model. Emotional demands have an impact onfeelings or emotions and are strongly related to interpersonalrelationships, including caring and concern for others [20].Social support by supervisors and colleagues helps workersto appreciate their own value and competencies and enablesthem to copewith upcoming demands and difficult situations[19]. Social support, as defined by Karasek andTheorell, is “anoverall level of helpful social interaction available on the jobfrom both coworkers and supervisors” [19].

More recently, the importance of including emotionaldemands in the job strain model has been emphasizedby several authors, particularly considering jobs involvinginteraction with clients [21–23]. In the context of thesediscussions, van Vegchel et al. [24] argue that the demand-control-support model would give an oversimplified imagefor human service work, proposing that emotional demandsare an essential complement to psychological demands, asworkload. Actually, emotional demands have been associatedwith physical symptoms, as low back pain [25], psychologicaloutcomes [26], and occupational injuries [23].

A number of investigations have reported increase ofsleep disturbances or insomnia due to high demands, lowjob control, and job strain [12, 13, 16, 27]. In addition,some authors have investigated the role of social support forinsomnia [14, 15, 28, 29] or the buffering effect of job controland social support against insomnia [7, 15, 17, 30, 31]. Somestudies have associated social support but not job strain withinsomnia. In contrast, some studies did not show a bufferingeffect of social support and high job control on better sleepquality [5, 30].

In the present study the relationship between job strainand self-reported insomnia symptoms among nurses is dealtwith through two aspects that deserve further investigation:Firstly, the partly conflicting study results regarding the roleof social support by itself and combined with the job strainmodel. Secondly, a detailed study of emotional demandssince this construct is relatively recent considering the jobstrain model. To our knowledge, there have been no studiesfocusing on the combination of emotional demands andjob control in relation to insomnia, although the relationsbetween sleep and other approaches of emotional demandshave already been addressed [32–34]. Job stress among nursesis approached in the present study by using two models:

(i) the job strainmodel, derived from psychological demandsand job control and (ii) the combination of job control withemotional demands.This study integrated social support intothe models under the assumption that the prevalence ofself-reported insomnia symptoms is higher among nurseswith high psychological and emotional job demands andwith low decision latitude and low social support, comparedwith nurses without these stressors. Three objectives guidedthe study: (i) to investigate the association between the jobstrain model and self-reported insomnia symptoms, (ii) toevaluate the combination of emotional demands and jobcontrol regarding the associationwith self-reported insomniasymptoms, and (iii) to analyze the influence of social supporton these associations.

2. Methods

2.1. Subjects. Nurses from the 18 largest public hospitals(with over 150 beds) in the city of Rio de Janeiro, Brazil,participated in this cross-sectional study. Each nurse wascontacted personally by a team of interviewers (in mostcases, nurses themselves), who explained the objectives of thestudy and invited them to participate. After signing the con-sent form, the nurses received a comprehensive self-reportquestionnaire to complete. The questionnaire was dividedinto three blocks, corresponding to (i) variables related towork (number of jobs, weekly work hours, work schedule,intention to leave the profession, recovery after work scale,effort-reward imbalance scale, psychological and emotionaldemands, job control, and social support from workers andfrom colleagues), (ii) behaviours related to health and lifestyle(self-reported diagnosis of hypertension and cardiovasculardiseases, sickness absenteeism, sleep duration, self-reportedinsomnia symptoms, sleep satisfaction, weight, height, prac-tice of physical activity, smoking, consumption pattern ofalcoholic beverages, coffee consumption, common mentaldisorder), and (iii) sociodemographic data and informationon the domestic sphere (school degree, income, maritalstatus, number and age of children, type of domestic duties,and number of hours of domestic work). Data collectionwas carried out from March 2010 to November 2011. Trainedreviewers entered, reviewed, and coded the informationcontained in the returned questionnaire into a database usingthe software Microsoft Access 2010.

2.2. Study Variables

2.2.1. Self-Reported Insomnia Symptoms. Individuals thatreported insomnia symptoms were defined as those whoanswered often or always to any of the questions concerning“Difficulty to sleep or get to sleep,” “Waking up during thenight (more than three times),” or “Waking up too early inthemorning and having trouble getting back to sleep” [7].Theresponse categories for the questions were “Never,” “Rarely (afew times per year),” “Sometimes (several times per month),”“Often (several days per week),” and “Always (every day).”

2.2.2. Job Stress. Perceived stress at work was measured bythe Portuguese version of the Job Content Questionnaire 2.0


(JCQ 2.0) [35]. This scale measures psychological demands,job control, and social support from supervisors and cowork-ers. It also includes questions on emotional demands, whichwere evaluated by a three-item scale regarding being con-fronted with emotionally demanding work, having to hideone’s emotions during work or demanding a lot of negotia-tion, discussion, or understanding with others.

The response categories were a Likert-scale rangingfrom 1 (strongly disagree) to 4 (strongly agree) and werereverse-coded afterwards if necessary.The following were theCronbach’s alpha coefficients: psychological demands (0.71),emotional demands (0.59), job control (0.75), supervisorsupport (0.74), and coworker support (0.81).

Two stress models were analyzed: the job strain, based onpsychological demands and job control, and the emotionaldemand-control model. The first step in the analysis wasthe description of each individual dimension, which wasbased on the categorization into approximately equal-sizedterciles. Second, to compute the job strain both psychologicaldemands and job control were dichotomized at the medianvalue into high and low groups, leading to four job categories:low-strain, passive job, active job, and high-strain job [19]. Asimilar procedure was conducted for the emotional demand-control model. In a further step, the social support categories(low and high) were added to the low-strain and the high-strain quadrants of the job strain model, leading to fourcombinations: low-strain with high social support (referencegroup), low-strain with low social support, high-strain withhigh social support, and high-strain with low social support.An analogous procedure was performed for the emotionaldemand-control model, also resulting in four combinations.In this case, the reference group comprised workers with lowemotional demand-high job control plus high social support.

2.2.3. Covariates. Relevant covariates associated with bothself-reported insomnia symptoms and job strain wereconsidered in the present study as possible confounders[2, 15–17]. Sociodemographic covariates included age, sex,per capita income, marital status (married/cohabiting,divorced/separated, or never married/cohabiting), andchildren less than six years old. Work-related variablesincluded night work (current night-worker, former night-worker, or never worked nights), as well as working hoursper week (<40, 40–59, and ≥60). Health-related factorswere the risk of alcohol consumption (frequency and doseof alcohol per week based on the recommendation of theNational Institute on Alcohol Abuse and Alcoholism) [36],smoking habits (smoker, ex-smoker, and nonsmoker), andcoffee intake (numbers of cups of coffee per day). Body massindex (BMI) was defined by self-reported weight (kg)/height(m2). BMI was categorized as underweight (<18.49), normalweight (18.50–24.99), overweight (25.00–29.99), and obese(≥30.00).

2.3. Statistical Analysis. Descriptive analyses of sociodemo-graphic variables as well as variables related to work andhealth were based on chi-square tests. Multivariate logisticregression analyses, with 95% confidence intervals (95% CI),

were used to examine the associations between eachmodel ofjob stress and self-reported insomnia symptoms. Both crudeand adjusted odds ratios were presented (significance at 𝑝 <0.05). All data were analyzed using the software IBM Statisti-cal Package for the Social Sciences version 19.0 (IBM SPSS).

3. Results

The number of completed questionnaires returned was 3,229(82.7% of the total number of nurses) and these were part ofthe analysis for this present paper. The 675 (17.3%) who didnot participate were due to refusals (𝑛 = 478), nurses beingunavailable across multiple visits over two months (𝑛 = 128),and absences because of holidays (𝑛 = 69). Consideringthat the analyses are based on scales, data from participantswho did not answer a single question were excluded fromdatabank, resulting in a final sample size of 3013 workersfor the analysis of job strain, and 3035 for the emotionaldemand-control model. We performed a sensitivity analysisin which we assumed that persons with missing informationfor job strain did not have an exposure in that category (i.e.,they were classified in the low strain group). This procedurewas adopted to test the influence of missing data on theresults [37]. Results of the multivariate logistic regressionanalysis showed that the direction and the strength of theassociations between job strain and self-reported insomniasymptoms were nearly identical as the previous ones. Thesame result was observed as regards the emotional demand-control model (data not shown). Thus, the loss of theseparticipants did not alter significantly any of the above-reported results.

The participants were predominantly female (87.3%) withmean age of 39.9 years (±10.0; range: 22 to 68 years).Overall the prevalence of self-reported insomnia symptomswas 34.3%. The highest prevalence of self-reported insomniasymptoms was observed for the oldest groups, women,married or divorced, and participants with children undersix years old. Persons working 40 hours per week or less andformer night workers showed higher self-reported insomniasymptoms prevalence compared with their counterparts.Participants with a very low or high BMI and who werenot physically active had a higher prevalence of self-reportedinsomnia symptoms. In addition, the prevalence of self-reported insomnia symptoms was highest among high-riskdrinkers, current smokers, and persons with a coffee intakeof more than three cups per day. Neither race nor educationlevel was associated with self-reported insomnia symptoms(Table 1).

Table 2 shows that odds ratio corresponding to all stressdimensions presented a gradient. Thus, compared with lowdemands, adjusted OR for medium and high psychologicaldemands were 1.33 and 1.82, respectively. The correspondingvalues for emotional demands were 1.19 and 1.59, respectively.A gradient from higher to lower degree of job controland social support were also observed. Low job controlcorresponded to OR = 1.40 (CI 95% 1.14–1.72) and low socialsupport corresponded to OR = 1.59 (CI 95% 1.29–1.95).

According to Table 3, participants with high demandsand low control showed a higher prevalence of self-reported


Table 1: Description of the whole group; bivariate analysis between sample characteristics and self-reported insomnia symptoms, based onchi-square tests. Rio de Janeiro, Brazil, 2011.

Sample characteristics Whole group Without insomnia symptoms With insomnia symptoms𝑝

𝑛 % 𝑛 % 𝑛 %Age 0.003≤34 years 1153 35.7 799 38.1 354 30.935–54 years 1755 54.4 1131 53.9 624 35.9≥55 years 277 8.6 167 8.0 110 40.1

Sex 0.013Male 411 12.7 294 13.8 117 28.8Female 2818 87.3 1839 86.2 979 35.1

Race 0.730Black 347 10.7 222 10.6 125 36.2White 1775 55.0 1181 56.2 594 33.7Mixed 956 29.6 624 29.7 332 35.2Others 113 3.5 73 3.5 40 35.7

Per capita income per month 0.070≤US$1,740 877 27.2 559 27.8 318 36.5US$1,741–US$3,050 1194 37.0 783 38.9 411 34.9≥US$3,051 978 30.3 671 33.3 307 31.6

Education level 0.224University degree 795 24.6 537 25.6 258 33.0Postgraduation 2165 67.0 1403 66.8 762 35.4Master’s or PhD 231 7.2 160 7.6 71 30.9

Marital status 0.016Married/cohabiting 1833 56.8 1180 56.1 653 36.1Divorced/separated 593 18.4 387 18.4 206 35.5Never married/cohabiting 764 23.7 535 25.5 229 30.1

Children <6 years <0.001Yes 587 18.2 349 16.7 238 40.7No 2583 80.0 1740 83.3 843 33.0

Working hours 0.032<40 hours 753 23.3 472 21.5 281 37.540–59 hours 1565 48.5 1277 58.1 288 31.6≥60 hours 789 24.4 449 20.4 340 33.0

Night work 0.085Current night work 1979 61.3 1329 62.3 650 33.2Former night work 1036 32.1 657 30.8 379 37.0Never night work 214 6.6 147 6.9 67 31.8

BMI 0.004Underweight 39 1.2 21 1.0 18 46.2Normal weight 1426 44.2 982 47.9 444 31.2Overweight 1001 31.0 654 31.9 347 35.0Obese 633 19.6 391 19.1 242 38.4

Physical activity <0.001Yes 1017 31.5 726 34.3 291 28.8No 2182 67.6 1389 65.7 793 36.7

Risk of alcohol consumption 0.009No risk 1083 33.5 679 40.1 404 34.1Low risk 1563 48.4 961 56.8 602 33.3High risk 119 3.7 53 3.1 66 45.8

Smoking habits <0.001Nonsmoker 2424 75.1 1637 77.4 787 32.8Ex-smoker 501 15.5 319 15.1 182 36.5Smoker 281 8.7 160 7.6 121 43.5

Coffee intake <0.001No coffee 672 20.8 464 22.1 208 31.1≤3 cups per day 1452 45.0 988 47.0 464 32.2>3 cups per day 1057 32.7 648 30.9 409 39.2


Table 2: Crude and adjusted OR for the association between job strain dimensions and self-reported insomnia symptoms, based onmultivariate logistic regression tests. Rio de Janeiro, Brazil, 2011.

Job stress model dimensions With insomnia symptoms OR crude Multivariatemodel 1

Multivariatemodel 2

Multivariatemodel 3

𝑛 %Psychological demands

Low 274 29.4 1.0 1.0 1.0 1.0Medium 339 32.8 1.24 (1.01–1.51) 1.33 (1.08–1.65) 1.33 (1.08–1.65) 1.33 (1.07–1.64)High 451 40.0 1.68 (1.38–2.05) 1.90 (1.54–2.34) 1.89 (1.54–2.34) 1.82 (1.48–2.27)

Emotional demandsLow 242 28.3 1.0 1.0 1.0 1.0Medium 326 33.3 1.24 (0.99–1.53) 1.20 (0.97–2.07) 1.20 (0.97–1.50) 1.19 (0.96–1.49)High 512 39.5 1.60 (1.31–1.95) 1.69 (1.38–2.07) 1.69 (1.38–2.07) 1.59 (1.29–1.96)

Job controlHigh 334 32.4 1.0 1.0 1.0 1.0Medium 345 32.3 1.00 (0.82–1.21) 0.98 (0.80–1.19) 0.99 (0.80–1.20) 1.00 (0.82–1.22)Low 372 39.8 1.36 (1.11–1.65) 1.36 (1.11–1.66) 1.37 (1.12–1.68) 1.40 (1.14–1.72)

Social supportHigh 311 29.4 1.0 1.0 1.0 1.0Medium 397 34.1 1.17 (0.96–1.41) 1.14 (0.94–1.38) 1.14 (0.94–1.39) 1.18 (0.97–1.44)Low 365 41.1 1.58 (1.29–1.92) 1.58 (1.29–1.93) 1.58 (1.29–1.93) 1.59 (1.29–1.95)

Multivariate model 1: adjustments for sociodemographics: sex, age, income, marital status, and children under 6 years.Multivariate model 2: model 1 + adjustments for work-related factors: working hours and night work.Multivariate model 3: model 2 + adjustments for health-related factors: BMI, smoking habits, alcohol consumption, physical activity, and coffee intake.

insomnia symptoms than the respective reference groupsboth regarding job strain (OR = 2.20) and the emotionaldemand-control model (OR = 1.99). For both models, theinclusion of low social support resulted in an increasedprevalence of self-reported insomnia symptoms, comparedwith the combination with high social support, particularlyfor the emotional demand-control model, with an increasefrom OR = 1.41 to OR = 2.47.

4. Discussion

An original contribution of this study is the analysis of thecombination of high emotional demands and low job control,which was shown to be associated with self-reported insom-nia symptoms. Emotional demands have recently been addedto the 2.0 version of JCQ [35], but studies on this constructlinked to the demand-control model are yet scarce, and noonewas identified dealingwith self-reported insomnia symp-toms. The results seem to be plausible given that emotionallydemanding occupations are related to health outcomes andburnout [26], although the underlying mechanisms are stillunknown [22]. Actually, nursing is a profession with highemotional demands throughout the daily working routine[21, 38]. Results on the emotional demand-control modelfollow the ones obtained in previous research on the rela-tionship between job strain and self-reported evaluations ofsleep, such as insomnia symptoms [12] and sleep disturbances[27]. According to the results here presented, studies onjob stress and self-reported insomnia symptoms should alsoincorporate emotional demands, particularly among nurses.

Considering the scale dimensions, the odds for self-reported insomnia symptomswere slightly higher for personswith high psychological demands than for those with highemotional demands. Additionally, positions with a lack ofcontrol and social support revealed significant higher oddsfor self-reported insomnia symptoms as well. These resultsare convincing because they are similar to the outcomes ofprevious research suggesting higher risk for insomnia dueto high demands, low control, and social support separately[14, 29].

The present research integrated the social support dimen-sion into the job stress models. This approach allowedconfirming the assumption that social support makes a dif-ference, which seems to be higher for the emotional demand-control model than for the job strain model as judged bydifferences in odds between groups with high and low socialsupport. The results showed that the three dimensions com-bined (high demands, low control, and low social support)increased the chances of self-reported insomnia symptoms.Furthermore, these results showed that the combinationincreases the odds of self-reported insomnia symptomsmoresignificantly than each of the three dimensions separately.

Some limitations of this study should be mentioned. Thestudy design was a cross-sectional one, which always has thedisadvantage of measuring correlation rather than causality.Therefore, there is the possibility of reverse causality betweenjob strain, lack of support, and self-reported insomnia symp-toms. There is a real possibility that stress and low socialsupport do not lead to insomnia, but that preexisting self-reported insomnia symptoms can contribute to a bad mood,


Table3:Cr

udeandadjuste

dORforthe

associationbetweenDem

and-Con

trol-S

uppo

rtMod

els,andself-repo

rted

insomniasymptom

s,basedon

multiv

ariatelogisticregressio

ntests

.Rio

deJaneiro

,Brazil,2011.

Mod

elso

fjob

stress

With

insomnias

ymptom

sORcrud

eMultiv

ariatemod

el1

Multiv

ariatemod

el2

Multiv

ariatemod

el3

𝑛%

Jobstr

ain

Low-strain

231

27.4

1.01.0

1.01.0

Passive

206

32.3

1.32(1.04–

1.68)

1.32(1.03–1.6

8)1.3

3(1.04–

1.70)

1.33(1.04–

1.71)

Activ

e269

36.5

1.63(1.30–

2.05)

1.75(1.38–2.20)

1.75(1.39

–2.21)

1.69(1.33

–2.15

)High-str

ain

327

42.4

1.99(1.60–

2.49)

2.20

(1.75–2.77)

2.22

(1.76

–2.80)

2.20

(1.74

–2.78)

Jobstr

ain+socialsupp

ort

Low-strain+high

supp

ort

386

29.7

1.01.0

1.01.0

Low-strain+lowsupp

ort

314

35.0

1.19(0.82–1.17)

1.17(0.81–1.7

0)1.19(0.82–1.7

2)1.17(0.80–

1.71)

High-str

ain+high

supp

ort

9036.0

1.62(1.16

–2.26)

1.81(1.2

8–2.55)

1.83(1.29–

2.58)

1.78(1.26–

2.52)

High-strain

+lowsupp

ort

235

45.6

2.40

(1.84–

3.14)

2.60

(1.98–3.42)

2.64

(2.00–

3.48)

2.64

(1.99–

3.49)

Emotionald

emand-controlm

odel

Lowdemand+high

control

276

29.6

1.01.0

1.01.0

Lowdemand+lowcontrol

274

33.3

1.18(0.95–1.4

7)1.18(0.94–

1.47)

1.19(0.95–1.4

8)1.2

0(0.96–

1.51)

Highdemand+high

control

231

34.7

1.25(0.99–

1.56)

1.32(1.05–1.6

6)1.3

2(1.04–

1.66)

1.25(0.99–

1.58)

Highdemand+lowcontrol

266

44.9

1.89(1.50–

2.37)

2.08

(1.64–

2.62)

2.08

(1.65–2.63)

1.99(1.57–2.53)

Emotionald

emand-controlm

odel+socialsupp

ort

Lowdemand/high

control+

high

supp

ort

416

30.3

1.01.0

1.01.0

Lowdemand/lowcontrol+

lowsupp

ort

358

35.1

1.10(0.79–

1.5)

1.10(0.80–

1.54)

1.12(0.81–1.5

6)1.13(0.81–1.5

9)Highdemand/high

control+

high

supp

ort

6936.3

1.32(0.92–1.9

0)1.4

8(1.02–2.15)

1.49(1.02–2.16)

1.41(0.96–2.05)

Highdemand/lowcontrol+

lowsupp

ort

197

49.4

2.33

(1.77–3.07)

2.53

(1.91–3.35)

2.55

(1.93–3.39)

2.47

(1.86–

3.29)

Multiv

ariatemod

el1:adjustm

entsforsociodemograph

ics:sex,age,income,marita

lstatus,andchild

renun

der6

years.

Multiv

ariatemod

el2:mod

el1+

adjustm

entsforw

ork-relatedfactors:working

hoursa

ndnightw

ork.

Multiv

ariatemod

el3:mod

el2+adjustm

entsforh

ealth

-related

factors:BM

I,sm

okinghabits,

alcoho

lcon

sumption,

physicalactiv

ity,and

coffeeintake.


which can lead to difficulties in facing demands, but also totreating colleagues poorly, which may cause the lack of socialsupport. Moreover, some potential confounders were notincluded in the adjustment of the results, such as informationon depression, which may influence sleep disturbance [4].Another potential confounder is the use of sleeping pills,which has already been associated with work-related stressamong female nurses [39]. Furthermore, no data on socialsupport from family and friends, which can have a significantimpact on sleep quality [40], were collected.

Among the strengths of the current research is the use ofa large sample, which enabled reliable results. The inclusionof emotional demands as a psychosocial factor linked to thedemand-control model allowed the exploration of this rela-tively new concept, thus providing an original contributionto the knowledge of work-related sleep disturbances.

Besides approaching the “classic” job strain model ofpsychological demands and job control, this study tookemotional demands and social support into account. Bothaspects are vitally important for occupations like nursing,which require caring personal service [41]. This is one ofthe professions where high emotional demands are causedby dealing with patients, seeing their suffering and some-times their death, but also where one must rely on one’steam, coworkers, and supervisors in order to do the jobprofessionally. Therefore, it is important to include bothaspects in the association between job stress and self-reportedinsomnia symptoms. Although the precise directional rela-tionship between job stress, social support, and self-reportedinsomnia symptoms in this study is unclear, it is advisableto perform certain stress management and team buildingactivities to improve individual coping stress strategies, thecollectivework climate, and solidarity in theworkforce.Thesemeasures would benefit workers’ health, the hospitals them-selves through the quality of assistance, and the economy ingeneral.

Ethical Approval

The study was approved by the appropriate committees andofficials. Approval to conduct the research was granted bythe ethics committee from the hospitals. The study wasbriefly explained to participants and they were informedthat involvement was completely voluntary and that theycould withdraw at any time with no negative implications.Participants signed consent forms.



Acknowledgments

Rosane Harter Griep and Lucia Rotenberg are recipients ofresearch productivity grants from CNPq (Brazilian Councilfor Scientific and Technological Development). The authorswould like to acknowledge the Brazilian National Research

Council (CNPq) and Carlos Chagas Filho Foundation forResearch Support in the State of Rio de Janeiro (FAPERJ).Rosane Harter Griep and Lucia Rotenberg are fellows ofthe Irving Selikoff International Fellows of the Mount SinaiSchool of Medicine ITREOH Program. Their work wassupported in part by Grant 1 D43 TW00640 from the FogartyInternational Center of the National Institutes of Health.

References

[1] D. Cunnington, M. F. Junge, and A. T. Fernando, “Insomnia:prevalence, consequences and effective treatment,”TheMedicaljournal of Australia, vol. 199, no. 8, pp. S36–S40, 2013.

[2] P. P. Doghramji, “Integratingmodern concepts of insomnia andits contemporary treatment into primary care,” PostgraduateMedicine, vol. 126, no. 5, pp. 82–101, 2014.

[3] L. A. Irish, A. L. Dougall, D. L. Delahanty, and M. H. Hall,“The impact of sleep complaints on physical health and immuneoutcomes in rescue workers: a 1-year prospective study,” Psycho-somatic Medicine, vol. 75, no. 2, pp. 196–201, 2013.

[4] J. Fernandez-Mendoza, S. Shea, A. N. Vgontzas, S. L. Calhoun,D. Liao, and E. O. Bixler, “Insomnia and incident depression:role of objective sleep duration and natural history,” Journal ofSleep Research, 2015.

[5] Y. Li, A. N. Vgontzas, J. Fernandez-Mendoza et al., “Insomniawith physiological hyperarousal is associated with hyperten-sion,” Hypertension, vol. 65, no. 3, pp. 644–650, 2015.

[6] M. Sands-Lincoln, E. B. Loucks, B. Lu et al., “Sleep duration,insomnia, and coronary heart disease among postmenopausalwomen in the women’s health initiative,” Journal of Women’sHealth, vol. 22, no. 6, pp. 477–486, 2013.

[7] M. C. Gadinger, J. E. Fischer, S. Schneider, G. C. Fischer, G.Frank, and W. Kromm, “Female executives are particularlyprone to the sleep-disturbing effect of isolated high-strain jobs:a cross-sectional study inGerman-speaking executives,” Journalof Sleep Research, vol. 18, no. 2, pp. 229–237, 2009.

[8] C. S. Lopes, J. R. Robaina, and L. Rotenberg, “Epidemiology ofinsomnia: prevalence and risk factors,” in Can’t Sleep? Issues ofBeing an Insomniac, S. Sahoo, Ed., InTech, 2012.

[9] A. Palm, C. Janson, and E. Lindberg, “The impact of obesity andweight gain on development of sleep problems in a population-based sample,” Sleep Medicine, vol. 16, no. 5, pp. 593–597, 2015.

[10] M. M. Thakkar, R. Sharma, and P. Sahota, “Alcohol disruptssleep homeostasis,”Alcohol, vol. 11, no. 14, pp. 20115–20117, 2014.

[11] K. Kurotani, T. Kochi, A. Nanri et al., “Dietary patterns andsleep symptoms in Japanese workers: the Furukawa Nutritionand Health Study,” Sleep Medicine, vol. 16, no. 2, pp. 298–304,2015.

[12] E. Yoshioka, Y. Saijo, T. Kita, H. Satoh, M. Kawaharada, and R.Kishi, “Effect of the interaction between employment level andpsychosocial work environment on insomnia in male Japanesepublic service workers,” International Journal of BehavioralMedicine, vol. 20, no. 3, pp. 355–364, 2013.

[13] T. Akerstedt, P. Fredlund, M. Gillberg, and B. Jansson, “Workload and work hours in relation to disturbed sleep and fatiguein a large representative sample,” Journal of PsychosomaticResearch, vol. 53, no. 1, pp. 585–588, 2002.

[14] H.-C. Kim, B.-K. Kim, K.-B. Min, J.-Y. Min, S.-H. Hwang, andS.-G. Park, “Association between job stress and insomnia inKorean workers,” Journal of Occupational Health, vol. 53, no. 3,pp. 164–174, 2011.


[15] K. Nomura, M. Nakao, T. Takeuchi, and E. Yano, “Associationsof insomnia with job strain, control, and support among maleJapanese workers,” Sleep Medicine, vol. 10, no. 6, pp. 626–629,2009.

[16] A. H. de Lange, M. A. J. Kompier, T. W. Taris et al., “A hardday’s night: a longitudinal study on the relationships among jobdemands and job control, sleep quality and fatigue,” Journal ofSleep Research, vol. 18, no. 3, pp. 374–383, 2009.

[17] A. Ota, T. Masue, N. Yasuda et al., “Psychosocial job charac-teristics and insomnia: a prospective cohort study using theDemand-Control-Support (DCS) and Effort-Reward Imbal-ance (ERI) job stress models,” Sleep Medicine, vol. 10, no. 10, pp.1112–1117, 2009.

[18] R. Karasek, C. Brisson, N. Kawakami, I. Houtman, P. Bongers,and B. Amick, “The Job Content Questionnaire (JCQ): aninstrument for internationally comparative assessments of psy-chosocial job characteristics,” Journal of Occupational HealthPsychology, vol. 3, no. 4, pp. 322–355, 1998.

[19] R. Karasek and T. Theorell, Healthy Work—Stress, Productivity,and the Reconstruction of Working Life, Basic Books, New York,NY, USA, 1990.

[20] J. De Jonge, N. Van Vegchel, A. Shimazu, W. Schaufeli, and C.Dormann, “A longitudinal test of the demand-control modelusing specific job demands and specific job control,” Interna-tional Journal of Behavioral Medicine, vol. 17, no. 2, pp. 125–133,2010.

[21] T. S. Kristensen, H. Hannerz, A. Høgh, and V. Borg, “TheCopenhagen Psychosocial Questionnaire—a tool for the assess-ment and improvement of the psychosocial work environment,”Scandinavian Journal of Work, Environment and Health, vol. 31,no. 6, pp. 438–449, 2005.

[22] J. de Jonge, P. M. Le Blanc, M. C. W. Peeters, and H. Noordam,“Emotional job demands and the role ofmatching job resources:a cross-sectional survey study among health care workers,”International Journal of Nursing Studies, vol. 45, no. 10, pp. 1460–1469, 2008.

[23] H. A. Johannessen, H. M. Gravseth, and T. Sterud, “Psychoso-cial factors at work and occupational injuries: a prospectivestudy of the general working population in Norway,” AmericanJournal of Industrial Medicine, vol. 58, no. 5, pp. 561–567, 2015.

[24] N. van Vegchel, J. de Jonge, M. Soderfeldt, C. Dormann, andW. Schaufeli, “Quantitative versus emotional demands amongSwedish human service employees: moderating effects of jobcontrol and social support,” International Journal of StressManagement, vol. 11, no. 1, pp. 21–40, 2004.

[25] H. Gonge, L. D. Jensen, and J. P. Bonde, “Do psychosocial strainand physical exertion predict onset of low-back pain amongnursing aides?” Scandinavian Journal ofWork, Environment andHealth, vol. 27, no. 6, pp. 388–394, 2001.

[26] D. Zapf, C. Seifert, B. Schmutte, H. Mertini, and M. Holz,“Emotion work and job stressors and their effects on burnout,”Psychology and Health, vol. 16, no. 5, pp. 527–545, 2001.

[27] T. Theorell, A. Perski, T. Akerstedt et al., “Changes in job strainin relation to changes in physiological state: a longitudinalstudy,” Scandinavian Journal of Work, Environment and Health,vol. 14, no. 3, pp. 189–196, 1988.

[28] A. Ota, T. Masue, N. Yasuda, A. Tsutsumi, Y. Mino, and H.Ohara, “Association between psychosocial job characteristicsand insomnia: an investigation using two relevant job stressmodels—the demand-control-support (DCS) model and theeffort-reward imbalance (ERI) model,” Sleep Medicine, vol. 6,no. 4, pp. 353–358, 2005.

[29] N. Nishitani and H. Sakakibara, “Job stress factors, stressresponse, and social support in association with insomnia ofjapanesemale workers,” Industrial Health, vol. 48, no. 2, pp. 178–184, 2010.

[30] E. Pelfrene, P. Vlerick, F. Kittel, R. P. Mak, M. Kornitzer, and G.De Backer, “Psychosocial work environment and psychologicalwell-being: assessment of the buffering effects in the jobdemand-control(-support) model in BELSTRESS,” Stress andHealth, vol. 18, no. 1, pp. 43–56, 2002.

[31] D. C. Jarrin, I. Y. Chen, H. Ivers, and C. M. Morin, “The roleof vulnerability in stress-related insomnia, social support andcoping styles on incidence and persistence of insomnia,” Journalof Sleep Research, vol. 23, no. 6, pp. 681–688, 2014.

[32] P. C. Winwood and K. Lushington, “Disentangling the effectsof psychological and physical work demands on sleep, recoveryand maladaptive chronic stress outcomes within a large sampleof Australian nurses,” Journal of Advanced Nursing, vol. 56, no.6, pp. 679–689, 2006.

[33] S. A. Burgard and J. A. Ailshire, “Putting work to bed: stressfulexperiences on the job and sleep quality,” Journal of Health andSocial Behavior, vol. 50, no. 4, pp. 476–492, 2009.

[34] M. Loft and L. Cameron, “The importance of sleep: relation-ships between sleep quality and work demands, the prioriti-zation of sleep and pre-sleep arousal in day-time employees,”Work and Stress, vol. 28, no. 3, pp. 289–304, 2014.

[35] T. M. de Araujo and R. Karasek, “Validity and reliability ofthe job content questionnaire in formal and informal jobs inBrazil,” Scandinavian Journal of Work, Environment and Health,Supplement, no. 6, pp. 52–59, 2008.

[36] National Institute on Alcohol Abuse and Alcoholism (NIAAA),Moderate & binge drinking, 2013, http://www.niaaa.nih.gov/alcohol-health/overview-alcohol-consumption/moderate-binge-drinking.

[37] V. J. Felitti, R. F. Anda, D. Nordenberg et al., “Relationship ofchildhood abuse and household dysfunction to many of theleading causes of death in adults: the adverse childhood expe-riences (ACE) study,” American Journal of Preventive Medicine,vol. 14, no. 4, pp. 245–258, 1998.

[38] T. S. B. Magnago, M. T. L. Lisboa, R. H. Griep, R. C. Zeitoune,and J. P. Tavares, “Condicoes de trabalho de profissionais daenfermagem: avaliacao baseada nomodelo demanda-controle,”Acta Paulista de Enfermagem, vol. 23, no. 6, pp. 811–817, 2010.

[39] J. Dorrian, J. Paterson, D. Dawson, J. Pincombe, C. Grech,and A. E. Rogers, “Sleep, stress and compensatory behaviors inAustralian nurses and midwives,” Revista de Saude Publica, vol.45, no. 5, pp. 922–930, 2011.

[40] J. A. Ailshire and S. A. Burgard, “Family relationships andtroubled sleep among U.S. adults: examining the influences ofcontact frequency and relationship quality,” Journal of Healthand Social Behavior, vol. 53, no. 2, pp. 248–262, 2012.

[41] S. A. Stansfeld, C. Pike, S. McManus et al., “Occupations, workcharacteristics and common mental disorder,” PsychologicalMedicine, vol. 43, no. 5, pp. 961–973, 2013.

Research ArticleConsequences of Job Insecurity on the Psychological andPhysical Health of Greek Civil Servants

Dimitra Nella,1 Efharis Panagopoulou,1 Nikiforos Galanis,2

Anthony Montgomery,3 and Alexis Benos1

1School of Medicine, Faculty of Health Sciences, Aristotle University of Thessaloniki, University Campus, 55131 Thessaloniki, Greece2Department of Orthopaedics, “Papageorgiou” General Hospital, Medical School, Aristotle University of Thessaloniki,56403 Thessaloniki, Greece3Department of Educational and Social Policy, University of Macedonia, Egnatia Street 156, 54636 Thessaloniki, Greece

Correspondence should be addressed to Anthony Montgomery; [email protected]

Received 22 December 2014; Revised 18 May 2015; Accepted 19 May 2015


Copyright © 2015 Dimitra Nella et al. This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The aim of this study was to estimate the short term consequences of job insecurity associated with a newly introduced mobilityframework in Greece. In specific, the study examined the impact of job insecurity on anxiety, depression, and psychosomatic andmusculoskeletal symptoms, two months after the announcement of the mobility framework. In addition the study also examinedthe “spill over” effects of job insecurity on employees not directly affected by the mobility framework. Personal interviews usinga structured questionnaire were conducted for 36 university administrative employees awaiting repositioning, 36 coworkers notat risk, and 28 administrative employees of a local hospital not at risk. Compared to both control groups the employees in theanticipation phase of labor mobility had significantly worse scores for perceived stress, anxiety, depression, positive affect, negativeaffect, social support,marital discord, common somatic symptoms, and frequency ofmusculoskeletal pain.This study highlights theimmediate detrimental effects of job insecurity on the physical, psychological, and social functioning of employees.There is a needfor the development of front line interventions to prevent these effects from developing into chronic conditions with considerablecost for the individual and society in general.

1. Introduction

Since 2008 and the beginning of the Economic Crisis thathas affected most European countries including Greece, newforms of flexible and marginal employment have emergedresulting in a considerable increase in job insecurity. Job inse-curity is a social phenomenon, meaning that it is experiencedas a subjective perception about employment and unemploy-ment, and reflects the insecurity, uncertainty, powerlessnessand helplessness that occur when an individuals lacks theassurances that their job will remain stable [1]. It has beenstated that job insecurity is the most stressful aspect of theprocess leading to unemployment with a worse impact onemployees than unemployment itself [2].

Job insecurity has been defined as the subjectively per-ceived and undesired possibility to lose the present jobin the future, as well as the fear or worries related to

the possibility of job loss [1, 3]. It can be differentiatedbetween cognitive and affective job insecurity with the firstreferring to the cognitive probability of losing one’s job andthe latter referring to the fear and worry of losing one’s work.Another way to differentiate job insecurity is differentiatingbetween quantitative insecurity which refers to worryingabout the loss of job itself and qualitative which refers toworrying about losing important aspects of job, for example,salary, health insurance, and social life [1, 3, 4].

Job insecurity varies by race, ethnicity, and immigrationstatus [5]. In two nationally representative US samples, moreBlacks than non-Blacks experienced perceived job insecurity[6]. Immigrant women in Sweden were more likely to workin temporary jobs than native born women [7]. Erlinghagen[8] analyzed self-perceived job insecurity among 17 Europeancountries and found significant cross-country differencesin individuals’ perception of job insecurity. Not only were


http://dx.doi.org/10.1155/2015/673623


the findings driven by social-structural or institutional differ-ences, but also the perception of job insecurity was influencedby nation-specific unobserved characteristics (e.g., religious-ness, general assessment of job security, and basic trust infellow human beings).

Risk factors for experiencing job insecurity include beingmale, a finding linked to the traditional role of the breadwin-ner, age between 30–44 years old, lower educational status,being self-employed or working in the private practice, bluecollar work, and working in the manufacturing sector [9, 10].

Job insecurity has been linked to several adverse healthoutcomes. In regard to mental health it has been associatedwith psychosomatic symptoms, loss of self-esteem, anxiety,andminor psychiatric symptoms [1, 11–14]. In regard to phys-ical health, job insecurity has been associated with increasedmorbidity, lower levels of self-reported health, increasedincidence rates of hypertension, coronary heart disease, andmyocardial death [15, 16]. In addition, job insecurity hasbeen found to lead to restricted physical activity due tomusculoskeletal disorders such as low back pain and neckpain [17, 18]. In addition, anticipation of redundancy hasbeen shown to affect health behaviors such as exercise,dietary habits, and sleep [19–21]. Moreover, job insecurityis associated with increased use of healthcare services anddecreased compliance with occupational safety regulations[22].

Three theoretical models have been suggested to explainthe paths that lead from the negative consequences of jobinsecurity to employees’ health and wellbeing, Jahoda’s latentdeprivation model [23], the psychological contract theory[24, 25], and the vitamin model suggested by Probst andBrubaker [22]. Jahoda’s model suggests that the possibilityof losing one’s job threatens the satisfaction of needs suchas income and social contacts and leads to frustration. Thesecond model, concerning psychological contract theory,suggests that insecurity about retaining their posts is per-ceived by employees as a violation of an untold contract onbehalf of their employer and therefore their commitmentto their business and their well-being is affected. Finally,according to the vitamin model, job insecurity affects neg-atively employees’ wellbeing due to the associated feelings ofunpredictability and uncontrollability [26].

In 2012, the Greek government launched a reform of thenational civil service, as part of dealing with the ongoingfinancial crisis. The published law 4093/2012 highlightedthe need for the suspension of almost 2000 civil servantsthroughout the Greek public sector through a suggested“mobility framework for personnel.”The framework dictatedthat a percentage of employees of the civil service, followingan assessment process, would be transferred to other postswithin the countrywith 75%of their current salary. Followingthe transfer, and a second evaluation process, a percentage ofthose employees would eventually be made redundant fromthe civil service.

The reform programme in Greece has resulted in height-ened feelings of job instability and job insecurity. A reviewof the literature indicates that job instability/insecurity isassociated with psychologically ill health and impaired phys-ical health [5]. Indeed, chronic job insecurity appears to

have a dose-response relationship with self-reported healthand physical symptoms and increases the risk of minorpsychiatric morbidity [27, 28]. Greece has been profoundlyaffected by the global financial and economic crisis, withwide-ranging economic, social, and political consequences.In terms of job insecurity, the picture in Greece is bleak. In2015, the country is entering its seventh year of recessionand is still operating within severely constricted fiscal limits.Public and nonprofit mental health service providers havescaled back operations, shut down, or reduced staff; plansfor the development of child psychiatric services have beenabandoned; and state funding for mental health decreased by20% between 2010 and 2011, and by a further 55% between2011 and 2012 [29].There has been a substantial deteriorationin mental health status, with population surveys suggestinga significant increase in the prevalence of major depression,from 3.3% in 2008 to 8.2% in 2011, with economic hardshipbeing a major risk factor [30].

The aim of this study was to estimate the short termconsequences of job insecurity associated with the newlyintroduced mobility framework. In specific, the study exam-ined the impact of job insecurity on anxiety, depression, psy-chosomatic andmusculoskeletal symptoms, twomonths afterthe announcement of themobility framework. In addition thestudy also examined the “spill over” effects of job insecurityin employees not directly affected by themobility framework.

2. Materials and Methods

Employees of the administration department of the AristotleUniversity of Thessaloniki, expecting to be made redun-dant after the publication of the mobility framework, wererecruited for the study. In order to be included in the studyemployees had to be still working in their current post. Thecomparison group consisted of employees of the administra-tion department of an academic hospital not at risk of jobredundancy at the time of the study. To examine the “spillover” effects of job insecurity, a second comparison groupwas created, consisting of employees of the administrationdepartment of the same University, not at risk of losing theirjobs.

The study was approved by the Ethical board of theMedical School of the Aristotle University of Thessaloniki.All employees were informed about the study with an emailinviting them to participate and explaining the purpose of thestudy. A telephone contact number was given for employeesinterested in participating. After the first telephone contactwith the research team, an appointment was set for theinterview. Interviews took place during working hours in aprivate room in the employees’ working place. All interviewswere conducted by the same researcher. At the beginning ofthe interview participants were ensured about the anonymityof the procedure and gave their verbal consent. At theend of the interview all respondents were given contactinformation for counseling and support groups. Participantswere informedof their right towithdraw from the study at anypoint. Collected data were stored appropriately in a securelocation.


Table 1: Sample characteristics.

University officeworkers in labor

mobility

University officeworkers not at risk

Hospital officeworkers not at risk 𝑝

∗

𝑁 = 36 𝑁 = 36 𝑁 = 36

GenderMale 12 (33%) 13 (36%) 9 (32%) NSFemale 24 (67%) 23 (64%) 19 (68%)

Age 43.7 ± 7.5† 41.6 ± 7.4† 41.5 ± 7.9† NS∗∗

Marital statusNot married/not in relationship 8 (22%) 9 (25%) 6 (21%) NSMarried/in relationship 28 (78%) 27 (75%) 22 (79%)

EducationSecondary 23 (64%) 9 (25%) 9 (32%)

0.001Postsecondary 11 (31%) 15 (42%) 7 (25%)tertiary 2 (5%) 12 (33%) 12 (43%)

Employment contractFixed term 0 2 (6%) 1 (4%)

0.03Not fixed term 36 (100%) 27 (75%) 24 (85%)Permanent 0 7 (19%) 3 (11%)

∗

𝜒2 test (with Monte Carlo method when needed).∗∗ANOVA.†Mean ± SD.NS: nonsignificant.

The present study is a prospective study that investigatesa change in job security in a sample of employees, whowere compared with a suitable cohort of employees whodid not experience a change in job security. The employedmethodology is consistent with researchers who emphasizethe importance of exploring job insecurity via “naturalexperiments” [31, 32].

The questionnaire consisted of nine parts. For parts 1 to 7the Greek versions of the Perceived Stress Scale (PSS-10) [33,34], Hospital Anxiety and Depression Scale (HADS) [35–37],Multidimensional Scale of Perceived Social Support (MSPSS)[38, 39], the Beier-Sternberg Discord Questionnaire (DQ)[40], the Positive and Negative Affect Schedules (PANAS)[41], the Pennebaker Inventory of Limbic Languidness (PILL)[42], and the Health Behavior Inventory (HBI) [43, 44] wereused to assess perceived stress, anxiety and depression, socialsupport, marital discord, positive and negative affect, com-mon somatic symptoms, and health behaviors, respectively.Part 8 consisted of a body diagram with all main joints toestimate the frequency of musculoskeletal pain in a scale of1 to 10 based on the Nordic questionnaire [45] (MS scale)and part 9 consisted of demographic and medical historyinformation.

Statistical analysis was performed using the statisticalpackage IBM SPSS Statistics Standard v.20 and statisticalsignificance was set up at 𝑝 < 0.05. Patient demographicswere summarized using descriptive statistics. Scores for allscales used were calculated. For the HADS questionnaire,the cutoff score of 8 was applied suggested by the authorsto identify clinical cases of anxiety disorder and depression.

Due to the fact that none of the scales were normallydistributed, nonparametric statistics were employed. Medianand interquartile range were chosen as descriptive measures.Kruskal Wallis tests were used to compare all study groups.In case of significant differences, Mann-Whitney𝑈 tests wereused to compare scores between two groups.

3. Results

Thefinal sample consisted of 36 out of 97 (37% response rate)of university employees at risk of being made redundant, 36out of 114 (31.5% response rate) of university administrativeemployees not at present risk of labor mobility, and 28 out of105 hospital administrative employees (26% response rate).

Demographic characteristics of the sample are summa-rized in Table 1.

Results showed that, with the exception ofmarital discord(DQ), scores for all other aspects of psychosocial and physicalhealth were significantly different between all three groups(Table 2).

Further comparisons using Mann-Whitney 𝑈 test indi-cated that the “job insecurity” group had higher scores forperceived stress (𝑝 < 0.001), anxiety (𝑝 < 0.001), depression(𝑝 < 0.001), negative affect (𝑝 < 0.001), common somaticsymptoms (𝑝 < 0.001), andmusculoskeletal pain (𝑝 < 0.001)and lower scores for positive affect (𝑝 < 0.001) and perceivedsocial support (𝑝 = 0.011 and 𝑝 = 0.006) compared to eachone of the control groups, while no difference was shownbetween the two control groups (Table 3).


Table 2: Scores for all scales used concerning psychological and physical health and social support.

Investigation group Control group 1 Control group 2

𝑝∗University office workers in labor mobility University office workers not at risk Hospital office workers not at risk

𝑁 = 36 𝑁 = 36 𝑁 = 36Median (interquartile range)

PSS-10 27.0 (4.5) 8.5 (5.75) 6.5 (11.5) <0.001HADSA 16.0 (3.75) 4.0 (4.5) 3.5 (5.25) <0.001HADSD 10.0 (4.0) 2.0 (1.75) 2.0 (3.0) <0.001PANAS PA 28.0 (7.5) 41.0 (5.75) 44.0 (6.25) <0.001PANAS NA 27.0 (8.75) 15.5 (3.0) 17.0 (5.0) <0.001MSPSS 5.79 (1.15) 6.33 (1.02) 6.42 (1.29) 0.008DQ∗∗ 2.7 (2.1) 2.1 (1.05) 2.45 (1.0) NSPILL 4.23 (2.52) 1.1 (0.39) 1.15 (0.5) <0.001MS 2.67 (2.13) 1.08 (1.0) 1.0 (0.67) <0.001∗Kruskal Wallis test.∗∗For scale DQ group sizes are 27, 26, and 22, respectively.NS: nonsignificant.

Table 3: Comparisons between groups for all scales investigated.

Investigation groupversus control

group 1

Investigation groupversus control

group 2

Control group 1versus control

group 2

Investigation groupversus controlgroup 1 + 2

PSS-10 U = 35.0 𝑈 = 30.5 𝑈 = 463.5 𝑈 = 65.5p < 0.001 p < 0.001 p = NS p < 0.001

HADSA U = 25.5 𝑈 = 31.0 𝑈 = 487.5 𝑈 = 56.5p < 0.001 p < 0.001 p = NS p < 0.001

HADSD U = 45.0 𝑈 = 32.0 𝑈 = 486.0 𝑈 = 77.0p < 0.001 p < 0.001 p = NS p < 0.001

PANAS PA U = 136.0 𝑈 = 110.5 𝑈 = 356.0 𝑈 = 246.5p < 0.001 p < 0.001 p = NS p < 0.001

PANAS NA U = 62.5 𝑈 = 71.5 𝑈 = 406.0 𝑈 = 134.0p < 0.001 p < 0.001 p = NS p < 0.001

MSPSS 𝑈 = 423.0 𝑈 = 303.0 𝑈 = 454.5 𝑈 = 726.0p = 0.011 p = 0.006 p = NS p = 0.002

DQ 𝑈 = 242.0 𝑈 = 228.5 𝑈 = 246.5 𝑈 = 470.5p = NS p = NS p = NS p = 0.05

PILL 𝑈 = 64.5 𝑈 = 38.0 𝑈 = 483.5 𝑈 = 102.5p < 0.001 p < 0.001 p = NS p < 0.001

MS 𝑈 = 238.5 𝑈 = 182.0 𝑈 = 486.0 𝑈 = 420.5p < 0.001 p < 0.001 p = NS p < 0.001

𝑈 = Mann-Whitney 𝑈 test.NS: nonsignificant.

In terms of clinical cases of depression and anxiety,35 (97%) of participants in the job insecurity group wereclassified as anxiety disorder cases compared to the 16 (25%)of participants in each one of the control groups (𝜒2 = 51.9,𝑝 < 0.001). This corresponds to an OR = 105 probabilities(95% CI 13.3–829.4, 𝑝 < 0.001) for office workers in the jobinsecurity group to develop an anxiety disorder. Likewise forsubscale HADS-D for depression, 31 (86%) of participants inthe job insecurity group were defined as cases for depression

against 2 (3%) participants in both control groups (𝜒2 =71.8, 𝑝 < 0.001 Monte Carlo method), correspondingto a probability of OR = 192.2 (95% CI 35.3–1047.4, 𝑝 <0.001) for participants in the job insecurity group to developdepression.

Protective and high risk health behaviors and health careuse were assessed and compared between the investigationgroup and the two groups together (Table 4). Office workersin the job insecurity group reported eating more fast food


Table 4: Health behavior and health care service use.

Investigation group Control group 1 + 2𝑝∗

𝑁 = 36 𝑁 = 64

Median (interquartile range)Protective health behaviorsBreakfasts per week 3.5 (6.75) 5.0 (4.0) NSRegular meals per week 2.0 (2.0) 2.0 (1.0) NSHours of night sleep per day 6.0 (2.0) 7.0 (1.0) <0.001Workout sessions per week 0.0 (0.0) 1.5 (3.0) 0.001High risk health behaviorsSnacks per week 4.0 (4.0) 2.0 (3.75) 0.001Cigarettes smokes per day 10.0 (20.0) 0.0 (10.0) 0.001Coffees per day 2.0 (2.0) 1.5 (1.0) 0.001Frequency of alcohol drinking per week 1.0 (2.0) 0.0 (1.0) NSAlcoholic drinks per week 0.5 (1.75) 0.0 (1.0) NSSleeping pills per week 0.3 ± 1.2∗∗ 0.1 ± 0.9∗∗ NSTranquilizers per week 0.3 ± 1.2∗∗ 0.1 ± 0.9∗∗ NSNonpharmaceutical methods to achieve

sleep per week 1.0 ± 1.8∗∗ 0.4 ± 1.4∗∗ 0.015Painkillers per week 1.8 ± 2.1∗∗ 0.3 ± 1.0∗∗ <0.001

Health care services useMedical consultations per month 0.4 ± 0.8∗∗ 0.2 ± 0.6∗∗ 0.048Sick days per month 0.3 ± 0.9∗∗ 0.0 ± 0.1∗∗ 0.034∗Mann-Whitney 𝑈 test.∗∗Median ± SD.NS: nonsignificant.

meals perweek thanworkers not at risk (𝑝 = 0.001).They alsoconsumed more coffees (𝑝 = 0.001) and cigarettes per day(𝑝 = 0.001). In addition they engaged in less physical exerciseper week (𝑝 = 0.001), slept less during the night (𝑝 < 0.001),and usedmoremedication for pain relief (𝑝 < 0.001). Finally,employees feeling insecure about their job visited the doctormore often during the past month (𝑝 = 0.048) and reportedmore absent days due to sickness (𝑝 = 0.034) compared tothe comparison groups.

4. Discussion

Results show that employees at risk of losing their jobsshowed higher levels of perceived stress, anxiety, depression,and negative feelings and lower levels of positive feelingscompared to employees not at risk of losing their jobs. Thesefindings are in agreement with previous studies [11, 31].Results also showed that 97% and 86% of the group at riskof joblessness were classified as clinical cases of anxiety anddepression, respectively. As this study was carried out withinthe first three months of the announcement of the laborshortage measures, results of the study reflect the immediate,short term reaction. However, our results need to be treatedwith caution as we are unaware of the long term impact of theeconomic crisis inGreece [46] and its potential impact on oursample.

Employees in the job insecurity group also reportedreceiving less social support. Additionally, while the resultwas not statistically significant the job insecurity group didreport a higher degree of discord with their spouses. Jenkinset al. have also reported the negative impact of job insecurityon employees’ marriages resulting in higher rate of divorces[47]. However, the present data indicates that the relationshipbetween job insecurity and health problems in not related tomarital discord.

Employees in the job insecurity group reported higherfrequency of common somatic symptoms. Among the symp-toms that were most frequently reported were chest pains,racing heart, and choking sensation, symptoms indicativeof cardiovascular impairment. Previous studies have alsoexplored the link between job insecurity and cardiovascu-lar symptoms [14] and it is reported that feeling insecureabout retaining one’s job is a risk factor for coronary heartdisease [48]. Both control groups showed significantly lowerfrequency of musculoskeletal pain compared to employees inanticipation of job loss in accordance with studies reportingthat job insecurity increases the risk for low back pain [16]and is a predictor for musculoskeletal pain in the limbs [49].

In terms of health behaviors participants in the jobinsecurity group smoked more cigarettes per day, exercisedless, slept fewer hours every night (under 6 hours per night),and used more frequently painkillers, a finding which can belinked to the increased reported frequency ofmusculoskeletal


pain in the job insecurity group. These results are in agree-ment with previous studies showing that job insecurity wasa risk factor for work-related sleep problems [18] and thatmore insecure employees tend to drink alcohol, smoke, andnot exercise [19, 20, 50].

The reported increased use of health resources by partic-ipants in the job insecurity group is in contrast to previousstudies showing that the economic crisis in Greece hasresulted in a reduction in visits to the doctor [46].

Results showed no spillover effects of the negative conse-quences of job insecurity on employees working in proximitybut not currently at risk of losing their job. This could betime-related as secondary effects of insecurity might needmore time to develop. For example in the study of Lang et al.musculoskeletal problems of employees that survived adownsizing of staff were stronger relative to musculoskeletalsickness absences measured for an extended period coveringtwo subsequent years after downsizing [17], suggesting thatnegative effects may be more pronounced later on.

One of the main study limitations is that no subjectivemeasure of job insecurity was used to assess job insecurity.Employees whose names were included in the first mobilityscheme which was introduced by the Greek Governmentwere considered as experiencing job insecurity. Our jobinsecurity group, individuals on the mobility scheme, wereself-selected but we were unable to control for the long termimpact of the crisis our participants [46]. This methodologyhas been previously used in other studies assessing the effectsof attributed job insecurity [19, 20, 51]. Future studies shouldalso include subjective assessments of insecurity in orderto potentially identify risk and protective factors that canaggravate or alleviate the threat of an objective situation ofanticipated job loss. In addition the small, nonrandomisedsample size does not permit generalizability of findings toother sectors. Finally, the cross-sectional nature of the studyand the fact that it relied on self-report limit our ability to beconclusive. However, the data collected on musculoskeletalpain, demographics, and medical information used a reliablemeasure that reduced the effect of self-report.

5. Conclusion

This study highlights the immediate detrimental effects of jobinsecurity on physical, psychological, and social functioningof employees. Results of the study also highlight the needfor development of front line interventions to prevent theseeffects from developing into chronic conditions with con-siderable cost for the individual and society in general. Interms of developing interventions, education appears to be acritical factor. For example, Perlman and Bobak [52] assessedthe contribution of unstable employment to mortality inPosttransition Russia and found that unemployment andjob insecurity were significant predictors of mortality. Oneaspect of the study is particularly noteworthy with regardto the importance of education. Perlman and Bobak foundthat education seemed to provide a protective factor againstsome indicators of unstable employment, independently ofoccupation. Although the reasons are uncertain, it is possible

that education may provide resilience or coping skills, assuggested by the association between education and higherperceived control [53] or depressive symptoms [54].



References

[1] H. de Witte, “Job insecurity: review of the international litera-ture on definitions, prevalence, antecedents and consequences,”SA Journal of Industrial Psychology, vol. 31, no. 4, pp. 1–6, 2005.

[2] R. S. Lazarus and S. Folkman, Stress Appraisal and Coping,Springer, New York, NY, USA, 1984.

[3] T. Vander Elst, H. de Witte, and N. de Cuyper, “The jobinsecurity scale: a psychometric evaluation across five Europeancountries,” European Journal of Work and Organizational Psy-chology, vol. 23, no. 3, pp. 364–380, 2014.

[4] H. Chung and W. van Oorschot, “Employment insecurity ofEuropean individuals during the financial crisis. A multi-level approach,” in Reconciling Work and Welfare in Europe—A Network of Excellence of the European Commision’s SixthFramework Programme, The University of Edinburgh, 2010.

[5] P. A. Landsbergis, J. G. Grzywacz, and A. D. Lamontagne,“Work organization, job insecurity, and occupational healthdisparities,”American Journal of Industrial Medicine, vol. 57, no.5, pp. 495–515, 2014.

[6] S. A. Burgard, J. E. Brand, and J. S. House, “Perceived jobinsecurity andworker health in theUnited States,” Social Scienceand Medicine, vol. 69, no. 5, pp. 777–785, 2009.

[7] S. Akhavan, C. Bildt, and S. Wamala, “Work-related healthfactors for female immigrants in Sweden,” Work, vol. 28, no. 2,pp. 135–143, 2007.

[8] M. Erlinghagen, “Self-perceived job insecurity and social con-text: a multi-level analysis of 17 European countries,” EuropeanSociological Review, vol. 24, no. 2, pp. 183–197, 2008.

[9] K. Naswall and H. De Witte, “Who feels insecure in Europe?Predicting job insecurity frombackground variables,”Economicand Industrial Democracy, vol. 24, no. 2, pp. 189–215, 2003.

[10] T. Alterman, S. E. Luckhaupt, J. M. Dahlhamer, B. W. Ward,and G. M. Calvert, “Job insecurity, work-family imbalance,and hostile work environment: prevalence data from the 2010National Health Interview Survey,” American Journal of Indus-trial Medicine, vol. 56, no. 6, pp. 660–669, 2013.

[11] S. A. Burgard, L. Kalousova, and K. S. Seefeldt, “Perceived jobinsecurity and health: the michigan recession and recoverystudy,” Journal of Occupational and Environmental Medicine,vol. 54, no. 9, pp. 1101–1106, 2012.

[12] M. Sverke, J. Hellgren, and K. Naswall, “No security: a meta-analysis and review of job insecurity and its consequences,”Journal of Occupational Health Psychology, vol. 7, no. 3, pp. 242–264, 2002.

[13] M. Sverke, J. Hellgren, and K. Naswall, Job Insecurity: ALiterature Review, edited by: Joint Programme forWorking LifeResearch in Europe (SALTSA), National Institute for WorkingLife, Stockholm, Sweden, 2006.

[14] G. H.-L. Cheng and D. K.-S. Chan, “Who suffers more from jobinsecurity? A meta-analytic review,” Applied Psychology, vol. 57,no. 2, pp. 272–303, 2008.


[15] S. Lee, G. A. Colditz, L. F. Berkman, and I. Kawachi, “Prospec-tive study of job insecurity and coronary heart disease in USwomen,” Annals of Epidemiology, vol. 14, no. 1, pp. 24–30, 2004.

[16] K. D. Laszlo, K. Engstrom, J. Hallqvist, A. Ahlbom, and I. Jan-szky, “Job insecurity and prognosis after myocardial infarction:the SHEEP Study,” International Journal of Cardiology, vol. 167,no. 6, pp. 2824–2830, 2013.

[17] J. Lang, E. Ochsmann, T. Kraus, and J.W. B. Lang, “Psychosocialwork stressors as antecedents of musculoskeletal problems:a systematic review and meta-analysis of stability-adjustedlongitudinal studies,” Social Science and Medicine, vol. 75, no.7, pp. 1163–1174, 2012.

[18] M. Kivimaki, J. Vahtera, J. E. Ferrie, H. Hemingway, and J.Pentti, “Organisational downsizing and musculoskeletal prob-lems in employees: a prospective study,” Occupational andEnvironmental Medicine, vol. 58, no. 12, pp. 811–817, 2001.

[19] J. B. Park, A. Nakata, N. G. Swanson, and H. Chun, “Organi-zational factors associated with work-related sleep problems ina nationally representative sample of Korean workers,” Interna-tional Archives of Occupational and Environmental Health, vol.86, no. 2, pp. 211–222, 2013.

[20] J. E. Ferrie, M. J. Shipley, M. G. Marmot, S. A. Stansfeld, andG. D. Smith, “An uncertain future: the health effects of threatsto employment security in white-collar men and women,”American Journal of Public Health, vol. 88, no. 7, pp. 1030–1036,1998.

[21] J. E. Ferrie, M. J. Shipley, M. G. Marmot, S. Stansfeld, and G. D.Smith, “Health effects of anticipation of job change and non-employment: longitudinal data from the Whitehall II study,”British Medical Journal, vol. 311, no. 7015, pp. 1264–1269, 1995.

[22] T. M. Probst and T. L. Brubaker, “The effects of job insecurityon employee safety outcomes: cross-sectional and longitudinalexplorations.,” Journal of Occupational Health Psychology, vol. 6,no. 2, pp. 139–159, 2001.

[23] M. Jahoda, Employment and Unemployment. A Social-Psychological Analysis, The Psychology of Social Issues,Cambridge University Press, Cambridge, UK, 1982.

[24] M. Sverke, J. Hellgren, K. Naswall, A. Chirumbolo, H. deWitte, and S. Goslinga, Job Insecurity and Union Membership:European Unions in the Wake of Flexible Production, Work andSociety, Peter Lang Publisher, Brussels, Belgium, 2004.

[25] N. De Cuyper and H. De Witte, “The impact of job insecurityand contract type on attitudes, well-being and behaviouralreports: a psychological contract perspective,” Journal of Occu-pational and Organizational Psychology, vol. 79, no. 3, pp. 395–409, 2006.

[26] P. Warr, “A vitamin model of the environment and mentalhealth,” in Work, Unemployment and Mental Health, p. 320,Clarendon Press, Oxford, UK, 1987.

[27] M. Marmot, K. Newman, J. Ferrie, and S. Stansfeld, TheContribution of Job Insecurity to Socio-Economic Inequalities,Lancaster University, Health Variations Programme, Lancaster,UK, 2001.

[28] C. Muntaner, H. Chung, I. Kim, and J. Benach, “Populationsat special health risk: workers,” in International Encyclopedia ofPublicHealth, K.Heggenhougen and S.Quah, Eds., pp. 285–301,Academic Press, San Diego, Calif, USA, 2008.

[29] D. C. Anagnostopoulos and E. Soumaki, “The state of childand adolescent psychiatry in Greece during the internationalfinancial crisis: a brief report,” European Child and AdolescentPsychiatry, vol. 22, no. 2, pp. 131–134, 2013.

[30] M. Economou, M. Madianos, L. E. Peppou, A. Patelakis, and C.N. Stefanis, “Major depression in the era of economic crisis: areplication of a cross-sectional study across Greece,” Journal ofAffective Disorders, vol. 145, no. 3, pp. 308–314, 2013.

[31] J. E. Ferrie, M. J. Shipley, S. A. Stansfeld, and M. G. Marmot,“Effects of chronic job insecurity and change in job security onself reported health, minor psychiatric morbidity, physiologicalmeasures, and health related behaviours in British civil servants:theWhitehall II study,” Journal of Epidemiology and CommunityHealth, vol. 56, no. 6, pp. 450–454, 2002.

[32] S. V. Kasl and B. A. Jones, “An epidemiological perspectiveon research design, measurement and surveillance strategies,”in Handbook of Occupational Health Psychology, J. C. Quickand L. E. Tetrick, Eds., American Psychological Association,Washington, DC, USA, 2002.

[33] S. Cohen, T. Kamarck, and R. Mermelstein, “A global measureof perceived stress,” Journal of Health and Social Behavior, vol.24, no. 4, pp. 385–396, 1983.

[34] E. Andreou, E. C. Alexopoulos, C. Lionis et al., “Perceived stressscale: reliability and validity study in Greece,” InternationalJournal of Environmental Research and Public Health, vol. 8, no.8, pp. 3287–3298, 2011.

[35] A. S. Zigmond and R. P. Snaith, “The hospital anxiety anddepression scale,” Acta Psychiatrica Scandinavica, vol. 67, no. 6,pp. 361–370, 1983.

[36] R. P. Snaith and A. S. Zigmond, The Hospital Anxiety andDepression Scale Manual, NFER-Nelson, 1994.

[37] I. Michopoulos, A. Douzenis, C. Kalkavoura et al., “Hospitalanxiety and depression scale (HADS): validation in a Greekgeneral hospital sample,” Annals of General Psychiatry, vol. 7,article 4, 2008.

[38] G. D. Zimet, N. W. Dahlem, S. G. Zimet, and G. K. Farley, “Themultidimensional scale of perceived social support,” Journal ofPersonality Assessment, vol. 52, no. 1, pp. 30–41, 1988.

[39] P. Theofilou, “Translation and cultural adaptation of the Mul-tidimensional Scale of Perceived Social Support (MSPSS) forGreece,” Health Psychology Research. In press.

[40] E. G. Beier and D. P. Sternberg, “Marital communication,”Journal of Communication, vol. 27, no. 3, pp. 92–97, 1977.

[41] D. Watson, L. A. Clark, and A. Tellegen, “Development andvalidation of brief measures of positive and negative affect: thePANAS scales,” Journal of Personality and Social Psychology, vol.54, no. 6, pp. 1063–1070, 1988.

[42] J. W. Pennebaker, “The PILL: a trait measure,” inThe Psychologyof Physical Symptoms, pp. 169–171, Springer, New York, NY,USA, 1982.

[43] I. Moustou, E. Panagopoulou, J. Montgomery, and A. Benos,“Burnout predicts health behaviors in ambulance workers,”TheOpen Occupational Health and Safety Journal, vol. 2, pp. 16–18,2010.

[44] E. Tsiga, E. Panagopoulou, and D. Niakas, “Health promotionacross occupational groups: one size doesn’t fit all,” Occupa-tional Medicine. In press.

[45] I. Kuorinka, B. Jonsson, A. Kilbom et al., “Standardised Nordicquestionnaires for the analysis of musculoskeletal symptoms,”Applied Ergonomics, vol. 18, no. 3, pp. 233–237, 1987.

[46] A. Kentikelenis, M. Karanikolos, I. Papanicolas, S. Basu, M.McKee, andD. Stuckler, “Health effects of financial crisis: omensof a Greek tragedy,”TheLancet, vol. 378, no. 9801, pp. 1457–1458,2011.


[47] R. Jenkins, A. MacDonald, J. Murray, and G. Strathdee, “Minorpsychiatric morbidity and the threat of redundancy in a profes-sional group,” Psychological Medicine, vol. 12, no. 4, pp. 799–807,1982.

[48] J. E. Ferrie, M. Kivimaki, M. J. Shipley, G. Davey Smith, and M.Virtanen, “Job insecurity and incident coronary heart disease:the Whitehall II prospective cohort study,” Atherosclerosis, vol.227, no. 1, pp. 178–181, 2013.

[49] J. Bugajska, D. Zołnierczyk-Zreda, A. Jędryka-Goral et al.,“Psychological factors at work and musculoskeletal disorders:a one year prospective study,” Rheumatology International, vol.33, no. 12, pp. 2975–2983, 2013.

[50] G. M. H. Swaen, U. Bultmann, I. Kant, and L. G. P. M.van Amelsvoort, “Effects of job insecurity from a workplaceclosure threat on fatigue and psychological distress,” Journal ofOccupational & EnvironmentalMedicine, vol. 46, no. 5, pp. 443–449, 2004.

[51] J. E. Ferrie, “Is job insecurity harmful to health?” Journal of theRoyal Society of Medicine, vol. 94, no. 2, pp. 71–76, 2001.

[52] F. Perlman and M. Bobak, “Assessing the contribution ofunstable employment to mortality in posttransition Russia:prospective individual-level analyses from the Russian longi-tudinal monitoring survey,” The American Journal of PublicHealth, vol. 99, no. 10, pp. 1818–1825, 2009.

[53] J. Mirowsky and C. E. Ross, Education, Social Status and Health,Aldine de Gruyter, New York, NY, USA, 2003.

[54] A. Nicholson, H. Pikhart, A. Pajak et al., “Socio-economicstatus over the life-course and depressive symptoms inmen andwomen in Eastern Europe,” Journal of Affective Disorders, vol.105, no. 1–3, pp. 125–136, 2008.

Research ArticleTowards a Job Demands-Resources Health Model: EmpiricalTesting with Generalizable Indicators of Job Demands, JobResources, and Comprehensive Health Outcomes

Rebecca Brauchli, Gregor J. Jenny, Désirée Füllemann, and Georg F. Bauer

Division of Public and Organizational Health, Epidemiology, Biostatistics and Prevention Institute, University of Zurich,Hirschengraben 84, 8001 Zurich, Switzerland

Correspondence should be addressed to Rebecca Brauchli; [email protected]

Received 16 February 2015; Accepted 27 April 2015


Copyright © 2015 Rebecca Brauchli et al. This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.

Studies using the Job Demands-Resources (JD-R) model commonly have a heterogeneous focus concerning the variables theyinvestigate—selective job demands and resources as well as burnout and work engagement. The present study applies the rationaleof the JD-R model to expand the relevant outcomes of job demands and job resources by linking the JD-R model to the logic ofa generic health development framework predicting more broadly positive and negative health. The resulting JD-R health modelwas operationalized and tested with a generalizable set of job characteristics and positive and negative health outcomes among aheterogeneous sample of 2,159 employees. Applying a theory-driven and a data-driven approach, measures which were generallyrelevant for all employees were selected. Results from structural equationmodeling indicated that themodel fitted the data.Multiplegroup analyses indicated invariance across six organizations, gender, job positions, and three times ofmeasurement. Initial evidencewas found for the validity of an expanded JD-R health model. Thereby this study contributes to the current research on jobcharacteristics and health by combining the core idea of the JD-R model with the broader concepts of salutogenic and pathogenichealth development processes as well as both positive and negative health outcomes.

1. Introduction and Study Aim

In the field of occupational health and safety it is wellknown that job characteristics affect workers’ health andwell-being [1, 2]. With the introduction of new working methodsand procedures during the 20th century a number of newhealth and safety hazards at work emerged. Many countries,especially in the European Union, aim to systematicallyidentify factors that lead to occupational health [1]. In theUK, for example, the Health and Safety Executive (HSE)Management Standards Indicator Tool was developed andis increasingly used by organizations to monitor workingconditions that can lead to stress [3].

Besides, a lot of research was conducted not only toassess indicators for work-related stress and well-being butalso to identify the underlying mechanisms that lead fromjob characteristics to health and well-being. Among others,

very well established is the so-called demand-control model(DCM) [4]; a combination of high job demands and low jobcontrolwill lead to job strain.An alternativemodel, the effort-reward imbalance (ERI) model [5], assumes that job strainis the result of an imbalance between effort and reward andmay lead to negative health outcomes, such as cardiovasculardiseases. However, most studies on the DCM and ERI modelhave been restricted to a very limited set of independent vari-ables thatmay not be relevant for all kinds of jobs and persons[6]. To meet this limitation the Job Demands-Resources (JD-R)model [6–8]was developed at the beginning of the century.At its heart lies the assumption that job characteristics can beclassified into two general categories: job demands and jobresources [6]. Job demands refer to physical, mental, social,or organizational job characteristics that require sustainedphysical or psychological effort, thus being associated withphysiological and/or psychological costs [7]. Job resources


http://dx.doi.org/10.1155/2015/959621


refer to those physical, mental, social, or organizationaljob characteristics that may be functional in meeting jobrequirements and thus reduce the associated physiologicaland/or psychological costs and stimulate personal growthand development [6, 7, 9]. The second assumption of themodel is that the two categories of job characteristics evoketwo relatively independent psychological processes whichare considered to play a crucial role in the development ofburnout and engagement [6, 10].The first process—the healthimpairment process—explains the exhausting impact of jobdemands, such as poorly designed jobs (e.g., jobs with lowjob control) or chronic job demands (e.g., work overloador time pressures) on burnout [6, 11]. The second process—the motivational process—suggests that job resources exerta motivating potential and lead to high work engagement.There has been considerable empirical support for thesetwo processes and their impact on burnout and engagementand consequently also on organizational outcomes [12–14].Further, the JD-R model assumes that job resources are notonly related towork engagement but also to burnout, whereasjob demands are strongly related to burnout but not or onlyweakly related to engagement [11, 15].

The JD-R model has been offered as a generic frameworkto overcome the limited focus of previous stress modelssuch as DCM and ERI model [6]. Thereby it provides broadcategories of physical, mental, social, or organizational jobcharacteristics to be included. Since studies using this modelare highly diverse regarding the job demands and resourcesincluded, it is difficult to compare and aggregate findingsregarding key job demands and resources across differentstudies.

Moreover, in line with the origin of the JD-Rmodel, moststudies using thismodel focus on burnout and engagement asrelevant outcomes [6]. An increasing number of studies con-sider further positive outcomes such as innovativeness, lifesatisfaction, organizational commitment, perceived health,workability, or happiness and negative outcomes, such asabsenteeism, accidents, unsafe behaviors, physical ill health,or turnover intention [15].However, these studiesmostly con-sider these outcomes as being mediated by work engagementand burnout as suggested by the JD-R model.

To study the direct effect of job characteristics and biopsy-chosocial health outcomes other than burnout and engage-ment, a balanced conceptualization of the elements and pathsof the JD-R model is needed. Currently, on the predictorside, the model indeed comprehensively acknowledges thatboth job demands and job resources can be physical, mental,and social in nature [6]. In contrast, the dependent variablesdo not explicitly consider physical, mental, and social healthoutcomes but are mostly limited to psychological outcomes,mainly burnout and engagement. Consequently, the pathslinking job characteristics to these focused outcomes are alsoprimarily conceptualized as psychological processes (healthimpairment and motivational process).

2. Developing the JD-R Health Model

The present study aims to develop and test a comprehensiveJD-R health model with paths linking job demands and

job resources to both negative and positive biopsychosocialhealth outcomes. Thereby we link the JD-R model with theconcept of salutogenesis [16] and a generic health develop-ment framework applying the conception of positive andnegative health [17].

2.1. Processes of Health Development: Pathogenesis and Salu-togenesis. Antonovsky [16, 18] proposed to complete theconcept of pathogenesis, which examines how diseasesdevelop, by the concept of salutogenesis, which explainshow health is maintained or strengthened. Antonovsky, whounderstood health as a “ease-disease-continuum” rangingfrom minimal to maximal health [16, 18], researched how“Generalized Resistance Resources (GRR)” and “Sense ofCoherence (SoC)” contributed to the maintenance of healthin potentially harming environments. The health devel-opment model [17] links the salutogenic and pathogenicprocesses by showing how risk factors are related to diseaseoutcomes, whereas resources are related to positive healthoutcomes. This dual path is in the same line of thinkingas general stress research and in particular as the JD-Rmodel: demands that are appraised as threats can lead to ahealth impairment process—a pathogenic process—straininga person physically, draining him/her mentally, and isolatinghim/her socially, thus harming his/her self-reproduction.This process can bemitigated by the presence of resources (or“GRR” in Antonovsky’s terminology) and SoC, influencingappraisal as well as coping and recovery processes—in theterms of Antonovsky a salutogenic process. Additionally,as the JD-R model has postulated and empirically proven,resources stimulate personal growth and development: Aperson draws on resources not only to be resilient in theface of potential harmful situations and events, but also tostrengthen his/her standing in life and work and to achievehis/her goals. This process leads to a state of energy andvigor, which can be understood in terms of positive healthas self-fulfillment. This again puts a person in the positionto further build resources and thus protect him/her fromnegative health outcomes as well as strengthen his positivehealth status, as research on gain cycles shows [19, 20].Although Antonovsky’s original salutogenic model limits theterm salutogenesis to the coping and recovering process asdescribed above, we propose to apply the term also to theprocess of resources leading to positive health [17].

2.2. Negative andPositiveHealth. Positive andnegative healthrelate to corresponding conceptions spanning more than 60years, from the preamble of the World Health Organization(WHO) in 1946 [21] to Seligman’s proposal for positivehealth in 2008 [22]. The WHO (1946) defines health as a“(. . .) state of complete physical, mental and social well-being and not merely the absence of disease or infirmity”(p. 2). Similar definitions have been developed early formental health [23], subsequently building evidence thathuman beings can “flourish” emotionally and socially despitemental disorders [24]. Similarly, Seligman [22] proposed aconceptual framework for positive health comprising threecategories: subjective (“when a person feels great, defined


by high ends of measures of several psychological states”),biological (“the positive ends of physiological function andanatomical structure distributions”), and functional (“howwell does the individual function?”). Taken together, healthhas positive and negative axes, covers physical, mental, andsocial facets of human life, and has emotional and functionalcomponents. Following this, we define negative health asimpaired physical, mental, and social self-reproduction, whichis traditionally linked tomedical classification systems. Anal-ogously, we define positive health as physical, mental, andsocial self-fulfillment, forwhichmuch research has emerged inrecent years (see Journal of Positive Psychology). Examples ofnegative health are painful musculoskeletal disorders inhibit-ingmovement, anxiety states anddepressivemood, and socialalienation and exclusion. Examples of positive health areenergetic fitness, joy and happiness, and being embedded inharmonious relationships. Clearly, both positive and negativehealth are interrelated but still have independent characteris-tics; that is, one can be physically impaired and still mentallyand socially fulfilled. Further, these aspects of impaired self-reproduction and of self-fulfillment can be operationalizeddomain-specific (e.g., joy resulting from work situations) orunspecific (e.g., general happiness in life).

2.3. JD-R HealthModel. The JD-R healthmodel assumes thatjob demands directly lead to negative health via a pathogenicpath whereas job resources directly lead to positive healthas well as negative health via a salutogenic path. Therebywe assume that job resources have a beneficial impact onnegative health since themore resources people have availablethe easier they recover from demands. In accordance withUK Health and Safety Executive [3] we assume that it isreasonable to identify a generalizable set of indicators (jobcharacteristics) predicting work-related and general health.If we are able to demonstrate the model’s stability we increaseits public health impact.

From both the perspectives of salutogenesis and occu-pational health, it would be very interesting to considerindividual characteristics such as personal resources (senseof coherence, general and specific self-efficacy, coping skills,optimism, self-esteem, hope, or resilience) potentially rel-evant for either the salutogenic or the pathogenic processor both in this general model. However, in this study, wefocus on job demands and job resources, thereby exclud-ing personal resources since personal resources can act asmoderators, mediators, and/or direct predictors of health[15]. Therefore the integration of personal resources wouldoverload the model conceptually and methodologically.

3. Testing the Model: Study Hypotheses

To test this JD-R health model we first identified a commonset of indicators of job demands and resources, whichare potentially suitable for explaining negative and positivehealth outcomes within a broad range of occupations andorganizations (see Section 4). Second, a set of negative andpositive health indicators equally applicable to diverse groupsof employees was selected. Based on these generalizable indi-cators, we (1) tested themodel in a heterogeneous sample and

(2) validated it for different subgroups—that is, in six differentorganizations, among female andmale employees and amongemployees with and without managerial function—as well asacross time.

3.1. Hypothesis 1—Model Testing. We expect to find the dualpathways between (a) job demands and negative healthand (b) job resources and positive health. Further, as itis assumed by the JD-R model, we also expect differentcross-links between these processes: (c) Job demands andjob resources are negatively related. (d) Job resources arenegatively related to negative health. (e) Negative and positivehealth are negatively correlated.

3.2. Hypothesis 2—Invariance Testing. We expect that themodel holds true for different subgroups, that is, for all of thesix organizations, for male as well as for female employees,and for employees with and without managerial function.Further, we expect that it is invariant across time.

4. Method

4.1. Sample. The present three-wave study with a 1-year timeinterval used data collected in the context of a large-scalestress management intervention program (see Acknowledg-ments) implemented between 2008 and 2010 in Switzerlandin six medium and large Swiss organizations in diverse busi-ness sectors. All members of the organizations were invitedto participate. Response rates for the three waves were 70.2%,64.9%, and 62.3% for t1, t2, and t3, respectively. Analysis ofrespondents revealed only minor selective dropout in regardto gender (lower for male) and job resources (lower foremployees with better resources) [25].

Except for the invariance testing across time, analyses inthis study were conducted with the first wave data (t1). Thisbaseline sample consisted of 2,159 employees who worked insix organizations that included three industrial productioncompanies (29.5%; 13.2%; and 18.3%), one food processingcompany (13.9%), one public administration service (15.3%),and one hospital (9.7%). The sample included 1,392 male(64.5%) and 767 female employees (35.5%), with an averageage at t1 of 39.3 years (SD = 11.11). In addition, 42.3% hada higher education degree (college or university). Organiza-tional tenure was 9.0 years (SD = 9.33) with an average of5.1 years (SD = 6.19) in the present job. The heterogeneityof organizations contributes to significant variations in thestudy variables such as organizations largely differed con-cerning, first, their gender ratio (male : female): organization1: 86.6% : 13.4%; organization 2: 45.3% : 54.7%; organization3: 91.4% : 8.6%; organization 4: 42.1% : 57.9%; organization 5:58.5% : 41.5%; organization 6: 17.8% : 82.2%. This ratio clearlydepends on the business sector where the organizationsoperate (see above). Second, they differ concerning the ratioof employees with andwithoutmanagerial position (yes : no):organization 1: 25.1% : 74.9%; organization 2: 24.3% : 75.7%;organization 3: 47.2% : 52.8%; organization 4: 46.5% : 53.5%;organization 5: 25% : 75%; or organization 6: 28.4% : 71.6%.Finally, the organizations differ concerning the core variables


in this study job demands (𝑀 = 2.53 to 𝑀 = 2.79), jobresources (𝑀 = 3.49 to𝑀 = 3.98), negative health (𝑀 = 2.18to𝑀 = 2.45), and positive health (𝑀 = 3.04 to𝑀 = 3.75).

Participants completed a newly developed online sur-vey (see Acknowledgments) that included questions onwork characteristics and health outcomes. Participants wereassured of the anonymity of the data in the introductionto the questionnaire. Participation in the survey was on avoluntary basis, with the questionnaire being administeredduring working time which took about 30 minutes to fillout the basic section of scales (see below). Employees loggedinto the survey system and received an immediate, automatedfeedback of their results in form of a “traffic-light” display(red, orange, and green), detailed percentile rankswith regardto benchmark values, and tips for the highlighted topic.

4.2. Process of Selection of Scales and Items. In order toidentify job demands and resources as well as indicatorsof negative and positive health which are not organization-specific we followed a stepwise approach: The survey com-prised a basic section with 35 validated scales on diversejob demands and resources and a broad range of well-being and health indicators, for which data of all employeeswere available. On the basis of these 35 scales we first ana-lyzed qualitative information about the presence of variousworking conditions in the participating companies fromthe consultants who implemented the stress managementintervention in these companies. In this first step theseconsultants figured as raters (nonparticipative observers) ofthe qualitative process information which was not acces-sible via traditional quantitative survey methods as it wasimplemented in this study.The consultants were asked whichnegative and positive factors were most salient and relevantwithin a specific organization during the study. Based on thisqualitative information, we excluded specific job demandsand job resources restricted to a particular organization (suchas physical and environmental demands or client-relatedissues) and selected those which are expected to be relevantfor all the different organizations. Second,we used descriptivestatistics (means and standard deviations) to quantitativelyidentify the most relevant job demands and job resourcesacross all organizations. Finally, using bivariate correlationsand principle component analyses, we grouped the workcharacteristic indicators and excluded scales which did notload distinctively on one factor (such as effort-reward imbal-ance which is an aggregate rather than a demand in itself).Further, we excluded scales which did not strictly measure acharacteristic of the job (such as work-home interference) orloaded negatively on a positive factor (such as social stressorson the social resources factor). This procedure yielded fivegroups of job demands and job resources: (1) quantitative and(2) qualitative task-related demands, job resources related to(3) manager and (4) peer behavior (support/appreciation),and (5) task-related resources.

We proceeded in a similar way (second step only) with(6) positive and (7) negative health outcomes. Principalcomponent analysis including all scales on health and well-being yielded two factors mainly comprising psychosomatic

disorders and psychosocial well-being, respectively. Again,scales loading negatively on a positive factor (such as negativefeelings on the positive health factor) were excluded toachieve a maximum of distinctiveness (see also PreparatoryAnalyses).

4.3. Measures. The selection procedure described aboveresulted in the following measures used for model testing.

(1) Quantitative Task-Related Job Demands. Two scales canbe subsumed under the factor of quantitative job demands.Time pressure and work interruption were assessed with fouritems each ranging from 1 = very rarely/never to 5 = veryoften/constantly [26]. An example of an item for time pressureis “At work, how often is a rapid pace of work required?”and for work interruption “How often does it occur that youcannot work on something in peace because something elsealways comes in between?”

(2) Qualitative Task-Related Job Demands. Two scales mea-sured qualitative job demands. Qualitative overload isassumed to occur when someone has to fulfill tasks whichare too complicated and too difficult. The three items wereassessed using a 5-point scale from 1 = almost never/notat all true to 5 = almost always/fully true [27]. This is asample item from this scale: “It happens that the work istoo difficult for me.” Uncertainty at work is characterizedby unclear or ambiguous instructions and by the absence ofsufficient information to make decisions [26]. This scale usesfour items: three on a 5-point scale ranging from 1 = veryrarely/never to 5 = very often/constantly and one item on a 5-point scale from 1 = from nobody to 5 = from more than threepersons. An example item is “From how many people do youregularly receive instructions?”

Job Resources Related to (3) Manager and (4) Peer Behavior.Four scales measure manager behavior; another two scalesmeasure peer behavior. Supportive leadership including thedegree to which the supervisor is available, the degree towhich the supervisor’s behavior is respectful and fair, andperformance feedback was measured by five items drawnfrom a questionnaire by Udris and Rimann [27]; for example,“The line manager lets one know how well a job has beendone.”This responsewas scored on a 5-point rating scale from1 = almost never/not at all true to 5 = almost always/fully true.Interpersonal fairness describes the manner of interpersonaltreatment by supervisors during decision-making processes[28]. This scale comprises four items responded on a 5-pointscale from 1 = to a small extent to 5 = to a large extent. Anexample item is “He/she treated you with respect?”Managerand peer support were assessed by one item, each drawnfrom a scale assessing social support received from differentpersons at work [29]. Participants had to assess how muchthey can rely on different kinds of people in difficult situationsatwork, on their direct supervisor and their colleagues amongothers. Both items were scored on a 5-point scale rangingfrom 1 = not true at all to 5 = a lot. Manager and peerappreciation were also assessed by two single items [30]:“Overall, how satisfied are you with the appreciation of your


person shown by your line manager?” and “Overall, howsatisfied are you with the appreciation of your person shownby your colleagues?” These items were rated on a 7-pointgraphical scale using smileys.

(5) Task-Related Job Resources. Two scales captured task-related job resources. Task identity was assessed by a singleitem which was “In my job one can produce something orcarry out an assignment from A to Z?” rated on a 5-pointscale from 1 = almost never/not at all true to 5 = almostalways/fully true [27]. Job control was assessed using a scalewith six items ranging from 1 = very little/not at all to 5 =very much [26]. An example item is “Can you organize yourworkday independently?”

(6) Negative Health. Three scales assessed negative health.Exhaustion was measured using one dimension of theOldenburg Burnout Inventory [31]. The eight items of theexhaustion subscale refer to general feelings of emptinessat work, overtaxing from work, a strong need for rest fromwork, and a state of physical exhaustion through work [7].An example item is “After my work, I usually feel worn outand weary.” Four answer categories from 1 = totally disagreeto 4 = totally agree were used. Insomnia was measured usinga short version of the Insomnia Severity Index [32].The threeitems covered difficulty in falling asleep, difficulty in stayingasleep, and the problem of waking up too early, with oneitem each. Participants answered the items on a 5-point ratingscale from 0 = none to 4 = very. Psychosomatic disorderswere measured by seven items concerning headaches, neckor shoulder pain, back pain, pain in the joints and limbs,loss of appetite, stomach disorders, digestion problems, skinproblems, and eye problems [33]. Responses to these itemswere on a 5-point rating scale from 1 = never to 5 = constantly.

(7) Positive Health. Three scales captured positive health.Work engagement was measured using the nine-item versionof the Utrecht Work Engagement Scale [34], which includesthree subscales of three items each: vigor (e.g., “Atwork, I feel Iam bursting with energy”), dedication (e.g., “My job inspiresme”), and absorption (e.g., “Time flies when I’m working”).Work engagement was scored on a 7-point scale from 0 =never to 6 = always. Job satisfaction was measured by oneitem (How satisfied are you with your work in general?)whichwas rated on a 7-point graphical scale with smileys [35].Affective commitment was assessed by four items drawn fromAllen and Meyer (1990). Commitment was scored on a 7-item rating scale from 1 = not true at all to 7 = almost fullytrue. An example item is “I enjoy discussing my organizationwith people outside it.” Job-related enthusiasm was assessedby three items from Warr’s [36] measurement of well-beingasking whether the participant was optimistic, enthusiastic,and cheerful about the job.They responded on a 5-point scalefrom 1 = never to 5 = all the time.

4.4. Data Analyses. We tested the hypotheses with structuralequation modeling using AMOS 20 software package. Thefit of the model was assessed with the 𝜒2 statistics, thecomparative fit index (CFI), the root mean square error of

approximation (RMSEA), and the standardized root meansquare residual (SRMR). Values of 0.90 and higher areacceptable for the CFI, whereas values of 0.95 or higherare indicators of an excellent fit [37]. Values of up to 0.08for the RMSEA and SRMR represent reasonable errors ofapproximation, whereas values up to 0.05 and 0.01 indicategood and excellent fit, respectively [37, 38].

The invariance of our final model across different orga-nizations, gender, job positions, and time (Hypothesis 2)was tested using multiple group analysis with the AMOS20 software package. In this procedure, two constrainedmodels (one with equality constraints on regression pathsand covariances between latent variables and one additionallywith constraints on factor loadings) were compared to adefault model without cross group constraints. Traditionally,𝜒

2 difference tests are used to assess whether there is asignificant difference between the models [39]. However,because 𝜒2 is highly dependent on sample size, invariancedecisions were based on the differences in CFI, in RMSEA,and in SRMRwith aΔCFI≤ 0.01 supplemented by aΔRMSEA≤ 0.015 and a ΔSRMR ≤ 0.030 indicating invariance [40, 41].

5. Results

5.1. Preparatory Analyses. Because our study variables weremeasured via single-source self-reporting, we examined thedegree to which common method variance could threat ouranalyses. Thus, two tests were conducted: first, a Harmansingle factor test [42] was performed.The examination of theunrotated factor solution indicated the presence of at leastsix factors; that is, no single factor emerged whereby the firstfactor explained 30.07%, indicating that common methodeffects were not a likely contaminant of the results observed inour study. To confirm these results, we additionally controlledfor the effects of an unmeasured latent factor in our model[43]. The results indicated that whereas the method factordid improve the model fit, it accounted for a very smallproportion (nearly 0%) of the total variance. Both testssuggest that common method variance is not a pervasiveproblem in this study.

To test the assumed two-factor structure of health (nega-tive and positive health), we split our sample into two randomsubsamples (𝑛’s = 925 and 926) and conducted with one halfof the sample an exploratory factor analysis and with the sec-ond half a confirmatory factor analysis. For the exploratorypart, we conducted a principal component analysis on theseven health scales with oblimin rotation. Two componentsobtained eigenvalues over Kaiser’s criterion of 1, which incombination explained 62.37% of the variance. Besides, theanalyses of the scree plot yielded two factors as well. Theresulting two factors were as expected (1) sleep disorders,exhaustion, and psychosomatic disorders indicating negativehealth and (2) job satisfaction, affective commitment, work-related enthusiasm, andwork engagement indicating positivehealth. Next, we conducted a confirmatory factor analysiswith the second half of the sample.The two-factor solutionweextracted from the exploratory factor analysis was confirmedand showed an acceptable to good model fit (𝜒2(11) = 40.01,


𝑝 < 0.001, CFI = 0.98, and RMSEA = 0.05, SRMR = 0.03)when we allowed the residuals to be correlated between sleepdisorders and psychosomatic disorders as well as betweenaffective commitment and enthusiasm.The inclusion of thesecorrelations seems acceptable because the respective variablesare drawn from the same scales (see above). Factor loadingsranged between 0.40 (psychosomatic disorders) and 0.89(exhaustion). Moreover, the covariation between the latentfactors representing positive and negative health was −0.48.This solution fitted the data remarkably better than the one-factor solution (𝜒2(12) = 220.57, 𝑝 < 0.001, CFI = 0.86, andRMSEA = 0.14, SRMR = 0.08)

Bivariate correlations and internal consistencies (Cron-bach’s alpha, where suitable, that is, with exception of thesingle items) of all study variables are shown in Table 1. Notethat all scales were sufficiently reliable.

5.2. Model Testing (Hypothesis 1). First, in order to testHypothesis 1, the two processes were investigated inde-pendently—that is, without cross-links between job demandsand job resources as well as between negative and positivehealth (Model 1). As the results indicated (see Table 2), themodel did not fit the data very well since relevant modelparameters (𝜒2, df, CFI, and SRMR) were not acceptable.Second, in order to investigate cross-links, job demands andjob resources, as postulated by the JD-R model, were allowedto correlate (Model 2). This model showed a superior fit(Δ𝜒2(4) = 695.29; 𝑝 < 0.001).

The parameter estimates for the final model are shown inFigure 1. All relations are significant. As expected, the pathsfrom job demands to negative health (𝛽 = 0.41; 𝑝 < 0.001)and from job resources to positive health (𝛽 = 0.89; 𝑝 <0.001) were positive and significant even though the pathcoefficient from job demands to negative health was lowerthan the path coefficient from job resources to positive health.Furthermore, the cross-links between job demands and jobresources (𝛽 = −0.63; 𝑝 < 0.001), between job resources andnegative health (𝛽 = −0.36; 𝑝 < 0.001), and, finally, betweennegative and positive health (𝛽 = −0.37; 𝑝 < 0.001) werenegative and significant as expected. Thus Hypothesis 1 wassupported. Moreover, in this model, a total variance of 42.4%(𝑟2 = 0.424) in negative health and of 60.2% (𝑟2 = 0.602) inpositive health is explained by job demands and job resources.

5.3. Invariance Testing (Hypothesis 2). To cross validate thefindings, several multiple group analyses were conductedin order to test the assumed invariance of the final model(Hypothesis 2). For each of these analyses, the regressionpaths and covariances between the latent variables in ourmodel were constrained to be equal across groups. This con-strained model (Model 2) was compared with the free model(default model), in which parameter estimates were allowedto vary freely across groups. Next, in addition to constrainingthe regression paths between the latent variables, the factorloadings were constrained to be equal across groups (Model3) and this model was compared with the free model.

Across organizations, results of invariance testing showedthat both regression paths between the latent variables and

the factor loadings were invariant as CFI, RMSEA, and SRMRdifference tests (ΔCFI = 0.002 and 0.009; ΔRMSEA = 0.001and 0.002; ΔSRMR = 0.001 and 0.006) showed (see Table 3).

Also across gender as well as across job positions (man-agers versus employees) regression paths and covariancesbetween latent variables and factor loadings turned out to beinvariant (gender: ΔCFI = 0.001 and 0.003; ΔRMSEA = 0.004and 0.003; ΔSRMR = 0.001 and 0.000; job position: ΔCFI= 0.001 and 0.001; see Table 4; ΔRMSEA = 0.002 and 0.003;ΔSRMR = 0.001 and 0.004; see Tables 4 and 5).

Finally, results of multiple group testing as displayed inTable 6 indicate invariance across the three measurementpoints (ΔCFI = 0.000 and 0.000; ΔRMSEA = 0.001 and 0.002;ΔSRMR= 0.004 and 0.002).Thus, results support Hypothesis2.

Taken together these findings lend support to the pro-posed expanded JD-R health model explaining negativeand positive health. The hypothesized model was confirmed(Hypothesis 1) and could be cross validated across six differ-ent organizations, across gender, across job positions, and,finally, across time (Hypothesis 2).

6. Discussion

The aim of the present study was to develop and test anexpanded JD-R health model with a comprehensive set ofjob characteristics generalizable to diverse organizations.Thereby we integrate the concept of salutogenesis [16] and ageneric health development framework applying the concep-tion of positive and negative health [17].

The resulting model links job demands and resourcesthrough two broad paths—a pathogenic as well as a saluto-genic path—to both positive and negative biopsychosocialhealth outcomes. The study could build on a broad rangeof job demands and resources as well as health outcomescollectedwithin a large-scale stressmanagement interventionstudy. This allowed to apply qualitative and quantitativemethods to empirically select and group global job demandsand resources which were relevant for all employees: quanti-tative and qualitative task-related job demands, job resourcesrelated to supportive and appreciative manager and peerbehavior, respectively, and task-related resources.

By testing our hypotheses with structural equation mod-eling, we found evidence for the validity of an expandedJD-R health model predicated on a broad and heteroge-neous sample. Invariance testing indeed showed that themodel seems to be generalizable to diverse organizations andoccupations and, moreover, time invariant. Namely, multiplegroup analyses indicated invariance across six organizationsof various business sectors, across gender and job positions,and, finally, across three times of measurement.

We were able to support the two different pathogenicand salutogenic processes in analogy but expansion of thehealth impairment and motivational processes of the JD-R model [6, 10]. Our expanded model also supported thecross-links also predicted by the JD-R model [15]. First, wefound a strong negative relationship between job demandsand job resources, which is in line with other studies [12, 44].


Table1:Internalconsistencies

(Cronb

ach’s

alph

asin

italicso

nthed

iagonal)andcorrela

tions

amon

gthev

ariables.

12

34

56

78

910

1112

1314

1516

1718

191

Workinterrup

tion

0.805

2Timep

ressure

0.54∗∗

0.829

3Qualitativeo

verlo

ad0.18∗∗

0.25∗∗

0.788

4Uncertaintyatwork

0.34∗∗

0.37∗∗

0.35∗∗

0.733

5Supp

ortiv

eleadership−0.13∗∗−0.17∗∗−0.15∗∗−0.34∗∗

0.816

6Interpersonaljustice−0.05∗−0.10∗∗−0.16∗∗−0.30∗∗

0.57∗∗

0.813

7Manager

supp

ort−0.09∗∗−0.11∗∗−0.12∗∗−0.29∗∗

0.59∗∗

0.45∗∗

—8

Manager

appreciatio

n−0.10∗∗−0.15∗∗−0.16∗∗−0.33∗∗

0.69∗∗

0.55∗∗

0.56∗∗

—9

Peer

supp

ort

−0.07∗∗−0.05∗−0.09∗∗−0.09∗∗

0.16∗∗

0.18∗∗

0.35∗∗

0.12∗∗

—10

Peer

appreciatio

n−0.09∗∗−0.11∗∗−0.15∗∗−0.16∗∗

0.28∗∗

0.22∗∗

0.15∗∗

0.39∗∗

0.37∗∗

—11

Task

identity

−0.09∗∗−0.06∗−0.10∗∗−0.15∗∗

0.12∗∗

0.09∗∗

0.09∗∗

0.13∗∗

0.12∗∗

0.14∗∗

—12

Jobcontrol

0.11∗∗−0.03−0.14∗∗−0.15∗∗

0.20∗∗

0.25∗∗

0.13∗∗

0.22∗∗

0.04

0.14∗∗

0.23∗∗

0.859

13Insomnia

0.13∗∗

0.19∗∗

0.24∗∗

0.20∗∗−0.20∗∗−0.22∗∗−0.16∗∗−0.21∗∗−0.17∗∗−0.19∗∗−0.11∗∗−0.18∗∗

0.709

14Ex

haustio

n0.24∗∗

0.40∗∗

0.35∗∗

0.31∗∗−0.32∗∗−0.26∗∗−0.22∗∗−0.31∗∗−0.15∗∗−0.24∗∗−0.12∗∗−0.24∗∗

0.51∗∗

0.820

15Psycho

somaticdisorders

0.15∗∗

0.19∗∗

0.21∗∗

0.18∗∗−0.17∗∗−0.21∗∗−0.14∗∗−0.18∗∗−0.13∗∗−0.18∗∗−0.08∗∗−0.17∗∗

0.43∗∗

0.44∗∗

0.740

16Jobsatisfaction−0.16∗∗−0.18∗∗−0.20∗∗−0.30∗∗

0.47∗∗

0.38∗∗

0.33∗∗

0.53∗∗

0.14∗∗

0.36∗∗

0.14∗∗

0.24∗∗−0.27∗∗−0.43∗∗−0.24∗∗

—17

Affectivec

ommitm

ent−0.01

0.02−0.16∗∗−0.19∗∗

0.27∗∗

0.20∗∗

0.17∗∗

0.26∗∗

0.04

0.18∗∗

0.11∗∗

0.19∗∗−0.11∗∗−0.24∗∗−0.12∗∗

0.45∗∗

0.817

18Job-related

enthusiasm−0.10∗∗−0.13∗∗−0.19∗∗−0.22∗∗

0.36∗∗

0.25∗∗

0.24∗∗

0.39∗∗

0.14∗∗

0.30∗∗

0.09∗∗

0.18∗∗−0.30∗∗−0.48∗∗−0.25∗∗

0.56∗∗

0.39∗∗

0.836

19Workengagement−0.04

0.03−0.17∗∗−0.17∗∗

0.27∗∗

0.18∗∗

0.15∗∗

0.29∗∗

0.09∗∗

0.25∗∗

0.12∗∗

0.22∗∗−0.27∗∗−0.39∗∗

0.22∗∗

0.487∗∗

0.53∗∗

0.61∗∗

0.942

Note.𝑁

=1,8

51.C

ronb

ach’s

alph

asappear

onthed

iagonalswhere

approp

riate.∗𝑝≤0.05,∗∗𝑝≤0.01(tw

o-tailed).


Table 2: Fit statistics for alternative models (model-testing).

Model df 𝜒

2 (𝑁 = 1,850) 𝜒

2/df GFI CFI RMSEA SRMR Δ𝜒

2

1 (without cross-links) 142 1784.96 12.57 0.91 0.86 0.08 0.13 —2 (final model) 138 1089.67 7.90 0.94 0.92 0.06 0.05 695.29∗∗∗

Note. GFI = goodness-of-fit index; CFI = comparative fit index; and RMSEA = root mean square error of approximation.∗∗∗𝑝 < 0.001.

Table 3: Fit statistics for multigroup analyses and invariance tests across organizations (𝑁Org.1 = 543,𝑁Org.2 = 247,𝑁Org.3 = 247,𝑁Org.4 = 284,𝑁Org.5 = 337, and𝑁Org.6 = 169).

Model df 𝜒2 𝜒2/df GFI CFI RMSEA SRMR Δ𝜒

2Δdf ΔCFI ΔRMSEA ΔSRMR

OrganizationsModel 1 (default model) 342 1025.48 3.00 0.92 0.90 0.03 0.06 — — — — —Model 2 (regression paths and covariancesbetween latent variables constrained to beequal across groups)

372 1072.89 2.88 0.92 0.90 0.03 0.06 47.41∗ 30 0.002 0.001 0.001

Model 3 (Model 2 and factor loadingsconstrained to be equal across groups) 417 1163.2 2.79 0.91 0.89 0.03 0.07 137.72∗∗∗ 75 0.009 0.002 0.006

Note. GFI = goodness-of-fit index; CFI = comparative fit index; RMSEA = root mean square error of approximation; and SRMR = standardized root meansquare residual.∗𝑝 < 0.05, ∗∗𝑝 < 0.01, and ∗∗∗𝑝 < 0.001.

Table 4: Fit statistics for multigroup analyses and invariance tests across gender (𝑁male = 1,200,𝑁female = 651).

Model df 𝜒2 𝜒2/df GFI CFI RMSEA SRMRΔ𝜒



120 799.47 6.66 0.94 0.90 0.05 0.06 9.41(𝑝 = 0.152) 6 0.001 0.004 0.001

Model 3 (Model 2 and factor loadingsconstrained to be equal across groups) 129 822.26 6.37 0.94 0.90 0.05 0.05 32.20∗∗ 15 0.003 0.003 0.000


Table 5: Fit statistics for multigroup analyses and invariance tests across job position (𝑁managers = 600,𝑁employees = 1,250).

Model df 𝜒2 𝜒2/df GFI CFI RMSEA SRMRΔ𝜒



120 774.79 6.457 0.94 0.90 0.05 0.06 4.38(𝑝 = 0.625) 6 0.001 0.002 0.000

Model 3 (Model 2 and factor loadingsconstrained to be equal across groups) 129 792.065 6.14 0.94 0.90 0.05 0.06 21.655

(𝑝 = 0.117) 15 0.001 0.003 0.004


Second, there is a strong and significant negative cross-linkbetween job resources and negative health, and, third, thereis a negative, even thoughnot particularly strong, relationshipbetween negative and positive health. These results alsocorrespond with findings from other studies using the JD-Rmodel as a theoretical framework [8, 45].

Furthermore, results showed that the regression weightof the path from job demands to negative health is weakerthan that of the path from job resources to positive health.This differencemight be explained by the fact that the positivehealth indicators are closer to the working situation whereasnegative health indicators such as insomnia or psychosomatic


Table 6: Fit statistics for multigroup analyses and invariance tests across time (𝑁𝑡1= 1,858,𝑁

𝑡2= 1,913, and𝑁

𝑡3= 1,754).

Model df 𝜒

2𝜒

2/df GFI CFI RMSEA SRMR Δ𝜒2 Δdf ΔCFI ΔRMSEA ΔSRMROrganizations

Model 1 (default model) 414 3147.67 7.60 0.94 0.93 0.04 0.05 — — — — —Model 2 (regression paths and covariancesbetween latent variables constrained to beequal across groups)

432 3178.18 7.36 0.94 0.93 0.03 0.06 30.51∗ 16 0.000 0.001 0.004

Model 3 (Model 2 and factor loadingsconstrained to be equal across groups) 456 3220.44 7.06 0.94 0.92 0.03 0.05 72.77∗∗∗ 42 0.000 0.002 0.002


Work interruption

Time pressure

Qualitative overload

Manager support

Interpersonaljustice

Supportive leadership

Uncertainty at work

Manager appreciation

Peer support

Peer appreciation

Job control

Task identity

Job demands

Job resources

Manager behavior

Quantitative demands

Peer behavior

Task-related resources

Negative health

Positive health

Insomnia

Exhaustion

Psychosomatic disorders

Job satisfaction

Affective commitment

Job-related enthusiasm

0.66

0.64

0.84

0.52

0.66

0.89

0.41

0.18

0.76

0.58

0.54

0.57

0.90

0.49

0.72

0.63

0.80

e5

e6

0.85

0.66

0.67

0.83

0.47

0.85

0.39

0.60

Work engagement

0.74

−0.63

−0.36

−0.37

Figure 1: Standardized path coefficients of the final model in the whole sample (𝑁 = 1,851).

disorders are more general and influenced by many factorsapart from work characteristics (see below). As regards ourfinal goal, namely, to develop amodel that includedmeasuresof both positive and negative health, the available healthindicators were limited in various ways. With the exemptionof work-related exhaustion, negative health was assessed ina general way by looking at insomnia and psychosomaticdisorders, which are only partially influenced by the worksituation. On the other hand, positive health was mainly

assessed by work-related measures, which plausibly showedstronger relationships with job characteristics (instead ofglobal indicators such as life satisfaction). In the future,negative and positive health should be assessed equally work-specific or generic to produce more comparable results forthe pathogenic and the salutogenic process of the JD-Rhealth model. Specifically, either a narrower study shouldbe undertaken on the side of negative health (i.e., morework-related) or a broader one on the side of positive health


(i.e., less work-related). The latter approach could have theadvantage of showing the general significance and publichealth relevance of job characteristics and their beneficial andadverse combination, respectively.

Moreover, the measurement of work characteristics andhealth was based on self-reported data, which increases thepossibility that the relationships between job demands andjob resources as well as between negative and positive healthcould be due to common method variance. Despite the lackof method variance found, there still remains the possibilitythat this effect may differ across models. Therefore, futureresearch efforts need to consider using multiple methods andmeasures to eliminate the effect of this potential bias.

Even though we tested time invariance using three timesof measurement, this study is cross-sectional. Longitudinalstudies with specific research questions and explicit hypothe-ses concerning change over time would complement theinsights of this study [25].

Another limitation concerns the process how we iden-tified the indicators of job demands, job resources, andnegative and positive health. This process was explorativeand its systematic should be improved, for example, via pro-fessional judges or raters instead of consultants.

A third limitation is the lack of a positive impressionscale included in the questionnaire in order to exclude thatparticipants answered in a way that will be viewed positivelyby others which might have biased the results [46].

Fourth, emerging issues (such as stress caused by theeconomic crisis Europe is still facing) were not captured. Itwould be very interesting to include such highly relevantconcepts in a follow-up study.

Moreover, in this study, we only focused on job demandsand job resources as predictors of health even though itwould be interesting to investigate the role of individualcharacteristics such as personal resources in our model [47].More studies are needed to reveal the influence of personalresources such as self-efficacy or resilience on the interplaybetween job characteristics and health.

Finally, in order not only to control for the different orga-nizations but also to specifically investigate the influence theorganizational context might have on the variables includedin our model as well as the relationships between them,further studies applying amultilevel approachwill be needed.

7. Conclusions and Contributions

The present study made a step towards an expanded JD-R health model tested with a common set of indicators ofjob demands and job resources predicting a broad conceptof health, including physical, mental, and social health. Itcontributes to the current research on job characteristics andemployees’ health by expanding the JD-R model towardsa pathogenic and salutogenic path with both negative andpositive health outcomes. Building on the JD-Rmodel and ona broad data set with three times of measurement, the presentstudy combined both a theory-driven, deductive approachand a data-driven, inductive approach for the selection ofcommon indicators. We regard our generalized JD-R health

model as a contribution to the integration of indicator-focused, evidence-based risk (and resource) assessmentswithin comprehensive frameworks, as was called for byClausen et al. [48]. We aimed to make the JD-R modelcomprehensible and useful to researchers with biomedicaltraining by showing that its logic can be expanded from theoriginal health impairment as well as motivational processesto simultaneously study pathogenic and salutogenic healthdevelopment processes at work. At the same time, thestudy aimed to show for the first time that this model—operationalized with a manageable number of validatedindicators—is stable over time and applicable to diverseeconomic sectors and professional groups. Therefore, thismodel is highly useful to show to a broader communitywho is concerned with public health issues how health-related good psychosocial working conditions are. Moreover,the findings of this study not only showed the relevanceof this topic but also can indicate which issues should beaddressed when implementing successful population-basedpublic health interventions in the working population.




Rebecca Brauchli and Gregor J. Jenny contributed equiva-lently to the paper.

Acknowledgments

The project “SWiNG” (Stress Management—Effectivenessand Benefit of Workplace Health Promotion) was launchedand financed by Health Promotion Switzerland and theSwiss Assurance Association. The selection of evidence-based scales for the online survey (“S-Tool”) was con-ducted by the University of Bern (Chair Professor N. Sem-mer) in collaboration with stress management consultantsand Health Promotion Switzerland (for more details, seehttps://www.s-tool.ch/).

References

[1] D. Gagliardi, A. Marinaccio, A. Valenti, and S. Iavicoli, “Occu-pational safety and health in Europe: lessons from the past,challenges and opportunities for the future,” Industrial Health,vol. 50, no. 1, pp. 7–11, 2012.

[2] A. B. Bakker, M. van Veldhoven, and D. Xanthopoulou,“Beyond the demand-control model: thriving on high jobdemands and resources,” Journal of Personnel Psychology, vol.9, no. 1, pp. 3–16, 2010.

[3] J. A. Edwards, S. Webster, D. van Laar, and S. Easton, “Psycho-metric analysis of the UK health and safety executive’s man-agement standards work-related stress indicator tool,” Work &Stress, vol. 22, no. 2, pp. 96–107, 2008.

[4] R. A. Karasek, “Job demands, job decision latitude, and mentalstrain: Implications for job redesign,” Administrative ScienceQuarterly, vol. 24, no. 2, pp. 285–308, 1979.


[5] J. Siegrist, “Adverse health effects of high-effort/low-rewardconditions,” Journal of Occupational Health Psychology, vol. 1,no. 1, pp. 27–41, 1996.

[6] A. B. Bakker and E. Demerouti, “The Job Demands-Resourcesmodel: state of the art,” Journal of Managerial Psychology, vol.22, no. 3, pp. 309–328, 2007.

[7] E. Demerouti, A. B. Bakker, F. Nachreiner, and W. B. Schaufeli,“The job demands-resources model of burnout,” Journal ofApplied Psychology, vol. 86, no. 3, pp. 499–512, 2001.

[8] W. B. Schaufeli and A. B. Bakker, “Job demands, job resources,and their relationship with burnout and engagement: a multi-sample study,” Journal of Organizational Behavior, vol. 25, no. 3,pp. 293–315, 2004.

[9] A. B. Bakker, E. Demerouti, andW. S. Schaufeli, “Dual processesat work in a call centre: an application of the job demands-resourcesmodel,” European Journal ofWork andOrganizationalPsychology, vol. 12, no. 4, pp. 393–417, 2003.

[10] A. B. Barker and S. A. E. Geurts, “Toward a dual-process modelof work-home interference,”Work and Occupations, vol. 31, no.3, pp. 345–366, 2004.

[11] Q. Hu, W. B. Schaufeli, and T. W. Taris, “The job demands-resources model: an analysis of additive and joint effects ofdemands and resources,” Journal of Vocational Behavior, vol. 79,no. 1, pp. 181–190, 2011.

[12] J. J. Hakanen, A. B. Bakker, and W. B. Schaufeli, “Burnoutand work engagement among teachers,” Journal of SchoolPsychology, vol. 43, no. 6, pp. 495–513, 2006.

[13] C. Korunka, B. Kubicek, W. B. Schaufeli, and P. Hoonakker,“Work engagement and burnout: testing the robustness of thejob demands-resources model,” Journal of Positive Psychology,vol. 4, no. 3, pp. 243–255, 2009.

[14] S. Llorens, W. Schaufeli, A. Bakker, and M. Salanova, “Does apositive gain spiral of resources, efficacy beliefs and engagementexist?” Computers in Human Behavior, vol. 23, no. 1, pp. 825–841, 2007.

[15] W. B. Schaufeli and T. W. Taris, “A critical review of the jobdemands-resources model: implications for improving workand health,” in Bridging Occupational, Organizational and Pub-lic Health, pp. 43–68, Springer, Amsterdam, The Netherlands,2014.

[16] A. Antonovsky, Unraveling the Mystery of Health, Jossey-Bass,San Francisco, Calif, USA, 1987.

[17] G. Bauer, J. K. Davies, and J. Pelikan, “The EUHPID HealthDevelopment Model for the classification of public healthindicators,” Health Promotion International, vol. 21, no. 2, pp.153–159, 2006.

[18] A. Antonovsky, Health, Stress, and Coping, Jossey-Bass, SanFrancisco, Calif, USA, 1979.

[19] J. J. Hakanen, R. Perhoniemi, and S. Toppinen-Tanner, “Positivegain spirals at work: from job resources to work engagement,personal initiative and work-unit innovativeness,” Journal ofVocational Behavior, vol. 73, no. 1, pp. 78–91, 2008.

[20] E.Ouweneel, P.M. Le Blanc,W. B. Schaufeli, andC. I. vanWijhe,“Good morning, good day: a diary study on positive emotions,hope, and work engagement,” Human Relations, vol. 65, no. 9,pp. 1129–1154, 2012.

[21] World Health Organization, Constitution of the World HealthOrganization, World Health Organization, Geneva, Switzer-land, 1946.

[22] M. E. P. Seligman, “Positive health,” Applied Psychology, vol. 57,no. 1, pp. 3–18, 2008.

[23] M. Jahoda, Current Concepts of Positive Mental Health, BasicBooks, New York, NY, USA, 1958.

[24] C. L. M. Keyes, “Promoting and protecting mental health asflourishing: a complementary strategy for improving nationalmental health,”TheAmerican Psychologist, vol. 62, no. 2, pp. 95–108, 2007.

[25] R. Brauchli, W. B. Schaufeli, G. J. Jenny, D. Fullemann, and G. F.Bauer, “Disentangling stability and change in job resources, jobdemands, and employee well-being—a three-wave study on thejob-demands resources model,” Journal of Vocational Behavior,vol. 83, no. 2, pp. 117–129, 2013.

[26] N. K. Semmer, D. Zapf, and H. Dunckel, Assessing Stress atWork: A Framework and an Instrument, Office for OfficialPublications of the European Communities, Luxembourg, Lux-embourg, 1995.

[27] I. Udris and M. Rimann, “SAA and SALSA: two questionnaireson the subjective analysis of work,” inHandbook of PsychologicalWork Analyisis Techniques. A Practice Oriented Overview, pp.397–419, vdf Hochschulverlag, 1999.

[28] J. A. Colquitt, “On the dimensionality of organizational justice:a construct validation of a measure,” Journal of Applied Psychol-ogy, vol. 86, no. 3, pp. 386–400, 2001.

[29] M. Frese, “Gutekriterien der Operationalisierung von sozialerUnterstutzung am Arbeitsplatz,” Zeitschrift fur Arbeitswis-senschaft, vol. 43, no. 1, pp. 112–121, 1989.

[30] N. Jacobshagen, F. T. Amstad, N. K. Semmer, and M. Kuster,“Work-family Balance in top management: conflict betweenwork and family life as a mediator of the relationship betweenstressors and being,” Zeitschrift fur Arbeits- und Organisation-spsychologie, vol. 49, no. 4, pp. 208–219, 2005.

[31] E. Demerouti, Burnout: A Consequence of Specific WorkingConditions Among Human Service and Production Tasks, PeterLang, Frankfurt, Germany, 1999.

[32] C. H. Bastien, A. Vallieres, and C. M. Morin, “Validation of theinsomnia severity index as an outcome measure for insomniaresearch,” Sleep Medicine, vol. 2, no. 4, pp. 297–307, 2001.

[33] G. Bauer and M. Schmid, KMU-Vital: Ein Webbasiertes Pro-gramm zur Betrieblichen Gesundheitsforderung, Vdf Hochschul-verlag, Zurich, Switzerland, 2008.

[34] W. B. Schaufeli, A. B. Bakker, and M. Salanova, “The measure-ment of work engagement with a short questionnaire: a cross-national study,” Educational and Psychological Measurement,vol. 66, no. 4, pp. 701–716, 2006.

[35] N. K. Semmer, J. Baillod, and L. Ruch, “Das Modell ver-schiedener Formen von Arbeitszufriedenheit: Nach 15 Jahrenkein Grund zur Resignation,” in Bericht uber den 37. Kongressder DGfP in Kiel 1990, D. Frey, Ed., pp. 648–649, Hogrefe,Gottingen, Germany, 1990.

[36] P. Warr, “The measurement of well-being and other aspects ofmental health,” Journal of Occupational Psychology, vol. 63, no.3, pp. 193–210, 1990.

[37] L.-T. Hu and P. M. Bentler, “Cutoff criteria for fit indexes incovariance structure analysis: conventional criteria versus newalternatives,” Structural Equation Modeling: A MultidisciplinaryJournal, vol. 6, no. 1, pp. 1–55, 1999.

[38] R. C. MacCallum, M. W. Browne, and H. M. Sugawara,“Power analysis and determination of sample size for covariancestructure modeling,” Psychological Methods, vol. 1, no. 2, pp.130–149, 1996.

[39] N. Schmitt and G. Kuljanin, “Measurement invariance: reviewof practice and implications,” Human Resource ManagementReview, vol. 18, no. 4, pp. 210–222, 2008.


[40] F. F. Chen, “Sensitivity of goodness of fit indexes to lackof measurement invariance,” Structural Equation Modeling. AMultidisciplinary Journal, vol. 14, no. 3, pp. 464–504, 2007.

[41] G. W. Cheung and R. B. Rensvold, “Evaluating goodness-of-fit indexes for testing measurement invariance,” StructuralEquation Modeling, vol. 9, no. 2, pp. 233–255, 2002.

[42] P. M. Podsakoff, S. B. MacKenzie, J.-Y. Lee, and N. P. Podsakoff,“Common method biases in behavioral research: a criticalreview of the literature and recommended remedies,” Journalof Applied Psychology, vol. 88, no. 5, pp. 879–903, 2003.

[43] L. J. Williams, J. A. Cote, and M. R. Buckley, “Lack of methodvariance in self-reported affect and perceptions at work: realityor artifact?” Journal of Applied Psychology, vol. 74, no. 3, pp. 462–468, 1989.

[44] A. B. Bakker, E. Demerouti, and W. Verbeke, “Using thejob demands-resources model to predict burnout and perfor-mance,”HumanResourceManagement, vol. 43, no. 1, pp. 83–104,2004.

[45] W. B. Schaufeli, A. B. Bakker, and W. van Rhenen, “Howchanges in job demands and resources predict burnout, workengagement, and sickness absenteeism,” Journal of Organiza-tional Behavior, vol. 30, no. 7, pp. 893–917, 2009.

[46] N. Mucci, G. Giorgi, V. Cupelli et al., “Work-related stressassessment in a population of Italian workers. The StressQuestionnaire,” Science of the Total Environment, vol. 502, pp.673–679, 2015.

[47] B. van derHeijden, K. VonDam,D. Xanthopoulou, andA.H. deLange, “Individual characteristics and work-related outcomes,”in An Introduction to Contemporary Work Psychology, M. C. W.Peeters, J. de Jonge, and T. W. Taris, Eds., pp. 243–266, JohnWiley & Sons, Chichester, UK, 2014.

[48] T. Clausen, K. Nielsen, I. G. Carneiro, and V. Borg, “Jobdemands, job resources and long-term sickness absence in theDanish eldercare services: a prospective analysis of register-based outcomes,” Journal of Advanced Nursing, vol. 68, no. 1, pp.127–136, 2012.

Research ArticleThe Context, Process, and Outcome Evaluation Model forOrganisational Health Interventions

Annemarie Fridrich, Gregor J. Jenny, and Georg F. Bauer

Division of Public & Organizational Health, Epidemiology, Biostatistics and Prevention Institute, University of Zurich,Hirschengraben 84, 8001 Zurich, Switzerland

Correspondence should be addressed to Georg F. Bauer; [email protected]

Received 18 December 2014; Revised 9 February 2015; Accepted 23 March 2015


Copyright © 2015 Annemarie Fridrich et al. This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.

To facilitate evaluation of complex, organisational health interventions (OHIs), this paper aims at developing a context, process, andoutcome (CPO) evaluationmodel. It builds on previousmodel developments in the field and advances them by clearly defining andrelating generic evaluation categories for OHIs. Context is defined as the underlying frame that influences and is influenced by anOHI. It is further differentiated into the omnibus and discrete contexts. Process is differentiated into the implementation process, asthe time-limited enactment of the original intervention plan, and the change process of individual and collective dynamics triggeredby the implementation process. These processes lead to proximate, intermediate, and distal outcomes, as all results of the changeprocess that are meaningful for various stakeholders. Research questions that might guide the evaluation of an OHI according tothe CPO categories and a list of concrete themes/indicators and methods/sources applied within the evaluation of an OHI projectat a hospital in Switzerland illustrate the model’s applicability in structuring evaluations of complex OHIs. In conclusion, the modelsupplies a common language and a sharedmentalmodel for improving communication between researchers and companymembersand will improve the comparability and aggregation of evaluation study results.

1. Introduction

Recent years have seen a rise of work-related health inter-ventions targeting the entire organisation as complex socialsystems, thus referred to as organisational health inter-ventions (OHIs) [1]. This paper refers to comprehensiveOHIs, which are usually comprised of a mix of individual-directed interventions targeting employee stressmanagementcapacities and leadership behaviour combined with work-directed interventions, such as collective workshops targetingworking conditions and social relations. Such interventionsare portrayed as intricate change processes in complex socialsystems where implementation and outcomes are never apriori predictable [2, 3].

Currently available meta-analyses and reviews of eval-uation studies reveal mixed results for the effectiveness ofcomprehensive OHIs [4, 5]. Thus, systematic evaluations ofOHIs are considered to be ultimately needed, to understandintended and unintended and desirable and undesirablechanges and moreover to pass on process and outcome

knowledge to other researchers and practitioners in the field[2, 3, 6]. However, since OHIs vary in content, composi-tion, context, targeted audience, and the desired outcomesof change, it is difficult to develop generally applicableevaluation principles [7]. A variety of different evaluationapproaches for OHIs have evolved, which either focus onspecific aspects of interventions to be evaluated or presentgeneral rationales to be further specified and operational-ized in regard to a specific project [2, 7, 8]. Additionally,these approaches make use of a diverse terminology, all ofwhich provide a challenge for practitioners and interventionresearchers to integrate findings across studies and to selectan appropriate evaluation approach for their own interven-tion project.

So far, only a few researchers have published genericmodels for the evaluation of complex OHIs. Some modelsfocus on one evaluation category such as process evaluation[7], others are limited to special types of interventions suchas preventive occupational health and safety interventions[9], and still others are rather applicable to single component


http://dx.doi.org/10.1155/2015/414832


interventions than to multicomponent interventions [10].Many of them display context as the underlying frame of anintervention but limit it to its hindering or facilitating role[7, 8, 10]. Finally, most of the existingmodels and frameworks[8, 9] classify effectiveness research/effect evaluation as anautonomous phase after the intervention instead of consider-ing it as a process of continuous observation and assessmentthroughout the intervention.

All these models and frameworks demonstrate the con-siderable effort that has been made to standardise andimprove the evaluation of organisational health interven-tions. However, to make the results revealed by evaluationsof comprehensive interventions more comparable, a modelwith clearly defined evaluation categories is needed. Such amodel has to be applicable to different intervention types, likesingle component and multicomponent interventions, inter-ventions that address psychosocial or physiological aspects,and those that focus on the individual, the organisation, orboth. Thus, it must be general enough to cover all thesedifferent intervention types but at the same time concreteenough to distinguish between the context, process, andoutcome aspects at different organisational levels in differentinterventions phases.

Furthermore, the model has to consider context not onlyas a static and confounding factor that hinders or facilitatesthe implementation process, but also as a transformableand essential part of the intervention. The model shouldalso consider the outcome evaluation as a continuous pro-cess rather than as a particular, time-limited interventionphase. Moreover, it should be accompanied by examples andresearch questions that help intervention researchers conductcontext, process, and outcome evaluations.

All of the aforementioned models and frameworks onlypartially meet these needs. Thus, this paper aims at devel-oping a model defining and relating generic evaluationcategories for OHIs. First, we develop, define, and relatethe generic categories and subcategories of the model. Sec-ond, concepts to specify the main categories for developingmeasurable indicators are compiled, considering differentintervention approaches and principles from the literature.Third, research questions that might guide the evaluationof an OHI according to the CPO categories and a listof concrete themes/indicators and methods/sources appliedwithin the evaluation of an OHI at a hospital in Switzerlandillustrate the model’s applicability in structuring evaluationsof comprehensive OHIs.

2. Overview of the Model

The present model differentiates between the three categoriescontext, process, and outcome evaluation of organisationalhealth interventions and is thus labelled CPO evaluationmodel (see Figure 1). Context is seen as the underlyingframe within an organisational health intervention is imple-mented, change occurs, and outcomes emerge. We furtherdistinguish between the omnibus context which refers tothe general intervention and implementation setting and thediscrete context which refers to specific situational variables.In regard to the category of process, two subcategories

are differentiated: the implementation process as the time-limited enactment of all steps and elements of the originalintervention plan and the triggered change process as all theintended and unintended and observable and nonobservablemechanisms of alteration in the intervention context. Thisleads to outcomes, defined as all results of the change pro-cess observable and measurable in the intervention context.According to the phase of the change process, alterationsof proximate, intermediate, and long-term outcomes can bedistinguished.

Another important aspect of the model is its grid ofphases and levels applied to these categories and subcate-gories. Whereas phases refer to temporal distinctions withinthe implementation process and its discrete context (prepara-tion, action cycle, and appropriation), the levels refer to hier-archical aspects of the context and the intended interventionoutcomes, spanning from the individual to the organisation.The following description starts with this overlaying grid ofphases and levels and then proceeds to the categories andsubcategories of context, process, and outcome.

3. Intervention Phases and Levels

3.1. Intervention Phases. OHIs usually consist of severalelements such as participatory workshops, survey feedbacks,and information events.These planned intervention elementsare comprised within overall intervention architecture, thatis, the combination and sequence of intervention elements.Furthermore, intervention planning and evaluation conceptsusually distinguish between three and five interventionphases [8, 9, 11–13]. They share in common the notion thatan analysis is needed before actions can be planned andimplemented. To reduce complexity, the CPO evaluationmodel proposes three intervention phases: the preparationphase, the action cycle phase, and the appropriation phase.

The first, so-called preparation phase comprises all activ-ities needed to fit the intervention to the specific contextand to obtain the commitment of the organisation for thefollowing phases. This includes, for example, presentationsand workshops with decision makers, a qualitative analysisof the intervention context, the planning of the interventionarchitecture (who is involved, when, and how in the OHI),the establishment of a steering group, and project leader.

The second phase is referred to as the action cyclephase that comprises all activities needed to trigger a changeprocess that will improve organisational health on a broadscale. It encompasses the subphases analysis, action plan-ning, enactment, and monitoring. Analysis, for example,an organisation-wide stress assessment, is considered partof the action cycle phase as it not only serves to generateinformation but also is an active intervention element sinceits mere implementation might trigger small changes, suchas increased awareness, readiness for change, or sensibilityfor stressful issues [11]. It should be noted that we usethe term enactment to replace the commonly used termimplementation [8, 9, 12], as the activities in the preparationphase and the final appropriation phase should also beconsidered as part of the implementation process.


Discrete context

Preparationphase

Action cyclephase

Appropriationphase

Intervention(architecture and elements)

Omnibus context

OutcomesO

G

L

I

Implementation process

Org

anisa

tiona

l lev

els

Change process

Prox

imat

e out

com

es

Inte

rmed

iate

out

com

es

Dist

al o

utco

mes

Figure 1: The context, process, and outcome (CPO) evaluation model. O: organisation; G: group; L: leader; I: individual.

The third phase is named appropriation phase andcomprises all activities needed to ensure the continuation,advancement, and diffusion of the change process triggeredby the previous two phases. This phase refers to the periodwhen intervention implementers and researchers usuallyhave left the organisation. At that time, capacities for self-optimization have been built up and the organisation and itsmembers have to take over the responsibility for the contin-uation of the triggered change processes, for example, in theform of continued, repeated action cycles and optimisationprocesses. As appropriation is a precondition for achievingsustaining long-term effects, we consider it a crucial elementof any OHI evaluation.

3.2. Intervention Levels. OHI aims to impact different levelsof an organisation, often referred to as combined, individual-organisational or multilevel interventions [14]. In this regard,the CPO evaluation model distinguishes the levels of indi-viduals/leaders and groups/organisation (cf. the IGLO-levelsby Nielsen and colleagues [15]). Thus, outcome evaluationshould be conducted in consideration of these levels in orderto make differentiated statements concerning the effective-ness of an intervention. This differentiation is also importantfor the evaluation of the intervention context, where differentlevels can be of importance during different phases. Forexample, during the preparation phase, organisational levelfactors such as strategic goals in regard to employee healthare of particular importance; during the action cycle phase,leadership level factors such as linemanager attitudes are cru-cial; during the appropriation phase, group level factors suchas team climate and capacities for continued optimisation arerelevant.

4. The Main Categories andSubcategories of the Model

4.1. Context. For a long time, researchers have consideredcontext to be crucial for understanding the causes of success

and failure of interventions [16]. Context is often consideredas a process indicator [16–18], but in recent years, the per-ceived importance of and the attention to contextual issueshave increased. Slowly but surely, context has dissociateditself from process issues, becoming an autonomous andmeaningful factor in intervention research [19]. Researchfindings demonstrate that context is a very complex andintervention-specific factor [20] and that its effects can varyfrom subtle to powerful [21]. For instance, it might occurthat an intervention aiming to improve individual resourcesis implemented in two teams of different sizes, and outcomeevaluation may reveal the intervention to be effective in thesmaller team while ineffective in the larger team. In the casethat evaluators ignore further contextual aspects, they mightconclude that the intervention works only in small teams.However, a more differentiated context evaluation mightreveal that the line manager of the smaller team stronglysupported the intervention while the line manager of thelarger team tended to be critical of the intervention. Thisshows that neglect of contextual aspects could lead to funda-mental fallacies concerning the effectiveness of interventions.Accordingly, the concept of context is broadly discussedin OHI research [2] and more broadly in organisationalbehaviour [21–23]. Various researchers treat context as anunspecific setting parameter that concerns environmentaland situational aspects and limit its conceptualization to itsfacilitating and hindering functions [7, 18, 21]. Moreover, thecontext of an intervention is often considered as an undesired,uncontrollable, and unmanageable constraint that could beneither predicted nor controlled. For instance, Kompier andKristensen [24] state that “interventions always take placein context, and that this context is not under control ofscientists.”

The conceptualisation of context in the CPO evaluationmodel follows a taxonomy proposed by Bauer and Jenny[1] (in reference to human resource management practicesdistinguished by Delery and Doty [25]): “. . . the organization


as the context of [organisational health] interventions mightbe considered for selecting and targeting the intervention[universalistic approach], for adapting the intervention tothis context [contingency approach], or as the final target andactor of change [configurational approach].” Considering thisreciprocal, transformational relationship between contextand intervention, the CPO evaluation model defines contextas the underlying frame that influences and is influencedby an organisational health intervention. As such, contextis more than a static, nonchangeable boundary condition;it is a malleable riverbed directing the river of change andsimultaneously being shaped by the river. When we considercontext as a static boundary condition, we must accept thecritical linemanager in the above example as a nonchangeablehindering factor and run the risk of implementation failure.When we, on the other hand, consider context as an alterablecondition, we can initially adapt the intervention (river) tothe critical line manager (river bed) by assigning him/hera specific role, for example, by explicitly representing thecritical perspective in the steering committee of the project.By this active involvement, his/her critical position mightbe transformed into one of constructive support of theintervention.

4.1.1. Omnibus and Discrete Context. The CPO evaluationmodel distinguishes the omnibus context and the discretecontext, as recommended by Johns [21]. The omnibus con-text describes the general intervention and implementationsetting: occupation (who), location (where), time (when),and rationale (why), whereas the discrete context refersto “specific situational variables that influence behaviourdirectly or moderate relationships between variables” [21].According to Johns [21], specific situational variables maycomprise task aspects (autonomy, uncertainty, accountabil-ity, resources, etc.), social aspects (social density, structure,influence, etc.), and physical aspects (temperature, light, builtenvironment, decor, etc.). Relating to OHI, the omnibuscontext describes the overall setting in which the OHI takesplace independently of the three intervention phases. Forinstance, it may occur that an intervention at the team-level leads to positive changes in team climate, but thesechanges do not become apparent because of interfering effectstriggered by another project that is implemented at the sametime (e.g., introduction of shift work). In this case, if evalu-ators ignore contextual aspects, such as conflicting projects,they might conclude that the team-level intervention wasineffective; however, it actually had positive effects that weresuperimposed by another project.

The discrete context refers to specific individual, leader,group, and organisational (IGLO) aspects directly relevantto the implementation and change process. Randall andcolleagues [26] further distinguish between two temporalkinds of context: the baseline preintervention context and thecontext of activated intervention.The CPO evaluation modelconsiders this temporal differentiation by referring to thethree above-described phases. In the preparation phase, thediscrete context is often evaluated in regard to the questionof whether the organisation and its members are ready forthe next phase so that the implementation process can gain

momentum and flow in the action cycle phase as described inthe scenariowith the critical linemanager (see Section 4.1). Inthe action cycle phase, the discrete context is most commonlyevaluated in regard to the factors that, passively or actively,hindered or promoted the flow of the implementation andchange process. In the appropriation phase, the discretecontext is evaluated in regard to aspects that will maintainand further develop the induced changes or if they will bereversed.

The differentiation between the two kinds of contexts isimportantwith respect to the extent towhich the interventionimplementers can influence them. Often, aspects of theomnibus context are hardly or not at all manipulable byintervention implementers, for instance, economic develop-ments. It is nevertheless important to monitor and record theomnibus context for an understanding of the implementationprocess and an interpretation of intervention outcomes. Forexample, high dismissal rates caused by an economic crisis(omnibus context) will lead to lower absolute participationrates (implementation process) as a matter of course.

The discrete context, on the other hand, can be changed—not easily, but easier than the omnibus context—and, thus,should be considered as a target of change from the begin-ning. Due to the close proximity of the discrete context to theimplementation process, the discrete context (e.g., leaders’and employees’ readiness for change) can have an immediateand stronger influence on the implementation process thanthe omnibus context.

4.2. Process. Referring to Bauer und Jenny [1], process covers“both the implementation processes . . . and the intendedand unintended process of change triggered in organizationsand their employees, leading to alterations in . . .outcomes.”Following this understanding, the CPO evaluation modeldistinguishes between the implementation process and thechange process.

4.2.1. Implementation Process. The CPO evaluation modeldefines the implementation process as the time-limited enact-ment of all steps and elements of the original interventionplan. The intervention plan, or so-called overarching inter-vention architecture, concerns the combination and sequenceof single intervention elements that may vary in terms oflevel (IGLO), target audience (e.g., aging workers and specificdepartments), type (e.g., workshops, survey, presentations,and practical lessons), implementer (internal or externalconsultants), and so on. Further, the intervention elementscan be arranged in parallel (e.g., if they address differenttarget groups) or in a sequential order (e.g., if they build oneach other). There are a range of intervention architecturesfor OHIs, which usually also incorporate a set of generalimplementation principles to be followed during the process(cf. Table 1). For example, during the preparation phase,successful building of a strong coalition of project leaders,defining goals, and raising awareness and commitment areevaluated (all of which will be part of the subsequent discretecontext of the action cycle phase). During the action cyclephase, it is evaluated if the sequence and linkage of theintervention elements are implemented as planned, how


Table 1: Possible concepts to specify the main CPO categories.

CPO categoryDefinition

Selection of possible concepts in the intervention research literature(for concrete indicator examples see corresponding references)

Omnibus contextGeneral intervention andimplementation setting

(i) Occupation (who), location (where), time (when), and rationale (why) [21](ii) Economic developments, regulatory/trade/economic policies, technological innovations, andchanging worker demographics and labour supply (external context [44])(iii) Organizational restructuring, new quality and process management initiatives, alternativeemployment arrangements, work/life/family programs and flexible work arrangements, andchanges in benefit and compensation systems (organizational context [44])

Discrete contextSpecific individual, leader,group, and organisational(IGLO) aspects directlyrelevant to theimplementation and changeprocess

Leaders/individuals(i) Line manager attitudes, employee readiness, and intervention history [20](ii) Readiness for/stages of change [45, 46](iii) Mental models of stakeholders [7, 8]Groups/organisation(i) Awareness of norms, diversity, early role clarification, manager availability, and constructiveconflicts [47](ii) Climate and culture, task attributes, social-relational aspects of work, worker roles, and careerdevelopment (work context [44])(iii) Organisational resources, psychological resources, facilitating and obstructing elements ofthe design, and organization and management of work [48](iv) Task characteristics, social characteristics, and physical characteristics [21]

Implementation processTime-limited enactment of allsteps and elements of theoriginal intervention plan

Implementation of intervention elements(i) Recruitment, reach (e.g., number of workshop participants), dose delivered (e.g., number ofworkshops), dose received (e.g., engagement in workshops), fidelity of implementation asplanned, and participants attitudes to and satisfaction with the intervention [16, 17, 27, 30, 49, 50](ii) Participation in intervention decision, stakeholder appraisals of intervention plans andactivities, and observable and perceived exposure to intervention activities [48]Implementation of intervention architecture(i) Thorough diagnosis, definition of goals/vision, raising of shared problem awareness, buildingof coalition of leaders and drivers, lively communication, good time management, professionalproject organization and responsibilities, empowerment for self-optimisation, quick-wins andmotivation, process flexibility, monitoring and controlling, and anchoring of change (12 successfactors of change [51])(ii) Multilevel collaboration, active support from managers, explication of tacit knowledge,continuous evaluation and adjustment, visualisation of process and results, appointment offacilitator, and defined project period [52]

Change processAll intended and unintendedindividual and collectivedynamics triggered by theimplementation process,leading to alterations in theorganisation and its members

(i) Diffusion, shared meaning making, social identity building, social comparison processes,interpersonal influences, and social learning (psychosocial mechanisms of change [36])

Proximate outcomesAll results of the changeprocess that immediately arise

(i) Minor structural and strategic modifications (e.g., adapted agendas, rules of communication,and well-being checks [53])(ii) Changes in attitudes, values, and knowledge [8](iii) Individual competencies and collective capacities for self-optimisation in teams [54]

Intermediate outcomesAll results of the changeprocess with regard to factual(job-related) and social(people-related) processes

(i) Demand-control-support [55](ii) Effort-reward-Imbalance [56](iii) Job demands and resources [57] and ratio of resources and demands [58](iv) Team climate [59](v) Healthy organizational resources and practices: task resources, social resources, and healthypractices (HERO model [60])(vi) Collective general resistance resources [61](vii) Work-related sense of coherence [62]

Distal outcomesAll higher-level results of thechange process that evolveover time

(i) General health, mental health, and vitality (health and well-being scales of the COPSOQQuestionnaire [63])(ii) Healthy employees: efficacy beliefs, trust, positive emotions, resilience, and workengagement/healthy organizational outcomes: organizational commitment, high performance,customer loyalty/satisfaction, and corporate social responsibility (HERO model [60])(iii) Individual and collective sense of coherence [64]

Note. CPO: context, process, and outcome.


Preparationphase

Action cyclephase

Appropriationphase

Discrete context

Implementation process

What is done to fit the intervention to the

discrete context and prepare IGLOs for the

action cycle phase?

Is the intervention implemented as planned and perceived as favourable by

IGLOs?

What are IGLOs doing to maintain and further

develop the triggered changes?

Are IGLOs ready for the action cycle phase?

Do IGLOs facilitate/hinder the implementation process?

Are IGLOs capable of maintaining and further developing the triggered

changes?

Figure 2:Themain questions for evaluating the implementation process and the discrete context with regard to the three intervention phasesproposed by the CPO model. IGLO: individual, group, leader, and organisation.

employees, managers, and other stakeholders perceive theimplementation process, and if the intervention successfullyshapes a favourable discrete context for the appropriationphase. The latter means that capacities for self-optimisationare built up so that the organisation and its members arecapable and willing to further develop the triggers of changeprocesses autonomously. In the case of formative evaluationassignments, the progress of implementation is monitoredcontinuously in order to make adjustments to the originalintervention plan if necessary. In the appropriation phase,it is evaluated whether and how the further development,maintenance, and sustainability of the intervention effects areensured.

For the evaluation of the implementation of single inter-vention elements, many researchers focus on quantitativeindicators such as reach [27] or dose received [17]. Butqualitative implementation indicators are also applied [28].Commonly, researchers apply measures capturing the per-ceived quality of an intervention element [29, 30], whichhas proven to be an important factor when doing processevaluation [17, 30]. However, more research is needed onwhich indicators concerning the implementation of interven-tion elements are useful and how the appraisal of particularintervention elements influences the overall impact of anintervention.

Distinction between aspects of the implementation pro-cess and the discrete context is sometimes difficult; thus, inthe past, it has often been ignored. Figure 2 illustrates themain questions for evaluating the implementation processand the discrete context with regard to the three interventionphases.

4.2.2. Change Process. TheCPO evaluationmodel defines thechange process as all intended and unintended individual andcollective dynamics triggered by the implementation process,

leading to alterations in the organisation and its members.Thus, the change process potentially involves all levels of theintervention context from the individual to the organisation(and their environments). As a current overview of OHIapproaches shows [1], change processes include, for example,individual and organisational learning, social processes, tak-ing over others’ perspectives, realisation of jointly developedaction plans for improving work, organisational structure,and strategy.

Regarding the timeline of change, the CPO evaluationmodel is based on the assumption that the implementationand change processes inOHIs are initiatedwith the beginningof the preparation phase. During the action cycle phase andthe appropriation phase, the change process gains drive anddevelops its intended dynamics. This is illustrated in Figure 1by the increasing colour density of the change process arrow.Furthermore, the arrow suggests that the change process doesnot end with the appropriation phase. The implementationprocess should trigger changes in the everyday processesand structures in organisations leading to short- and long-term outcomes of interest. Due to the multilevel nature ofOHIs, change can spread beyond the single topics (e.g.,coping with stress, transformative leadership, and teamwork)or particular target audiences (e.g., single divisions and riskgroups) initially addressed by the intervention [31]—leadingto unforeseen desirable or undesirable side effects.

Evaluation of the change process can help to reveal whatchanges were triggered on which level by the implementationprocess and thus help to better understand the mecha-nisms of change. There is a lot of research on the differenttypes of change. Authors discuss whether change should beconsidered as episodic or continuous [32], exceptional ornatural [33], but when it comes to the assessment of change,most researchers focus on aspects of the discrete contextthat influence or shape the change process in a hindering


or facilitating way [34, 35]. As the change process is verycomplex and only partly observable, it still remains as akind of black box. In practice, intervention implementersusually compare baseline and follow-upmeasures of outcomevariables in order to make induced individual and organi-sational changes visible. Research on valid change processindicators is still scarce, but the psychosocial mechanismsof change described by Karanika-Murray and Biron [36]demonstrate a promising approach to fill this gap (cf. Table 1).They present “six psychosocial mechanisms that can explainhow an intervention can exert its impact on individualsand workgroups: diffusion, sharing, identifying, comparing,influencing and learning” [36]. Evaluating whether, how, andto which extent these mechanisms have evolved, for example,whether and how intervention effects at the individual levelhave diffused throughout an entire team, might help to makethe change process more visible.

4.3. Outcomes. Outcome evaluation is mainly concernedwith what effects an intervention has had [37]. The CPOevaluation model defines outcomes as all results of thechange process that are measurable and at the same timemeaningful for the organisation, its members, researchers,and other stakeholders.TheCPOmodel utilises the followingthree outcome categories: proximate outcomes, intermediateoutcomes, and distal outcomes. These categories are basedon a trichotomy of outcomes commonly used in the publichealth community [38–41]. As all outcomes manifest inthe intervention context, proximate, intermediate, and distaloutcomes can be further observed on all levels of individuals,leaders, groups, and the organisation.

Proximate outcomes, often also labelled as immediate-,initial- or short-term outcomes or first-level targets, refer, forexample, to individual skills and collective capacities neededfor the change process (i.e., as part of the discrete context) aswell as quick-wins in the form of minor but instant structuralchanges. As such, they can be classified as results of thechange process that immediately arise.

Intermediate outcomes, often also labelled as medium-term outcomes or second-level targets, comprise, for exam-ple, changes in job demands and resources with regard tofactual (job-related) processes such as work load or time pres-sure and social (people-related) processes such as leadershipbehaviour or social support [42]. As such, we can define themas results of the change process with regard to factual (job-related) and social (people-related) processes.

Distal outcomes refer to the distal objectives of the inter-ventions such as improved individual health and increasedorganisational performance, which often depend on changein the intermediate outcomes. These outcomes are oftenlabelled as the overall objectives/goals or, simply, impacts, andwe, thus, define them as higher-level results of the changeprocess that evolve over time.

In the literature, most intervention process cycles end upwith the evaluation phase measuring the outcome, impact,or effect of the intervention [8, 11, 12]. The CPO eval-uation model assumes that changes in outcomes happencontinuously as a result of the change process induced bythe continuous implementation process; therefore the CPO

model depicts evaluation of proximate, intermediate, anddistal outcomes in a separate box of the CPO model. Thus,outcome evaluation should be considered not only as animportant intervention element where results of the changeprocess are fed back into the organisational system at theend of the intervention, but also as a continuous observationand assessment of change results accompanying the entireimplementation and change process, for example, by con-tinuously measuring proximal outcomes to show successfulgrowth in these variables. In this regard, change in proximateoutcomes might already be visible after a project informationevent (preparation phase) or after a particular interventionelement, such as a stress management workshop (action cyclephase). Thus, collection and reflection of outcome data arealso parts of an intervention and can influence the changeprocess [11].

5. Practical Application ofthe CPO Evaluation Model

5.1. Possible Concepts to Specify the Main CPO Categories.The CPO model defines generic categories, which meansthat they have to be further specified and operationalisedby concrete indicators for the purpose of measurement andevaluation. Table 1 presents a variety of possible concepts andrelated indicators for these generic categories of the CPOmodel, considering different intervention approaches andprinciples. The list is not exhaustive and is mainly compiledon the basis of recent intervention and evaluation research inoccupational health psychology [1]. The concrete selection ofindicators will depend on the specific intervention theoriesand logic; interest of stakeholders; and available resourcessuch as budget, time, and manpower. The concepts proposedfor evaluating the main CPO categories reveal consider-able variance in specification, terminology, and objective—although common themes emerge and could be condensedin future work.

5.2. Practical Application of the CPOModel Using the Exampleof an Evaluation of an OHI at a Hospital in Switzerland.Table 1 presents a list of relatively general concepts to specifythe main CPO categories. In order to clarify how the the-oretical concepts of the CPO categories can be translatedinto practice and to show the practical application of theCPO evaluation model in a more concrete way, Table 2exemplarily illustrates the model’s use for the conceptu-alization of an OHI evaluation by an OHI project thatincluded 31 nursing divisions from a hospital in Switzerland(see Acknowledgments) implemented between 2013 and2015.

The data collection, data analyses, and the evaluationreport of this OHI were structured using the CPO logic. Inthe first column of Table 2, research questions that mightguide the evaluation of an OHI along the CPO categories arepresented. The second and third columns show the appliedthemes/indicators and methods/sources for each CPO cat-egory. To better understand the applied themes/indicatorsand methods/sources, a brief overview of the project with


Table 2: Evaluation of an organisational health intervention (OHI) with special focus on lean processes at a hospital in Switzerland.

CPO category and correspondingresearch question Themes/indicator Methods/source

Omnibus contextIn what kind of organisation (size,structure, etc.) is the interventionimplemented?How does the political, social, andeconomic environment look like?

(i) General information on the hospital: location,number of divisions and employees, hierarchicalstructure, financing, and so forth(ii) Specific information on nursing departments:structure of nursing departments, types of nursingprofessions, number and characteristics of nursingpersonnel, and so forth(iii) Current political and economic situation/changesthat are relevant for the nursing divisions(iv) Previous intervention projects in the nursingdivisions

(i) Discussions with projectleaders/head of nursing divisions(ii) Documentary analysis(documents provided by thehospital)

Discrete contextAre IGLOs ready for the actioncycle phase?Do IGLOs facilitate/hinder theimplementation process?Are IGLOs capable of maintainingand further developing thetriggered changes?

Leaders/individuals(i) Project commitment and readiness for changeGroups/organisation(i) Planned projects/alterations during the interventionduration(ii) Reasons for the intervention project(iii) Provided resources (time and budget), stability ofproject personnel, and information politics(iv) Openness for novelty, communication culture, andchannels

(i) Discussions with projectleaders/head of nursing divisions(ii) Four focus group discussions

Implementation processWhat is done to fit the interventionto the discrete context and prepareIGLOs for the action cycle phase?Is the intervention implemented asplanned and perceived as favourableby IGLOs?Are capacities for appropriationbuilt up? What are IGLOs doing tomaintain and further develop thetriggered changes?

Implementation of intervention elements(i) Implementation of workshops

(a) Number of workshops(b) Number of workshop participants(c) Composition of participants(d) Quality appraisals of workshop(e) Outcome expectancies(f) Satisfaction with measures (developed during theworkshop)(g) Output of measures for improving the worksituation

(ii) Implementation of employee surveys(a) Survey period(b) Reach

Implementation of intervention architecture(i) Open questions

(i) Intervention planning chart(ii) List of measures (developedduring the workshop)(iii) Short questionnaires appliedat the second and fourthworkshop days(iv) Participation rate in thethree-wave survey/informationon the team sizes given by theteam leaders(v) Focus group discussions

Change processDoes the implementation trigger aprocess of change?How do the triggered changesdisseminate among IGLOs?How do the triggered changesevolve over time?

(i) Transfer of workshop training and output to theteam: communication, actions of workshopparticipants, and reactions of nonworkshop participants(ii) Visibility of the implementation of measures tononworkshop participants(iii) Dynamics triggered with regard to interpersonalinfluences and social learning within the teams

(i) Focus group discussions

Proximate outcomesWhich effects arise immediately?

(i) Changes concerning waste of resources, efficient useof time, and collaboration (i) Focus group discussions

Intermediate outcomesWhich effects arise with regard tofactual (job-related) and social(people-related) processes?

(i) Changes in the resource-demands-ratio(ii) Changes in the interprofessional collaboration(iii) Changes in team climate, work organisation, andsupervisor behaviour

(i) Three-wave survey(ii) Focus group discussions

Distal outcomesWhich higher-level effects evolveover time?

(i) Positive psychosocial health (engagement andsatisfaction)(ii) Negative psychosocial health (stress symptoms andnegative feelings)

(i) Three-wave survey(ii) Focus group discussions

Note. CPO: context, process, and outcome; IGLO: individual, group, leader, and organisation.


regard to the intervention’s architecture, goals, elements,and evaluation instruments is presented in the followingparagraph.

The intervention focused on optimizing the workingprocesses in the nursing divisions with regard to lean prin-ciples (i.e., reducing waste and enhancing value), fosteringinterdisciplinary collaboration between nursing and medicalstaff, strengthening team climate within the divisions, andimproving the balance of resources and demands of thenursing teams. A four-day workshop was implemented ineach division by the hospital’s internal process managers asthe main intervention component. A representative selec-tion of employees participated in the workshop; workshopparticipants consisted of registered nurses at all hierarchicallevels, includingwardmanagers.Theworkshopwas evaluatedusing a paper-based evaluation questionnaire. To evaluatethe intervention as a whole, an online survey (comprisingitems on resources, demands, teamclimate, work-life balance,etc.) was applied three times at six-month intervals in eachdivision.

The evaluation had a waitlist control group design; thatis, the 31 nursing divisions were randomly assigned to eitheran intervention (𝑛 = 16) or control group (𝑛 = 15). Forthe intervention groups, the workshop took place four tosix weeks after the first online survey; for the control group,the workshop took place four to six weeks after the secondonline survey. Furthermore, four focus group discussionswere conducted at the end of the intervention project togather data on the implementation and change process as wellas on the discrete context.The focus group discussion partic-ipants were composed as follows: (1) workshop participantswho rated the intervention impact positive; (2) workshopparticipants who rated the intervention impact negative; (3)nonworkshop participants who rated the intervention impactpositive; and (4) nonworkshop participants who rated theintervention impact negative.

6. Limitations

A thorough evaluation of all CPO categories and subcate-gories is a difficult endeavour requiring a lot of resourcesand instruments. In many cases, it will not be possible toevaluate all these aspects due to scarce resources or limitedaccess to information. Furthermore, it can be difficult tospecify the CPO categories and develop suitable indicatorsdue to the variety of possible concepts. However, the CPOmodel at least offers an overview of which categories couldbe evaluated and how these categories could be specified.Depending on the particular intervention project, evaluatorscan then consciously decide which of these categories shouldbe evaluated and to what extent. It can sometimes bedifficult to distinguish between aspects of the implemen-tation process and the discrete context but the questionsdisplayed in Figure 2 may facilitate the distinction of thesetwo categories. The most arguable category of the CPOmodel is the change process, as it still remains a kind ofblack box. Consequently, more research is needed on howto assess the change process. Furthermore, the CPO modelwas only tested in two intervention projects so far. Use of

the CPO model across diverse OHIs and evaluation studiesmay produce a systematic evidence base, building on genericcategories for collecting and reporting data. This will helpresearchers and practitioners to develop more effective andsustainable interventions in the future. Finally, preexistingconcepts and indicators have to be tested and compared indifferent intervention projects in future research in order toidentify the critical concepts and develop corresponding validindicators.

7. Conclusions

The CPO evaluation model provides a basis for structuredevaluation of combined OHIs in the field by combiningcontext, process, and outcome evaluation. It offers genericevaluation categories and subcategories that are furtherdifferentiated in terms of time and hierarchy and displayedby a grid of intervention phases and organisational levelsthat facilitates detecting changes at different times and levels.Furthermore, it provides a clear taxonomy for a wide rangeof possible concepts specifying the evaluation categoriesand subcategories. Development, testing, and selection ofconcrete indicators need to be realized in future research.However, descriptions of the evaluation of the OHI projectat the hospital demonstrate how an evaluation guided andstructured by the CPO evaluation model might look. Theresearch questions presented in Table 2 will support inter-vention researchers in selecting appropriate indicators for aparticular intervention project.

In comparison to similar models, the CPO model uses aclearly defined terminology for OHIs, which might facilitatethe development of a common language for improving bothcommunication between researchers and company membersand the comparability and aggregation of evaluation studyresults.

Second, it distinguishes the implementation process fromthe change process. This distinction is essential as it helpsdifferentiate between cause (i.e., the implementation process)and effect (i.e., the triggered change process), which in turnhelps researchers and company members understand themechanics of change. Following Harachi and colleagues [43],it also helps identify the possible causes of interventionfailure. In the first instance, the failure may be caused byan implementation failure, which means “that the way theintervention was implemented was incomplete or designedin such a way that the intervention would have failed evenif the theory behind the intervention was correct” [7].Alternatively, the theory/programme failure may be basedon false assumptions about how the implementation of theintervention translates into desired outcomes through anassumed change process,meaning “that the theory behind theproblem did not address the problem” [7].

Third, it assumes a reciprocal relationship between theimplementation process and the discrete context. By doingso, it broadens the hitherto existing understanding of a staticcontext by defining it as a dynamic factor that needs to besystematically considered and that can be transformed duringthe intervention.


Fourth, it considers outcome evaluation as the con-tinuous observation and assessment of the change resultsaccompanying the entire intervention. Conducting an out-come evaluation continuously from the beginning will helpto better understand the dynamics of the change process andto prevent evaluation results from being subject to hindsightbiases.

Overall, the CPO evaluation model can serve as ashared mental model throughout the complex interventionevaluation process by supporting organisational members,project leaders, implementers, and evaluators in establishingobjectives, selecting possible evaluation concepts from OHIliterature, developing key indicators, gathering data, andreporting evaluation results in a structured and succinct way.


The authors have declared that no conflict of interests exists.

Acknowledgments

The intervention project described in the paper was financedby “Lotteriefonds,” Suva, as well as the University HospitalZurich. Annemarie Fridrich was supported by the SwissNational Science Foundation (SNSF).

References

[1] G. F. Bauer and G. J. Jenny, “From fidelity to figuration: currentand emerging approaches to organizational health interven-tion research,” in Salutogenic Organizations and Change: TheConcepts Behind Organizational Health Intervention Research,G. B. Bauer and G. J. Jenny, Eds., Springer, Dordrecht, TheNetherlands, 2013.

[2] C. Biron, M. Karanika-Murray, and C. L. Cooper, “Organiza-tional interventions for stress and well-being—an overview,”in Improving Organizational Interventions for Stress and Well-Being. Addressing Process and Context, C. Biron, M. Karanika-Murray, and C. L. Cooper, Eds., pp. 1–17, Routledge, Hove, UK,2012.

[3] T. Cox,M. Karanika, A. Griffiths, and J. Houdmont, “Evaluatingorganizational-level work stress interventions: beyond tradi-tional methods,”Work & Stress, vol. 21, no. 4, pp. 348–362, 2007.

[4] A. D. LaMontagne, T. Keegel, A. M. Louie, A. Ostry, and P. A.Landsbergis, “A systematic review of the job-stress interventionevaluation literature, 1990–2005,” International Journal of Occu-pational and Environmental Health, vol. 13, no. 3, pp. 268–280,2007.

[5] K. M. Richardson and H. R. Rothstein, “Effects of occupationalstress management intervention programs: a meta-analysis,”Journal of Occupational Health Psychology, vol. 13, no. 1, pp. 69–93, 2008.

[6] K. Nielsen, T.W. Taris, and T. Cox, “The future of organizationalinterventions: addressing the challenges of today’s organiza-tions,”Work & Stress, vol. 24, no. 3, pp. 219–233, 2010.

[7] K. Nielsen and R. Randall, “Opening the black box: presenting amodel for evaluating organizational-level interventions,” Euro-pean Journal ofWork andOrganizational Psychology, vol. 22, no.5, pp. 601–617, 2013.

[8] K. Nielsen and J. S. Abildgaard, “Organizational interventions:a research-based framework for the evaluation of both processand effects,”Work and Stress, vol. 27, no. 3, pp. 278–297, 2013.

[9] L. M. Goldenhar, A. D. LaMontagne, T. Katz, C. Heaney, andP. Landsbergis, “The intervention research process in occupa-tional safety and health: an overview from the National Occu-pational Research Agenda Intervention Effectiveness Researchteam,” Journal of Occupational and EnvironmentalMedicine, vol.43, no. 7, pp. 616–622, 2001.

[10] R. Pawson andN.Tilley,Realistic Evaluation, Sage, London,UK,1997.

[11] A. Inauen, G. J. Jenny, and G. F. Bauer, “Design principles fordata- and change-oriented organisational analysis in workplacehealth promotion,” Health Promotion International, vol. 27, no.2, pp. 275–283, 2012.

[12] P. Mendel, L. S. Meredith, M. Schoenbaum, C. D. Sher-bourne, and K. B. Wells, “Interventions in organizational andcommunity context: a framework for building evidence ondissemination and implementation in health services research,”Administration and Policy in Mental Health and Mental HealthServices Research, vol. 35, no. 1-2, pp. 21–37, 2008.

[13] E. Wimbush and J. Watson, “An evaluation framework forhealth promotion: theory, quality and effectiveness,”Evaluation,vol. 6, no. 3, pp. 301–321, 2000.

[14] N. K. Semmer, “Job stress interventions and the organization ofwork,” Scandinavian Journal of Work, Environment and Health,vol. 32, no. 6, pp. 515–527, 2006.

[15] K. Nielsen, M. Stage, J. S. Abildgaard, and C. V. Brauer,“Participatory intervention from an organizational perspective:employees as active agents in creating a healthy work environ-ment,” in Salutogenic Organizations and Change: The ConceptsBehind Organizational Health Intervention Research, G. F. Bauerand G. J. Jenny, Eds., Springer, Dordrecht, The Netherlands,2013.

[16] L. Linnan and A. B. Steckler, “Process evaluation for publichealth interventions and research: an overview,” in ProcessEvaluation for Public Health Interventions and Research, A.B. Steckler and L. Linnan, Eds., pp. 1–24, Jossey-Bass, SanFrancisco, Calif, USA, 2002.

[17] S. G. Murta, K. Sanderson, and B. Oldenburg, “Process evalua-tion in occupational stress management programs: a systematicreview,” The American Journal of Health Promotion, vol. 21, no.4, pp. 248–254, 2007.

[18] J. van Berkel, C. R. L. Boot, K. I. Proper, P. M. Bongers, andA. J. van der Beek, “Process evaluation of a workplace healthpromotion intervention aimed at improving work engagementand energy balance,” Journal of Occupational and EnvironmentalMedicine, vol. 55, no. 1, pp. 19–26, 2013.

[19] C. Biron, M. Karanika-Murray, and C. L. Cooper, Eds., Improv-ing Organizational Interventions for Stress and Well-Being:Addressing Process and Context, Routledge, Hove, UK, 2012.

[20] R. Randall, K.Nielsen, and S.D. Tvedt, “Thedevelopment of fivescales to measure employees’ appraisals of organizational-levelstress management interventions,”Work and Stress, vol. 23, no.1, pp. 1–23, 2009.

[21] G. Johns, “The essential impact of context on organizationalbehavior,” Academy of Management Review, vol. 31, no. 2, pp.386–408, 2006.

[22] M. A. Griffin, “Specifying organizational contexts: systematiclinks between contexts and processes in organizational behav-ior,” Journal of Organizational Behavior, vol. 28, no. 7, pp. 859–863, 2007.


[23] R. T. Mowday and R. I. Sutton, “Organizational behavior:linking individuals and groups to organizational contexts,”Annual Review of Psychology, vol. 44, no. 1, pp. 195–229, 1993.

[24] M. Kompier and T. S. Kristensen, “Organizational work stressinterventions in a theoretical, methodological and practicalcontext,” in Stress in the Workplace. Past, Present and Future,J. Dunham, Ed., pp. 164–190, Whurr Publishers, London, UK,2000.

[25] J. E. Delery and D. H. Doty, “Modes of theorizing in strategichuman resource management: tests of universalistic, contin-gency, and configurational performance predictions,” Academyof Management Journal, vol. 39, no. 4, pp. 802–835, 1996.

[26] R. Randall, T. Cox, and A. Griffiths, “Participants’ accounts of astress management intervention,”Human Relations, vol. 60, no.8, pp. 1181–1208, 2007.

[27] R. E. Glasgow, E. Lichtenstein, andA. C.Marcus, “Why don’t wesee more translation of health promotion research to practice?Rethinking the efficacy-to-effectiveness transition,” AmericanJournal of Public Health, vol. 93, no. 8, pp. 1261–1267, 2003.

[28] G. Baril-Gingras, M. Bellemare, and C. Brisson, “How canqualitative studies help understand the role of context andprocess of interventions on occupational safety and healthand on mental health at work?” in Improving OrganizationalInterventions for Stress and Well-Being: Addressing Process andContext, C. Biron, M. Karanika-Murray, and C. L. Cooper, Eds.,pp. 135–162, Routledge, Hove, UK, 2012.

[29] B. Aust, R. Rugulies, A. Finken, and C. Jensen, “When work-place interventions lead to negative effects: learning fromfailures,” Scandinavian Journal of Public Health, vol. 38, no. 3,supplement, pp. 106–119, 2010.

[30] K. Nielsen, R. Randall, and K. Albertsen, “Participants’appraisals of process issues and the effects of stressmanagementinterventions,” Journal of Organizational Behavior, vol. 28, no. 6,pp. 793–810, 2007.

[31] M. Karanika-Murray, C. Biron, and C. L. Cooper, “Concludingcomments. Distilling the elements of successful organizationalintervention implementation,” in Improving OrganizationalInterventions for Stress and Well-Being. Addressing Process andContext, C. Biron, M. Karanika-Murray, and C. L. Cooper, Eds.,pp. 353–361, Routledge, Hove, UK, 2012.

[32] L. L. Martins, “Organizational change and development,” inAPA Handbook of Industrial and Organizational Psychology, S.Zedeck, Ed., pp. 691–728, American Psychological Association,Washington, DC, USA, 2011.

[33] H. Tsoukas and R. Chia, “On organizational becoming: rethink-ing organizational change,” Organization Science, vol. 13, no. 5,pp. 567–582, 2002.

[34] A. A. Armenakis and S. G. Harris, “Reflections: our journey inorganizational change research and practice,” Journal of ChangeManagement, vol. 9, no. 2, pp. 127–142, 2009.

[35] L. E. Tetrick, J. C. Quick, and P. L. Gilmore, “Research in orga-nizational interventions to improve well-being: perspectives onorganizational change and development,” in Improving Orga-nizational Interventions for Stress and Well-Being. AddressingProcess and Context, C. Biron, M. Karanika-Murray, and C. L.Cooper, Eds., pp. 59–76, Routledge, Hove, UK, 2012.

[36] M. Karanika-Murray and C. Biron, “The nature of changein organizational health interventions: some observations andpropositions,” in Salutogenic Organizations and Change: TheConcepts Behind Organizational Health Intervention Research,G. F. Bauer and G. J. Jenny, Eds., pp. 239–258, Springer,Dordrecht, The Netherlands, 2013.

[37] R. Z. Goetzel and R. J. Ozminkowski, How to Evaluate Work-place Health Promotion Programs, American Journal of HealthPromotion, Keego Harbor, Mich, USA, 2000.

[38] D. A. Julian, A. Jones, and D. Deyo, “Open systems evaluationand the logic model: program planning and evaluation tools,”Evaluation and Program Planning, vol. 18, no. 4, pp. 333–341,1995.

[39] M. M. Mark and G. T. Henry, “Logic models and contentanalyses for the explication of evaluation theories: the case ofemergent realist evaluation,” Evaluation and Program Planning,vol. 38, pp. 74–76, 2013.

[40] J. A. McLaughlin and G. B. Jordan, “Logic models: a toolfor telling your program’s performance story,” Evaluation andProgram Planning, vol. 22, no. 1, pp. 65–72, 1999.

[41] D. Nutbeam, “Evaluating health promotion—progress, prob-lems and solutions,”Health Promotion International, vol. 13, no.1, pp. 27–44, 1998.

[42] G. J. Jenny and G. F. Bauer, “The limits of control: a systemicmodel-based approach to changing organisations towards bet-ter health,” in Salutogenic Organizations and Change: The Con-cepts behind Organizational Health Intervention Research, G. F.Bauer andG. J. Jenny, Eds., pp. 167–187, Springer,Dordrecht,TheNetherlands, 2013.

[43] T. W. Harachi, R. D. Abbott, R. F. Catalano, K. P. Haggerty, andC. B. Fleming, “Opening the black box: using process evaluationmeasures to assess implementation and theory building,” TheAmerican Journal of Community Psychology, vol. 27, no. 5, pp.711–731, 1999.

[44] National Institute forOccupational Safety andHealth (NIOSH),The Changing Organization of Work and the Safety and Healthof Working People: Knowledge Gaps and Research Directions,National Institute forOccupational Safety andHealth (NIOSH),Washington, DC, USA, 2002.

[45] B. J. Weiner, H. Amick, and S.-Y. D. Lee, “Review: concep-tualization and measurement of organizational readiness forchange. A review of the literature in health services research andother fields,” Medical Care Research and Review, vol. 65, no. 4,pp. 379–436, 2008.

[46] J. M. Prochaska, “A transtheoretical model for assessing organi-zational change: a study of family service agencies’ movementto time-limited therapy,” Families in Society, vol. 81, no. 1, pp.76–84, 2000.

[47] P. Ø. Saksvik, S. D. Tvedt, K. Nytrø et al., “Developing criteriafor healthy organizational change,”Work and Stress, vol. 21, no.3, pp. 243–263, 2007.

[48] R. Randall, “Process monitoring in intervention research: a‘dashboard’ with six dimensions,” in Salutogenic Organizationsand Change: The Concepts Behind Organizational Health Inter-vention Research, G. F. Bauer and G. J. Jenny, Eds., Springer,Dordrecht, The Netherlands, 2013.

[49] A. Linna, A. Vaananen, M. Elovainio, M. Kivimaki, J. Pentti,and J. Vahtera, “Effect of participative intervention on organ-isational justice perceptions: a quasi-experimental study onfinnish public sector employees,” International Journal ofHuman Resource Management, vol. 22, no. 3, pp. 706–721, 2011.

[50] T. L. Veach, R. H. Rahe, R. L. Tolles, and L. M. Newhall,“Effectiveness of an intensive stress intervention workshop forsenior managers,” Stress and Health, vol. 19, no. 5, pp. 257–264,2003.

[51] M.Gerkhardt andD. Frey, “Erfolgsfaktoren und psychologischeHintergrunde in Veranderungsprozessen: Entwicklung eines


integrativen psychologischen Modells,” OrganisationsEntwick-lung, vol. 25, no. 4, pp. 48–59, 2006.

[52] C. Ipsen and V. Andersen, “A multi-level and participatorymodel for prevention of work-related stress in knowledgework,” in Salutogenic Organizations and Change: The Conceptsbehind Organizational Health Intervention Research, G. F. Bauerand G. J. Jenny, Eds., pp. 127–148, Springer, Dordrecht, TheNetherlands, 2013.

[53] G. J. Jenny, A. Inauen, R. Brauchli, D. Fullemann, F. Muller,and G. F. Bauer, SWiNG Project—Final Report of the Evaluation,Public andOrganizationalHealthDepartment, Zurich, Switzer-land, 2011.

[54] G. J. Jenny and G. F. Bauer, “The limits of control: a systemic,modelbased approach to changing organisations towards betterhealth,” in Salutogenic Organizations and Change. The Conceptsbehind Organizational Health Intervention Research, G. F. Bauerand G. J. Jenny, Eds., Springer, Dordrecht, The Netherlands,2013.


[56] J. Siegrist, “Adverse health effects of high-effort/low-rewardconditions.,” Journal of Occupational Health Psychology, vol. 1,no. 1, pp. 27–41, 1996.

[57] A. B. Bakker and E. Demerouti, “The job demands-resourcesmodel: state of the art,” Journal of Managerial Psychology, vol.22, no. 3, pp. 309–328, 2007.

[58] G. J. Jenny, R. Brauchli, A. Inauen, D. Fullemann, A.Fridrich, and G. F. Bauer, “Process and outcome evaluationof an organizational-level stress management intervention inSwitzerland,” Health Promotion International, 2014.

[59] N. R. Anderson and M. A. West, “Measuring climate for workgroup innovation: development and validation of the teamclimate inventory,” Journal of Organizational Behavior, vol. 19,no. 3, pp. 235–258, 1998.

[60] M. Salanova, S. Llorens, E. Cifre, and I. M. Martınez, “We needa hero! toward a validation of the Healthy and Resilient Organi-zation (HERO) model,” Group and Organization Management,vol. 37, no. 6, pp. 785–822, 2012.

[61] L. Vaandrager and M. Koelen, “Salutogenesis in the workplace:building general resistance resources and sense of coherence,”in Salutogenic Organizations and Change: The Concepts BehindOrganizational Health Intervention Research, G. F. Bauer and G.J. Jenny, Eds., pp. 77–89, Springer, Dordrecht, The Netherlands,2013.

[62] K. Vogt, G. J. Jenny, and G. F. Bauer, “Comprehensibility,manageability andmeaningfulness atwork: construct validity ofa scale measuring work-related sense of coherence,” SA Journalof Industrial Psychology, vol. 39, no. 1, article 1111, 2013.

[63] T. S. Kristensen, H. Hannerz, A. Høgh, and V. Borg, “TheCopenhagen psychosocial questionnaire—a tool for the assess-ment and improvement of the psychosocial work environment,”Scandinavian Journal of Work, Environment & Health, vol. 31,no. 6, pp. 438–449, 2005.

[64] O. Idan, O. Braun-Lewensohn, and S. Sagy, “Qualitative, senseof coherence-based assessment of working conditions in apsychiatric in-patient unit to guide salutogenic interventions,”in Salutogenic Organizations and Change. The Concepts behindOrganizational Health Intervention Research, G. F. Bauer and G.J. Jenny, Eds., pp. 55–74, Springer, Dordrecht, The Netherlands,2013.

Research ArticleAssociations between Distal Upper Extremity Job PhysicalFactors and Psychosocial Measures in a Pooled Study

Matthew S. Thiese,1 Kurt T. Hegmann,1 Jay Kapellusch,2 Andrew Merryweather,3

Stephen Bao,4 Barbara Silverstein,4 and Arun Garg2

1RockyMountain Center for Occupational and Environmental Health (RMCOEH), University of Utah, Salt Lake City, UT 84108, USA2Center for Ergonomics, University of Wisconsin-Milwaukee, Milwaukee, WI 53201, USA3Department of Mechanical Engineering, University of Utah, Salt Lake City, UT 84112, USA4Safety andHealth Assessment and Research for Prevention (SHARP) Program,Washington State Department of Labor and Industries,Olympia, WA 98504, USA

Correspondence should be addressed to Matthew S. Thiese; [email protected]

Received 17 January 2015; Accepted 10 April 2015


Copyright © 2015 Matthew S. Thiese et al. This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.

Introduction. There is an increasing body of literature relating musculoskeletal diseases to both job physical exposures andpsychosocial outcomes. Relationships between job physical exposure measures and psychosocial factors have not been wellexamined or quantified. These exploratory analyses evaluate relationships between quantified exposures and psychosocialoutcomes. Methods. Individualized quantification of duration, repetition, and force and composite scores of the Strain Index (SI)and the Threshold Limit Value for Hand Activity Level (TLV for HAL) were compared to 10 psychosocial measures. Relationshipsand predicted probabilities were assessed using ordered logistic regression. Analyses were adjusted for age, BMI, and gender.Resultsand Discussion. Among 1834 study participants there were multiple statistically significant relationships. In general, as duration,repetition, and force increased, psychosocial factors worsened. However, general health and mental exhaustion improved withincreasing job exposures. Depression was most strongly associated with increased repetition, while physical exhaustion was moststrongly associated with increased force. SI and TLV for HAL were significantly related to multiple psychosocial factors. Theserelationships persisted after adjustment for strong confounders. Conclusion. This study quantified multiple associations betweenjob physical exposures and occupational and nonoccupational psychosocial factors. Further research is needed to quantify theimpacts on occupational health outcomes.

1. Introduction

Upper extremity musculoskeletal disorders (UEMSDs) areamongst the most prevalent and costly compensated dis-orders in worker’s compensation systems. In Washingtonsate, work-related musculoskeletal disorders of the upperextremity and spine occurring without discrete trauma havebeen estimated to encompass 42.5% of compensable claimsand lost time claims, respectively [1]. The highest rates ofreported hand/wrist musculoskeletal disorders have been inthe construction and manufacturing industries [1, 2].

A study fromWashington state using theNorthAmericanIndustry Classification System (NAICS) with the Prevention

Index reported that, among the top 25 industries, the highestmedian compensable costs per worker’s compensation claimfrom 2002 to 2010 were in construction ranging from $11,280to $30,101 [3]. Manufacturing costs per claim ranged from$8,869 to $10,914. Median costs in the services sector rangedfrom $5,687 to $10,053 [3].

Despite high prevalence rates, underreporting of injuriesis reportedly widespread in the US and France [4]. A recentsystematic analysis found that 90% of employers underreportoccupational injuries and illnesses in Washington state [5].The strongest predictors of underreporting included oper-ating multiple shifts and use of the data for supervisor orrespondent’s job performance.


http://dx.doi.org/10.1155/2015/643192


There have been reports investigating relationshipsbetween psychosocial factors and work-related muscu-loskeletal disorders, many of which include low back andneck pain [6–15]. These reports have detailed relation-ships between musculoskeletal outcomes and both work-organizational and personal psychosocial factors. Bongers etal. reported that a range of job organizational factors wererelated to both back pain and neck pain across the literature[9]. Another study found that multiple work organizationalfactors were associated with neck, shoulder, and low backpain, with strongest associations found with neck pain [12].Only a few studies have investigated relationships betweenpsychosocial factors and UEMSDS.

While there is increasing recognition of the importanceof psychosocial factors in occupational health and safety,reports of psychosocial factors in relation to job physicalfactors for UEMSDs are uncommon. The most commonlyreported associations are between surgical carpal tunnelsyndrome (CTS) patient’s disability and/or pain outcomeswith depression [16, 17], pain anxiety [18, 19], and catas-trophization [20, 21]. Depression and pain anxiety, but notneuroticism, are associated with worse upper extremityfunction in UEMSD patients that include some CTS patients[22]. However, whether the outcomes of surgical or medicalcase series mirror those in populations of workers is largelyunknown.

Studies of association between psychosocial factors andincreased risk of UEMSDs are limited and provide conflictingevidence. A prospective cohort study of CTS reported highjob strain and low social support reported CTS risks [23],while a second report of the same cohort found mostlynegative results from analyses of workplace psychosocialfactors such as decision latitude [24]. One cross-sectionalstudy of workers found associations between CTS and bothjob dissatisfaction and high job demands [25]. Another studyof workers with UEMSDs that included a minority of CTScases reported risks including low decision authority, highpsychological demand, and low supervisor support, althoughjob satisfaction and affective disorders were not reported[26]. There was no relationship between hand symptomsand job dissatisfaction in a study of hand therapists [27].Job dissatisfaction and poorer physical health have beenassociated with CTS in a case-control study [6]; however, aprospective cohort study found no association between jobsatisfaction and new UEMSDs [28]. A study among Frenchworkers found job dissatisfaction to be weakly associatedwith symptoms only CTS case definition in a cross-sectionalstudy without measured job exposure factors. That studyalso found that low job control was associated with one oftwo statistical models and psychological and psychosomatic“problems” are associated with CTS [29].

In contrast with psychosocial factors, there is an increas-ing body of literature prospectively quantifying relation-ships between job physical measurements and carpal tunnelsyndrome [23, 24, 30–36], trigger digit [32], and lateralepicondylalgia [37] using measurement tools such as theAmericanConference ofGovernmental IndustrialHygienists(ACGIH), Threshold Limit Value for Hand Activity Level(TLV for HAL), and Strain Index (SI). In those studies,

psychosocial factors have been largely treated as potentialconfounders, without assessment of magnitude of relation-ships and/or potential interactions between job physicalexposures and psychosocial factors.

The objectives of this report are to perform exploratoryanalyses for potential relationships between job physicalmeasures including (a) the TLV for HAL, (b) the SI, and (c)measures of force and repetition, with the psychosocial fac-tors of job satisfaction, coworker support, supervisor support,physical exhaustion, mental exhaustion, anxiety, depressivesymptoms, and general health. The general hypothesis is thatincreasing job physical exposures (e.g., higher force, higherrepetition) will be associated with worsening of psychosocialresponses.


This pooled study was approved by the Institutional ReviewBoards of Washington State, University of Wisconsin-Milwaukee and the University of Utah. Detailed descriptionsofmethods and data collection instruments used in this studyare available and have been previously published [31, 32, 37,38]; thus, abbreviated methods follow.

This study includes workers recruited from 35 diversefacilities representing 25 industries located in Illinois, Utah,Washington, and Wisconsin. These employees performedjobs in the manufacturing, food processing, healthcare,and office sectors. All workers provided written, informedconsented prior to enrollment.

2.1. Psychosocial Factors and Demographic Data. Psychoso-cial factors and demographic data, including medical history,were collected using electronic questionnaires. Body massindices were calculated from measured heights and weights.All data were collected by trained researchers who wereblinded to the job physical exposures of the workers.

A total of 10 psychosocial measures were commonbetween all three research sites. These included (1) generalhealth compared to others, (2) depressive symptoms, (3)physical exhaustion after work, (4) mental exhaustion afterwork, (5) how well participants get along with coworkers,(6) job satisfaction, (7) how well participants get along withtheir closest or immediate supervisor, (8) degree to whichparticipants would recommend their job to others, (9) ifparticipants would take the job again, and (10) degree towhich participants feel that their employer cares about theirhealth and safety on the job. Responses were categorizedinto 3 or 4 levels (Table 1). Questions 1, 2, and 10 wereadapted from the NIOSH Generic Job Stress Questionnaire[39], and questions 6, 8, and 9 were adapted from theJob Content Questionnaire [40]. The other questions weredeveloped by the research team for this study. While thesequestions have been used in other studies [41–43], they havenot been validated. We were unable to include extensivebatteries of questions due to enrollment time limits and


Table 1: Descriptive statistics for the pooled population and jobphysical factors for the typical job on the right hand (𝑛 = 1834).

Mean ± SD or 𝑛 (%) Min–maxAge (years) 41.13 ± 11.34 18.0–72.0Female gender 1096 (59.8%)Body mass index (kg/m2) 28.67 ± 6.49 15.9–58.6Never smoke tobacco 1104 (60.2%)Diabetes mellitus 87 (4.7%)Thyroid problems 114 (6.2%)Total duration (%) 66.1 ± 22.4 0–100Forceful duration (%) 40.5 ± 31.3 0–100Total exertions (per minute) 22.2 ± 16.7 0–112.9Forceful exertions (per minute) 13.8 ± 15.7 0–111.3Hand activity level rating 3.4 ± 1.9 0–7.9Overall force (Borg rating) 2.3 ± 1.6 0–9Peak force (Borg rating) 2.7 ± 1.7 0–10Strain Index 6.7 ± 9.9 0.3–117.0TLV for HAL 0.64 ± 0.63 0.00–8.00

having participants excessively removed from productionjobs.

2.2. Job Physical Exposures. Individual data used to calculatethe TLV for HAL [44, 45] and SI [31, 44, 46] were collected bytrained ergonomics analysts who were blinded to symptomsand health data. Job physical raw data included (a) videotapesof tasks, (b) analyst peak hand force rating [47], (c) individualtask duration, and (d) length of work shift.

Videos were analyzed to extract data of analyst’s overallforce ratings, exertion durations, postures, and work speed.Exertion, duration, and repetition were also assessed directlyfrom recorded video of multiple cycles of each participant’stasks. Expert ergonomists who were specifically trained andstandardized viewed each video and quantified individualduration of exertions, repetition, and overall force ratings forboth hands of each worker for SI score calculations. Trainedergonomics analysts took video recordings and providedhand-specific peak force ratings (using the Borg CR-10 scale)for each task performed by each worker. Video recordingswere later analyzed in laboratory to quantify (i) Borg CR-10 force ratings for each sub-task, (ii) verbal anchor HALratings [48], (iii) total frequency of exertion, (iv) frequency offorceful exertions, (v) total percent duration of exertion, (vi)percent duration of forceful exertions, and (vii) posture andspeed of work used to calculate SI scores. Forceful exertionswere defined as those rated as “light” or greater on the BorgCR-10 scale (i.e., Borg CR-10 ≥ 2). Analysts were blinded tothe health and psychosocial status of the workers.

Exertion requirements measured included (i) verbalanchor scale for HAL rating [45, 48], (ii) counts of effortsper minute, and (iii) % duration of exertion [46]. Methods todetermine efforts per minute, % duration of exertion, workspeed, and posture were published previously [46].

2.3. SI and TLV for HAL Scoring and Components. TLV forHAL and SI were calculated for each task that a worker

performed. TLV for HAL scores were calculated using theACGIH method as follows: Score = [Analyst Peak ForceRating on Borg CR-10 Scale/(10 − HAL Rating)]. We treatedTLV for HAL score as a continuous variable. TLV forHAL was also categorized using the ACGIH prescribed cut-points: below the Action Limit (AL) (score < 0.56), betweenthe AL and Threshold Limit Value (TLV) (0.56 ≤ score ≤0.78), and above the TLV (score > 0.78). Calculation of SIscores followed prior published methods and incorporatedthe analyst’s overall force rating, counts of efforts/min, %duration of exertion, posture, work speed, and task duration[46]. First, SI was treated as a continuous variable. Then, SIscore was categorized into low risk (SI ≤ 6.1) and high risk (SI> 6.1) based on the most recent recommendation by Mooreet al. [49]. TLV for HAL and SI were calculated for each taskthat a worker performed. TLV forHAL scores were calculatedas follows: Score = [Analyst Peak Force Rating on Borg CR-10 Scale/(10 − HAL Rating)]. SI scores were calculated in themanner described byMoore and Garg [46] using total effortsper minute and total percent duration of exertion.

A large proportion of workers (𝑛 = 710, 38.7%)performed multiple tasks as part of their job. We defined“typical exposure” (i.e., exposure from the task the workerperformed for the largest percentage of a work shift) as beingrepresentative of theworker’s daily exposure. For comparativepurposes we also explored the alternative techniques of “peakexposure” (i.e., exposure from the most stressful task per-formed) and time-weighted-average (TWA) exposure fromall tasks performed during a work shift. Details of these jobphysical exposure summarization techniques are describedelsewhere [31, 44].

2.4. Statistical Analyses. Ordered logistic regression was per-formed to assess the risk between worker physical exposuresand psychosocial factors. All analyses were performed usingSAS 9.4 software (Cary, NC). Statistical significance was at𝑃 < 0.05. All models included age, gender, and body massindex (BMI) as potential confounders.

For interpretive purposes, we also calculated predictedprobabilities of participants being in a given psychosocialcategory per unit ofmeaningful change in a given job physicalexposure measure. Meaningful changes in physical exposurewere defined as 4 efforts per minute, 5% duration of exertion,1 Borg CR-10 unit of force, 0.1 units of TLV for HAL score,and 3 units of Strain Index score.

Analyses were treated as exploratory and thus no cor-rections were made to the models or results to account formultiple comparisons.

3. Results and Discussion

A total of 1834 participants were included in this pooledanalysis. Most (59.8%) were female (see Table 1) with a meanage of 41.1 years and mean BMI of 28.7 kg/m2. Most (60.2%)had never smoked tobacco and relatively few had beendiagnosed with diabetes mellitus (4.7%) or thyroid problems(6.2%). Job physical exposure measures for the typical jobtask were similar for both the left and right hands and


thus only right hand data and results are reported (Table 1).Frequency and percentage of the 10 psychosocial questionsassessed show reasonable distribution across this pooledsample of workers (Table 2).

Several associations between quantified job physicalexposures and psychosocial factors were identified (Table 3).Similarly, there were strong associations between age, gender,and BMI and all psychosocial measures except willingnessto take the job again and recommending the job to others.Both the TLV for HAL and the Strain Index were associatedwith job satisfaction, supervisor support, whether a workerwould recommend the job to someone else and how likely theworker would be to take the job again (𝑃 ≤ 0.05). The TLVforHAL additionally was associated with physical exhaustionafter work and whether the employer was thought to careabout the worker’s health and safety on the job (𝑃 ≤ 0.01).The only psychosocial factor associated with the Strain Indexbut not the TLV for HAL was mental exhaustion after work(𝑃 ≤ 0.01). Neither model showed association with generalhealth status, feelings of depression, or supervisor support(𝑃 > 0.17). Other measures of physical exposure similarlyshowed broad association with multiple psychosocial factors.In general, peak force and forceful duration were morestrongly associated with more psychosocial outcomes thanother exposure measures. Forceful duration of exertion wasassociated with (𝑃 ≤ 0.05) or tending towards associationwith (𝑃 ≤ 0.20) all psychosocial outcomes.

The directionality of most of the relationships betweenphysical exposure and psychosocial outcomes was as hypoth-esized, where an increase in job physical exposure measure(e.g., higher force, higher repetition, and higher duration ofexertion) was associated with a worsening of psychosocialresponse (e.g., more physical exhaustion, less job satisfaction,and less likely to take this job again). Exceptions were forgeneral health and mental exhaustion where increasing jobphysical exposures tended to be associated with better psy-chosocial responses (e.g., better general health, less mentalexhaustion).

For comparative purposes, analyses were performedevaluating relationships between both peak and TWA jobphysical exposure summarization techniques and the resultswere essentially identical to typical job physical exposuremeasures (data not shown).

Figure 1 represents estimates in change of likelihood for aworker of mean age and BMI to be in a worse psychosocialcategory per unit increase in job physical exposure measurefor the typical job as compared to the probability in thebest psychosocial category. This figure demonstrates bothdirectionality and magnitude of the relationships betweenSI or TLV for HAL and psychosocial factors. For example,consider TLV for HAL rated exposure and reporting beingphysically exhausted; for each 0.1 units increase in TLV forHAL there is a 0.17% increased probability that an averageworker will report seldom being exhausted, 0.41% increasedprobability they will report often being physically exhausted,and 0.15% increased probability they are reporting alwaysbeing exhausted as compared to those reporting never being

Table 2: Descriptive statistics for the psychosocial factors for thepooled population.

How is your general health compared topeople your own age1 better 263 (14.3%)2 631 (34.1%)3 743 (40.5%)4 worse 197 (10.7%)

How often do you feel down, blue, ordepressed1 never 505 (27.5%)2 1028 (56.1%)3 269 (14.7%)4 always 32 (1.7%)

Physically exhausted after work1 never 212 (11.6%)2 873 (47.6%)3 562 (30.6%)4 always 187 (10.2%)

Mentally exhausted after work1 never 474 (25.9%)2 882 (48.1%)3 397 (21.7%)4 always 81 (4.4%)

Get along with your coworkers1 always/often 947 (51.6%)2 745 (40.6%)3 hardly ever/never 142 (7.7%)

Job satisfaction1 satisfied 521 (28.4%)2 941 (51.3%)3 dissatisfied 372 (20.3%)

How often does your supervisordemonstrate appreciation for the workyou do1 always 1207 (65.8%)2 527 (28.7%)3 never 100 (5.5%)

How likely would you recommend yourjob to someone else1 strongly recommend 278 (15.2%)2 894 (48.8%)3 435 (23.7%)4 not recommend 227 (12.4%)

How likely would you take this job again1 very likely 521 (28.4%)2 721 (39.3%)3 429 (23.4%)4 unlikely 163 (8.9%)

My employer cares about my health andsafety on the job1 strongly agree 450 (24.5%)2 1140 (62.2%)3 176 (9.6%)4 strongly disagree 68 (3.7%)


Table3:Statisticalsig

nificance

ofordinallogisticregressio

nresults

analyzingrelationships

betweentypicaljob

physicalexpo

sure

measuresinther

ight

hand

andpsycho

socialfactorsa

fter

adjustmentfor

age,gend

er,and

body

massind

ex.

Expo

sure

General

health

Dow

n,blue,or

depressed

Physically

exhausted

Mentally

exhausted

Job

satisfaction

Superviso

rsupp

ort

Getalon

gwith

your

coworkers

Recommend

job

Take

thisjob

again

Employer

cares

Age

<0.001∗∗

<0.001∗∗

<0.001∗∗

0.007∗∗

0.00

4∗∗

0.033∗

<0.001∗∗

0.227

0.595

0.027∗

BMI

<0.001∗∗

<0.001∗∗

<0.001∗∗

0.008∗∗

0.00

4∗∗

0.033∗

<0.001∗∗

0.223

0.594

0.029∗

Gender

<0.001∗∗

<0.001∗∗

<0.001∗∗

0.007∗∗

<0.00

4∗∗

0.034∗

<0.001∗∗

0.232

0.005∗∗

0.026∗

Totalduration

0.00

6∗∗

0.121

0.459

<0.031∗∗

<0.001∗∗

0.028∗

0.067∗

<0.001∗∗<0.005∗∗

0.178

Forceful

duratio

n0.00

6∗∗

0.091

0.178

<0.001∗∗

<0.001∗∗

0.028∗

0.017∗

<0.001∗∗

<0.001∗∗

0.089

Totalrepetition

0.147

0.001∗∗

0.714

0.172

<0.001∗∗

0.456

0.46

80.096

0.001∗∗

0.295

Forceful

repetition

0.176

0.001∗∗

0.198

0.008∗∗

<0.001∗∗

0.383

0.518

0.003∗∗

<0.001∗∗

0.096

HAL

0.065

0.001∗∗

0.739

0.020∗

<0.001∗∗

0.031∗

0.101

0.010∗

<0.001∗∗

0.099

Overallforce

0.589

0.829

0.003∗∗

0.019∗

0.00

6∗∗

0.787

<0.001∗∗

<0.001∗∗

0.002∗∗

0.002∗∗

Peak

force

0.075

0.868

<0.001∗∗

0.013∗

<0.001∗∗

0.073

0.005∗∗

<0.001∗∗

<0.001∗∗

<0.001∗∗

Strain

index

0.681

0.166

0.402

0.002∗∗

0.002∗∗

0.228

0.617

0.001∗∗

0.012∗

0.769

TLVforH

AL

0.439

0.213

0.00

6∗∗

0.118

<0.001∗∗

0.292

0.031∗

<0.001∗∗

0.041∗

<0.001∗∗

∗0.05≥𝑃>0.01;∗∗0.01≥𝑃.


0.00%

0.30%

0.54%

0.25%

0.00

0.20

0.40

0.60

0.80

1.00

1 stronglyrecommend(reference)

2 3 4 notrecommend

Recommend job and TLV for HAL

(%)

0.00%

0.57%

0.95%

0.43%

0.00

0.20

0.40

0.60

0.80

1.00

(%)

1 stronglyrecommend(reference)

2 3 4 notrecommend

Recommend job and SI

(%)

0.00%

−0.79%−0.85%

−0.19%

−1.00

−0.80

−0.60

−0.40

−0.20

0.00

1 never(reference) 2 3 4 always

Mental exhaustion and SI

(%)

0.00%

−0.22% −0.22%

−0.05%

−1.00

−0.80

−0.60

−0.40

−0.20

0.00

1 never(reference) 2 3 4 always

Mental exhaustion and TLV for HAL

(%)

0.00%

0.90%

0.66%

0.00

0.20

0.40

0.60

0.80

1.00

1 satisfied(reference)

3 dissatisfied

Job satisfaction and SI

(%)

0.00%

0.54%0.44%

0.00

0.20

0.40

0.60

0.80

1.00

1 satisfied(reference)

3 dissatisfied

Job satisfaction and TLV for HAL

(%)

0.00%

0.17%

0.41%

0.15%

0.00

0.10

0.20

0.30

0.40

0.50

1 never(reference)

2 3 4 always

Physical exhaustion and TLV for HAL

(%)

0.00%

0.10%

0.22%

0.08%

0.00

0.10

0.20

0.30

0.40

0.50

1 never(reference)

2 3 4 always

Physical exhaustion and SI

Figure 1: Estimates in change of likelihood for a worker with mean age and BMI to be in a worse psychosocial category with a unit icreasein job physical exposure measure for the typical job as compared to the change in the best psychosocial category (unit change for SI = 3, unitchange for TLV for HAL = 0.1).


physically exhausted. When comparing differences betweenSI and TLV for HAL, the directionality and relative relation-ships are similar formost psychosocial outcomes. Differencesinmagnitude (𝑦-axis) between SI and TLV forHALmeasuresmight simply be the result of unit differences for estimatingprobabilities.

4. Discussion

The results of this study show a relatively consistent statisticalassociation between increased job physical exposure andworsening of psychosocial outcomes notwithstanding thenoteworthy exceptions of general health compared to othersand mental exhaustion after work which showed generallymore positive responses associated with higher physicalexposures. Many prior studies have evaluated relationshipsbetween psychosocial factors and WMSDs; however, this isthe first study that we are aware of to assess relationshipsbetween job physical factors and psychosocial outcomes.

While most associations are consistent, such as poorerresponses to job satisfaction, recommending job to oth-ers, and taking job again, as exposures increase, there area few associations that stand out as potentially unique.For example, perhaps unsurprisingly, perceived physicalexhaustion appears to be most strongly related to force,but not necessarily repetition. Conversely and somewhatunexpectedly, depression appears to be most strongly relatedto repetition, but not force. This might suggest that moremonotonous work somehow provokes depressive symptoms.Perhaps contradictorily, to the seemingly consistent asso-ciation between job dissatisfaction and increased physicalexposures, the tendency of workers to report relatively bettergeneral health and lessmental exhaustion with increased jobphysical exposures suggests that at leastmoderately strenuousjobsmay somehow be beneficial to one’s perceived well-being(if not job satisfaction).

It is important to note that while the statistical asso-ciations between job physical exposures and certain psy-chosocial factors appear very strong, the relative impacton probability of response is relatively modest (Figure 1).This implies that there are likely several factors, other thanphysical exposures, that influence the psychosocial state ofmanufacturing workers. Thus, psychosocial factors shouldcontinue to be studied as possible independent risk factorsfor occupational injuries and illnesses, such as CTS.

Only a few studies evaluating relationships between psy-chosocial factors and UEMSDs have been able to statisticallycontrol the potential confounder of job physical factors [11,15] or have created theoretical constructs that account forjob both physical factors and psychosocial factors in theetiological pathway for UEMSDs [12, 13]. The psychosocialfactors assessed in the literature have focused on bothwork-organizational (e.g., job pace, job control, and jobsatisfaction) factors and personal (e.g., depressive symptomsor anxiety) factors.The paper by Huang et al. theorized aboutthe potential causal pathways and relationships betweenjob physical factors, psychosocial factors, and health out-comes [14]. Several studies have found statistical relationshipsbetween different measures of psychosocial factors, while

statistically controlling for job physical exposures; however,there has not been an established relationship between jobphysical factors and psychosocial factors. To the best of ourknowledge, this is the first study to quantify the relationshipbetween these two domains.

Study strengths include a large, multicenter study includ-ingworkers from4 diverse states that used highly comparablestudy methods. Workers also were enrolled from a widediversity of occupations and spectrum of job physical factors.The broad range of job physical factors suggests the study isreasonably powered to detect relationships based on thosefactors. Data collection instruments used identical or nearlyidentical measures. Questionnaires, psychosocial measures,health status, and job measurements were obtained in allworkers, regardless of symptoms. The job measurementteams and health measurement teams were blinded to eachother.

Study limitations include the exploratory and cross-sectional nature of this study which limit the study tohypothesis generation regarding potential associations. Theworkers were mostly in manufacturing, which may limitextrapolations to other industrial sectors.The healthy workereffect may have had some impact, although the enrollmentsintentionally sought workers regardless of symptoms. Thenon-Gaussian distribution of the answers to the psychosocialfactors likely somewhat limits the power to detect effects,especially for feelings of depression and coworker support.The number of psychosocial factors is also somewhat limited,although generally more robust than prior reports. Addition-ally, not all psychosocial measures were validated.

5. Conclusion

These analyses demonstrate multiple relationships betweenjob physical exposure measures and psychosocial outcomesafter adjustment for age, BMI, and gender. Higher job phys-ical exposures appear to elicit consistently worse responsesto job satisfaction, willingness to take the job again, and rec-ommending the jobs to others. Depressive symptoms appearto be more strongly related to increasing repetition measuresalone, while perceived physical exhaustion appears to bemore strongly related to force measures alone. Conversely,higher physical exposure results in relatively better perceivedgeneral health and mental exhaustion, implying that at leastmoderately demanding work may have a positive psycho-logical effect. Ultimately, these findings should help futureresearchers as they attempt to quantify associations betweenpsychosocial factors and various occupational injuries andillnesses.

Disclosure

No co-author reported a direct financial interest in the resultsof the research supporting this paper.


The authors declare that they have no conflict of interests.


Acknowledgments

This study was funded, in part, by Grants from the NationalInstitute for Occupational Safety and Health (NIOSH/CDC)R01-OH009712, NIOSH Education and Research CenterTraining Grant T42/CCT810426-10. The authors acknowl-edge the hundreds of workers who volunteered to participatein these studies. They also acknowledge the many years ofwork by dozens of technicians, assistants, and other researchpersonnel from the research study groups that made thecollection of the data for this paper possible.

References

[1] Occupational Health Indicators for Washington State, 1997–2004, 2008, http://www.lni.wa.gov/safety/research/file/occ-health/occhealthreport2008.pdf.

[2] B. Silverstein, E. Welp, N. Nelson, and J. Kalat, “Claimsincidence of work-related disorders of the upper extremities:Washington State, 1987 through 1995,” American Journal ofPublic Health, vol. 88, no. 12, pp. 1827–1833, 1998.

[3] N. J. Anderson, D. K. Bonauto, and D. Adams, “Prioritizingindustries for occupational injury prevention and research inthe Services Sector in Washington State, 2002–2010,” Journal ofOccupational Medicine and Toxicology, vol. 9, no. 1, article 37,2014.

[4] S. Riviere, E. Penven, H. Cadeac-Birman, Y. Roquelaure, andM. Valenty, “Underreporting of musculoskeletal disorders in10 regions in France in 2009,” American Journal of IndustrialMedicine, vol. 57, no. 10, pp. 1174–1180, 2014.

[5] S. E. Wuellner and D. K. Bonauto, “Exploring the relationshipbetween employer recordkeeping and underreporting in theBLS Survey of Occupational Injuries and Illnesses,”The Ameri-can Journal of Industrial Medicine, vol. 57, no. 10, pp. 1133–1143,2014.

[6] J. T. Goodson,M. S. DeBerard, A. J.Wheeler, andA. L. Colledge,“Occupational and biopsychosocial risk factors for carpal tun-nel syndrome,” Journal of Occupational and EnvironmentalMedicine, vol. 56, no. 9, pp. 965–972, 2014.

[7] K. Harkapaa, “Psychosocial factors as predictors for earlyretirement in patients with chronic low back pain,” Journal ofPsychosomatic Research, vol. 36, no. 6, pp. 553–559, 1992.

[8] L. A. MacDonald, R. A. Karasek, L. Punnett, and T. Scharf,“Covariation between workplace physical and psychosocialstressors: evidence and implications for occupational healthresearch and prevention,” Ergonomics, vol. 44, no. 7, pp. 696–718, 2001.

[9] P. M. Bongers, C. R. De Winter, M. A. J. Kompier, and V. H.Hildebrandt, “Psychosocial factors at work and musculoskele-tal disease,” Scandinavian Journal of Work, Environment andHealth, vol. 19, no. 5, pp. 297–312, 1993.

[10] NIOSH,NIOSHSafety andHealthTopic: Stress atWork, NIOSH,2007, http://www.cdc.gov/niosh/topics/stress/.

[11] C. Harris-Adamson, E. A. Eisen, A. M. Dale et al., “Personaland Workplace psychosocial risk factors for carpal tunnel syn-drome: a pooled study cohort,”Occupational and EnvironmentalMedicine, vol. 70, no. 8, pp. 529–537, 2013.

[12] E. Bru, R. J. Mykletun, and S. Svebak, “Work-related stressand musculoskeletal pain among female hospital staff,”Work &Stress, vol. 10, no. 4, pp. 309–321, 1996.

[13] P. Carayon, M. J. Smith, and M. C. Haims, “Work organization,job stress, and work-relatedmusculoskeletal disorders,”HumanFactors, vol. 41, no. 4, pp. 644–663, 1999.

[14] G. D. Huang, M. Feuerstein, and S. L. Sauter, “Occupationalstress andwork-related upper extremity disorders: concepts andmodels,” American Journal of Industrial Medicine, vol. 41, no. 5,pp. 298–314, 2002.

[15] P.-C. Wang, D. M. Rempel, R. J. Harrison, J. Chan, and B. R.Ritz, “Work-organisational and personal factors associated withupper body musculoskeletal disorders among sewing machineoperators,” Occupational and Environmental Medicine, vol. 64,no. 12, pp. 806–813, 2007.

[16] C. Fernndez-de-las-Peas, A. I. de la Llave-Rincn, J. Fernndez-Carnero, M. L. Cuadrado, L. Arendt-Nielsen, and J. A. Pareja,“Bilateral widespread mechanical pain sensitivity in carpaltunnel syndrome: evidence of central processing in unilateralneuropathy,” Brain, vol. 132, no. 6, pp. 1472–1479, 2009.

[17] S. A. Lozano Calderon, A. Paiva, and D. Ring, “Patient satisfac-tion after open carpal tunnel release correlates with depression,”The Journal of Hand Surgery, vol. 33, no. 3, pp. 303–307, 2008.

[18] S. J. E. Becker, H. S. Makanji, and D. Ring, “Expected andactual improvement of symptoms with carpal tunnel release,”The Journal of Hand Surgery, vol. 37, no. 7, pp. 1324.e5–1329.e5,2012.

[19] J. K. Kim and Y. K. Kim, “Predictors of scar pain after opencarpal tunnel release,”The Journal of Hand Surgery, vol. 36, no.6, pp. 1042–1046, 2011.

[20] J. Cowan, H. Makanji, C. Mudgal, J. Jupiter, and D. Ring,“Determinants of return towork after carpal tunnel release,”TheJournal of Hand Surgery, vol. 37, no. 1, pp. 18–27, 2012.

[21] S. Das De, A.-M. Vranceanu, and D. C. Ring, “Contributionof kinesophobia and catastrophic thinking to upper-extremity-specific disability,” The Journal of Bone and Joint Surgery—American Volume, vol. 95, no. 1, pp. 76–81, 2013.

[22] D. Ring, J. Kadzielski, L. Fabian, D. Zurakowski, L. R. Mal-hotra, and J. B. Jupiter, “Self-reported upper extremity healthstatus correlates with depression,” The Journal of Bone & JointSurgery—American Volume, vol. 88, no. 9, pp. 1983–1988, 2006.

[23] C. Harris-Adamson, E. A. Eisen, A. M. Dale et al., “Personaland Workplace psychosocial risk factors for carpal tunnel syn-drome: a pooled study cohort,”Occupational and EnvironmentalMedicine, vol. 70, no. 8, pp. 529–537, 2013.

[24] C. Harris-Adamson, E. A. Eisen, A. M. Dale et al., “The impactof gender on personal, health and workplace psychosocial riskfactors for carpal tunnel syndrome: a pooled study cohort,”in Proceedings of the Human Factors and Ergonomics SocietyAnnual Meeting, vol. 57, pp. 911–914, September-October 2013.

[25] B. A. Silverstein, Z. J. Fan, D. K. Bonauto et al., “The naturalcourse of carpal tunnel syndrome in a working population,”Scandinavian Journal of Work, Environment and Health, vol. 36,no. 5, pp. 384–393, 2010.

[26] Y. Roquelaure, C. Ha, C. Rouillon et al., “Risk factors for upper-extremity musculoskeletal disorders in the working popula-tion,” Arthritis Care and Research, vol. 61, no. 10, pp. 1425–1434,2009.

[27] D. S. Bash and R. S. Farber, “An examination of self-reportedcarpal tunnel syndrome symptoms in hand therapists, protec-tive and corrective measures and job satisfaction,”Work, vol. 13,no. 2, pp. 75–82, 1999.

[28] A. Descatha, Y. Roquelaure, J.-F. Chastang, B. Evanoff, D. Cyr,and A. Leclerc, “Description of outcomes of upper-extremity


musculoskeletal disorders in workers highly exposed to repeti-tive work,” Journal of Hand Surgery, vol. 34, no. 5, pp. 890–895,2009.

[29] A. Leclerc, P. Franchi, M. F. Cristofari et al., “Carpal tunnelsyndrome and work organisation in repetitive work: a crosssectional study in France. Study Group on Repetitive Work,”Occupational and Environmental Medicine, vol. 55, no. 3, pp.180–187, 1998.

[30] R. Bonfiglioli, S. Mattioli, T. J. Armstrong et al., “Validationof the ACGIH TLV for hand activity level in the OCTOPUScohort: a two-year longitudinal study of carpal tunnel syn-drome,” Scandinavian Journal ofWork, Environment andHealth,vol. 39, no. 2, pp. 155–163, 2013.

[31] A. Garg, K. T. Hegmann, J. J. Wertsch et al., “The WISTAHhand study: a prospective cohort study of distal upper extremitymusculoskeletal disorders,” BMC Musculoskeletal Disorders,vol. 13, article 90, 2012.

[32] J. M. Kapellusch JM, F. E. Gerr, E. J. Malloy et al., “Exposure–response relationships for the ACGIH threshold limit value forhand-activity level: results from a pooled data study of carpaltunnel syndrome,” Scandinavian Journal of Work, Environment& Health, vol. 40, no. 6, pp. 610–620, 2014.

[33] T.Armstrong,A.M.Dale, A. Franzblau, andB.A. Evanoff, “Riskfactors for carpal tunnel syndrome andmedian neuropathy in aworking population,” Journal of Occupational and Environmen-tal Medicine, vol. 50, no. 12, pp. 1355–1364, 2008.

[34] R. A. Werner, A. Franzblau, N. Gell et al., “Prevalence of upperextremity symptoms and disorders among dental and dentalhygiene students,” Journal of the California Dental Association,vol. 33, no. 2, pp. 123–131, 2005.

[35] B. Evanoff, A. M. Dale, E. Deych, D. Ryan, and A. Franzblau,“Risk factors for incident carpal tunnel syndrome: results of aprospective cohort study of newly-hired workers,”Work, vol. 41,no. 1, pp. 4450–4452, 2012.

[36] N. Gell, R. A. Werner, A. Franzblau, S. S. Ulin, and T. J.Armstrong, “A longitudinal study of industrial and clericalworkers: incidence of carpal tunnel syndrome and assessmentof risk factors,” Journal of Occupational Rehabilitation, vol. 15,no. 1, pp. 47–55, 2005.

[37] A. Garg, J. M. Kapellusch, K. T. Hegmann et al., “The StrainIndex and TLV for HAL: risk of lateral epicondylitis in aprospective cohort,” American Journal of Industrial Medicine,vol. 57, no. 3, pp. 286–302, 2014.

[38] S. Bao and B. Silverstein, “Estimation of hand force inergonomic job evaluations,” Ergonomics, vol. 48, no. 3, pp. 288–301, 2005.

[39] J. J. Hurrell Jr. and M. A. McLaney, “Exposure to job stress: anew psychometric instrument,” Scandinavian Journal of Work,Environment and Health, vol. 14, no. 1, pp. 27–28, 1988.

[40] R. Karasek, C. Brisson, N. Kawakami, I. Houtman, P. Bongers,and B. Amick, “The Job Content Questionnaire (JCQ): aninstrument for internationally comparative assessments of psy-chosocial job characteristics,” Journal of Occupational HealthPsychology, vol. 3, no. 4, pp. 322–355, 1998.

[41] A. Garg, K. T. Hegmann, J. S. Moore et al., “Study protocoltitle: a prospective cohort study of low back pain,” BMCMusculoskeletal Disorders, vol. 14, article 84, 2013.

[42] R. M. Shorti, A. S. Merryweather, M. S. Thiese, J. Kapellusch,A. Garg, and K. T. Hegmann, “Fall risk factors for commercialtruck drivers,” Journal of Ergonomics, vol. 9, pp. 1–9, 2014.

[43] M. S. Thiese, M. Hughes, and J. Biggs, “Electrical stimulationfor chronic non-specific low back pain in a working-age

population: a 12-week double blinded randomized controlledtrial,” BMCMusculoskeletal Disorders, vol. 14, article 117, 2013.

[44] A. Garg, J. Kapellusch, K. Hegmann et al., “The Strain Index(SI) and Threshold Limit Value (TLV) for Hand Activity Level(HAL): risk of carpal tunnel syndrome (CTS) in a prospectivecohort,” Ergonomics, vol. 55, no. 4, pp. 396–414, 2012.

[45] American Conference of Governmental Industrial Hygienists,Hand Activity Level TLV, ACGIH, Cincinnati, Ohio, USA, 2001.

[46] J. S. Moore and A. Garg, “The strain index: a proposed methodto analyze jobs for risk of distal upper extremity disorders,”American Industrial Hygiene Association Journal, vol. 56, no. 5,pp. 443–458, 1995.

[47] G. A. V. Borg, “Psychophysical bases of perceived exertion,”Medicine and Science in Sports and Exercise, vol. 14, no. 5, pp.377–381, 1982.

[48] W. A. Latko, T. J. Armstrong, J. A. Foulke, G. D. Herrin, R. A.Rabourn, and S. S. Ulin, “Development and evaluation of anobservational method for assessing repetition in hand tasks,”American Industrial Hygiene Association Journal, vol. 58, no. 4,pp. 278–285, 1997.

[49] J. S. Moore, G. A. Vos, J. P. Stephens, E. Stevens, and A. Garg,“The validity and reliability of the Strain Index,” in Proceedingsof the 16thWorld Congress on Ergonomics (IEA ’06), Maastricht,The Netherlands, August 2006.

Research ArticleModels of Workplace Incivility: The Relationships to InstigatedIncivility and Negative Outcomes

Kristoffer Holm, Eva Torkelson, and Martin Bäckström

Department of Psychology, Lund University, P.O. Box 213, 221 00 Lund, Sweden

Correspondence should be addressed to Kristoffer Holm; [email protected]

Received 23 January 2015; Revised 21 April 2015; Accepted 27 April 2015


Copyright © 2015 Kristoffer Holm et al.This is an open access article distributed under the Creative CommonsAttribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The aim of the study was to investigate workplace incivility as a social process, examining its components and relationships to bothinstigated incivility and negative outcomes in the form of well-being, job satisfaction, turnover intentions, and sleeping problems.The different components of incivility that were examined were experienced and witnessed incivility from coworkers as well assupervisors. In addition, the organizational factors, social support, control, and job demands, were included in the models. A totalof 2871 (2058 women and 813men) employees who were connected to the Swedish Hotel and RestaurantWorkers Union completedan online questionnaire. Overall, the results from structural equation modelling indicate that whereas instigated incivility to alarge extent was explained by witnessing coworker incivility, negative outcomes were to a high degree explained by experiencedsupervisor incivility via mediation through perceived low social support, low control, and high job demands. Unexpectedly, therelationships between incivility (experienced coworker and supervisor incivility, as well as witnessed supervisor incivility) andinstigated incivility were moderated by perceived high control and high social support. The results highlight the importance ofincluding different components of workplace incivility and organizational factors in future studies of the area.

1. Introduction

The aim of the present study was to explore workplaceincivility as a social process, including experienced as well aswitnessed incivility from coworkers and supervisors and itsrelationships to instigated incivility and negative outcomesin the form of well-being, job satisfaction, turnover inten-tions, and sleeping problems. The goal was to create com-prehensive models including direct relationships betweenworkplace incivility and its outcomes, as well as mediationand moderation of organizational factors. This adds to thecurrent literature through including different componentsof workplace incivility as well as organizational factors inthe same models to explain instigated incivility and negativeoutcomes. Workplace incivility has been defined as “. . . low-intensity deviant behavior with ambiguous intent to harm thetarget, in violation of workplace norms for mutual respect.Uncivil behaviors are characteristically rude and discourte-ous, displaying a lack of regard for others” [1]. Incivility, asa covert form of aggression, demarcates from other formsof overt workplace aggression in that it can be ambiguousand of lower intensity and does not necessarily need to be

intended to harm [2]. Despite this, incivility has been equatedto the severity of workplace bullying on the outcomes ofjob satisfaction and of turnover intentions [3]. Examples ofsuch covert behaviours are rude looks or ignoring someone,compared to overt behaviors like yelling [3].

Beyond experienced incivility, research has also beenrequested on perspectives focusing on the bystanders andperpetrators as well as the organizational context as compo-nents of the incivility process [4].

Studies have approached the social process of incivility,exploring it as a group-level phenomenon [1, 5]. Anderssonand Pearson [1] raise the issue of how incivility may manifestin the form of a reciprocal social process between involvedindividuals. The authors theorized about a negative spiral,where incivility can create escalating responses of grow-ing workplace aggression nourishing interpersonal conflicts.Further research has since supported this notion, indicatingthat the destructive spiral of workplace incivility, may be abuilding block in a negative work environment [6]. Beingtargeted by incivility has been shown to lead to negativeemotions that subsequently relate to aggression [7].


http://dx.doi.org/10.1155/2015/920239


When investigating the escalation of workplace aggres-sion, Taylor and Kluemper [8] reported findings additionallysupporting the relationship between perceived incivility andworkplace aggression, when incivility is seen as a mediatorbetween role stress and aggression. A stressful environmentwould thus induce higher ratings of instigated incivility, lead-ing to further reciprocal behaviours, resulting in increasedaggression. Some scholars have viewed incivility as a stressor(e.g., [5, 9]). In relation to stress and strain research KarasekandTheorell’s [10] demand-control-support (DCS)model hasbeen often applied in the literature of occupational healthpsychology. Earlier research [11] on the onset of bullying,as an overt form of aggression, included variables from theDCS model. Thus, it is interesting to include organizationalfactors from the model in the investigation of a covert formof aggression such as workplace incivility.

ConsideringAndersson and Pearson’s [1] reasoning aboutan uncivil spiral with “tit for tat” responses, the self-sustainingnature of such a spiral highlights the risk of instigatedincivility as an outcome, related to either experienced orwitnessed incivility in the workplace. In an interview-basedstudy by Pearson et al. [2], it was found that witnesses toincivility modelled their behaviour after their observations,retaliating uncivil acts. In line with this, Ferguson and Barry[12] reported that individuals in highly cohesive groups weremore likely to adopt uncivil behaviour if witnessing it. Inthe present study we investigated if experienced and wit-nessed incivility is related to instigated incivility and negativeoutcomes in models including four organizational factors,social support fromcoworkers, social support from superiors,control, and job demands. In the study, we tested bothdirect relations of experienced and witnessed incivility, aswell as mediation andmoderation of organizational variablestowards the outcomes. Thus, the first hypothesis tested was:

experienced and witnessed workplace incivility, fromcoworker or supervisor, is directly related to insti-gated incivility.

In the field of workplace incivility, more research on possiblemediators has been requested [13]. Negative emotions, how-ever, have been shown to mediate the relationship betweencoworker incivility and increased deviant behaviour [14].Schilpzand et al. [15] argue that most studies have not investi-gated the mediating mechanisms for why certain antecedentconstructs would lead to incivility. In addition, not muchwork has been conducted on organizational factors suchas job demands, control, and social support as mediatorsof workplace incivility. Testing possible mediation of thesefactors would be an addition to the field. Thus, the secondhypothesis was:

organizational factors (social support from coworker,social support from supervisor, control, and jobdemands) mediate the relationships between expe-rienced and witnessed workplace incivility (fromcoworker or supervisor) and instigated incivility.

Control has previously been shown to buffer effects of jobdemands on being targeted by bullying in the workplace

[11], and psychosocial factors have also been approachedas moderators in the relationship between incivility andinstigated counterproductive work behaviour [14]. In linewith this, the third hypothesis was:

organizational factors (social support from coworker,social support from supervisor, control, and jobdemands) moderate the relationships between expe-rienced and witnessed workplace incivility (fromcoworker or supervisor) and instigated incivility.

A high level of incivility has been linked to a numberof negative outcomes. In the present study, we focus onnegative outcomes in the form of low well-being, low jobsatisfaction, turnover intentions, and sleeping problems. Inci-vility is negatively related to both mental and physical well-being [9, 16]. Studies also consistently report that individualssubjected to workplace incivility, from both a target andan instigator perspective, experience lower job satisfaction[16, 17]. Lim et al. [9] found that incivility impact the entireorganization in form of lower levels of job satisfaction andmental health, even when controlling for job stress. Therelationship between job satisfaction and witnessed incivilityhas since been supported [18].

Being the victim of uncivil behaviour has been relateddirectly to turnover intentions [6, 19] and incivility froma supervisor has shown to be stronger related to turnoverintentions than coworker incivility [20].

Moreover, having troubles with sleep has previously beenshown to be strongly related to other types of workplaceaggression, such as bullying [21, 22]. Similarly, both expe-rienced and witnessed bullying has been studied, wherewitnessing bullying relates to detrimental outcomes [23, 24].As follows to this, the fourth hypothesis was:

experienced and witnessed workplace incivility, fromcoworker or supervisor, is directly related to employ-ees’ negative outcomes (well-being, job satisfaction,and turnover intentions, aswell as sleeping problems).

Emotional and organizational support has previously beenfound to mediate the effects between experienced workplaceincivility and negative outcomes [25]. Social and organiza-tional support has also been approached as both a mediatorand a moderator in the research on workplace bullying andnegative outcomes [23, 26]. The DCS model, concerningthe variables of support, control and job demands, is wellestablished in the workplace literature and has previouslybeen tied to well-being [27]. In light of this, it servesimportant to include these variables in the present study.Thus, the fifth hypothesis was:

organizational factors (social support from coworker,social support from supervisor, control, and jobdemands) mediate the relationships between expe-rienced and witnessed workplace incivility (fromcoworker or supervisor) and negative outcomes (well-being, job satisfaction, turnover intentions, and sleep-ing problems).


Additionally, we tested a sixth hypothesis:

organizational factors (social support from coworker,social support from supervisor, control, and jobdemands) moderate the relationships between expe-rienced and witnessed workplace incivility (fromcoworker or supervisor) and negative outcomes (well-being, job satisfaction, turnover intentions, and sleep-ing problems).

The population of the present study consisted of individualsemployed in the hotel and restaurant sector, representing thehospitality industry. Previous work has shown this sector tobe particularly subjected to workplace bullying with negativeoutcomes related to it and has been suggested to be a sectorwith an aggressive climate [28, 29], making it a suitablepopulation for the investigation of workplace incivility.


2.1. Participants. An online survey was completed by 2871(2058 women and 813 men) members of the Swedish Hoteland Restaurant Workers Union. Participants’ ages rangedfrom 16 to 72 and the mean age was 36.6 years (SD =12.3). The respondents had been at their current workplaceon average for 6.6 years (SD = 7.2), 410 (14%) employeeshad a managerial or executive position, a majority 2291(79.8%)were born in Sweden, and 2273 (79%)were in perma-nent employment. Of the sample, 1188 (41.4%) were servicepersonnel such as waiters/waitresses and receptionists, 1076(37.5%) kitchen personnel, 45 (12%) facility workers, and 264(9.2%) belonged to some other category of staff.

2.2. Measures

2.2.1.Workplace Incivility. Experienced incivility from super-visor and coworker was measured by the 7-item WorkplaceIncivility Scale, [19] which was translated into Swedish [30].The scale assessed the frequency of perceived incivility in thelast year, which is a shorter time frame than originally usedby Cortina et al. [19].

The scale was modified to measure witnessed workplaceincivility, using different stems for the same 7 items, inaccordance with Ferguson and Barry’s [12] adaptation of theInterpersonal Deviance Scale [31]. Employees were askedto rate how often they have witnessed each of the sevenbehaviour items in the scale. Example questionswere “Duringthe past year while employed in the current organization,have you been in a situation where you have observed anyof your superiors: Making demeaning or derogatory remarksabout others?” The perception of supervisors and coworkerswas rated separately as advocated by Smith et al. [32], forexperienced and witnessed incivility.

Consistent with Blau and Andersson [33], the scale wasmodified to measure instigated workplace incivility. Employ-ees were asked to rate their own behaviour for each of the 7items in the scale.The response alternatives for all of the inci-vility measures ranged from 0 (never) to 4 (most of the time).Cronbach’s alphas for experienced incivility from supervisor

was .94 and from coworker .92, witnessed incivility fromsupervisor .96 and coworker .95, and instigated incivility .83.

2.2.2. Organizational Factors. Subscales from the rigorouslytested and applied Copenhagen Psychosocial Questionnaire(COPSOQ II) [34] in a Swedish variant [35]was used to assesspsychosocial factors at work.The subscales were job demands(four items), social support from supervisor (three items),social support from colleagues (three items), and control(four items) having Cronbach’s alphas of .80, .90, .80, and.81, respectively. Response alternatives on these scales rangedfrom 0 (never/hardly ever) to 4 (always).

2.2.3. Negative Outcomes. Job satisfaction was measured byfour items from the COPSOQ subscale. Responses rangedfrom 1 (very unsatisfied) to 4 (very satisfied). Cronbach’s alphawas .87. Sleeping troubles were captured by four items ofperceived sleeping troubles in the last four weeks from theCOPSOQ subscale. Responses ranged from 0 (not at all) to 4(all the time). Cronbach’s alpha was .87.Three items were usedto measure turnover intentions among the employees [20].Responses ranged from 0 (I strongly disagree) to 4 (I stronglyagree). Cronbach’s alpha was .79.Well-being wasmeasured bythe WHO-Five Well-Being Index [36]. A Swedish version ofthe instrument was used [37]. The scale consisted of 5 items,ranging from 0 (never) to 5 (all of the time). Cronbach’s alphawas .87.

2.2.4. Demographic Variables. Demographic questions con-cerned gender, age, supervisor/nonsupervisor, born in Swe-den, temporary employment, and length of employment.

2.3. Procedure and Ethical Considerations. A link to theonline-based survey was presented in a letter directed tothe participant with information about the study along withcontact information. Participants were free to withdraw atany point. Completing the study was considered consentingto participation.

The survey with the cover letter was forwarded tothe Hotel and Restaurant Workers Union, where it wasdistributed by e-mail through the membership registries.After one week the survey was reissued. As the survey hadreached 6800 individuals, roughly 1600 had responded tothe questionnaire. As additional reminders went out, around1200 more members participated, finally resulting in 2871completed surveys. Ethical approval was granted through theSwedish Central Ethical Review Board.

2.4. Strategy of Analysis. To test our hypotheses, we createdtwo structural models, one for hypotheses 1-2 concerninginstigated incivility and one for hypotheses 4-5 concerningthe negative outcomes of experienced andwitnessed incivilityin the workplace (it was found that these latent variablescorrelated −.329, but in the models, when other variableswere included, the correlation was insignificant, suggestingthat negative outcomes did not add uniquely to instigatedincivility when the organizational variables were included,and therefore we decided to make separate models). The


incivility variables were estimated as latent variables with theitems of each scale as observed variables. Negative outcomeswere a latent variable measured by items from the scalesjob satisfaction, sleeping problems, turnover intentions, andwell-being as observed variables.The organizational conceptssocial support from coworkers and supervisor, control, andjob demands were defined as latent variables measured byitems from their respective scales.

Since many of the measurement models were based oncategorical variables, we estimated them with MPLUS v.7.11 using the categorical option, estimating with WeightedLeast Square with mean and variance adjusted 𝜒2 values.This estimation method has been suggested to performwell when variables are categorical. The only exception wasthe negative outcome measurement model, consisting ofsummarized scale values and not single items, where weused theMaximumLikelihood Estimator.The fit indices usedwere CFI values above .95 representing good fit [38] andRMSEA values below .05 representing excellent fit, and valuesbelow .07 representing acceptable fit [39], in themeasurementmodels we primarily relied on the CFI since the RMSEA wasvery unstable and CFI was very close to 1.0, representingalmost perfect model fit.

We tested the measurement models of all the latentvariables and found that almost all had an excellent fit tothe data (CFI > .98), the exception being the model forthe negative outcomes. In that model CFI was .92, but afterthe addition of one error correlation between well-beingand job satisfaction the fit was excellent. Loadings for theincivility dimensions were high for all latent variables, in therange between .70 and .95, with a mean loading of .89. Withthese very goodmeasurement models we were confident thatmisfit in the structural model could not be attributed to badmeasurement models.

The proposed research model for hypotheses 1 and 2 isdepicted in Figure 1 and the model for hypothesis 4 and 5 inFigure 3. The only difference is that the dependent variableis instigated incivility in Figure 1 and negative outcomes inFigure 3. We first tested the total fit for each model and afterthat, based on our hypotheses, we tested for direct effectsbetween the variables in each of the two models as well asmediation of the organizational factors. Since the sample waslarge the hypotheses were tested with an alpha level of .005.

To test the third hypothesis related to the moderationof the organizational variables in the relationships betweenincivility (experienced and witnessed) and instigated inci-vility, a number of latent interaction models were estimated[40]. It was not possible to test all of the interactions inthe same model; therefore, the interaction models weresimpler in that they consisted of two independent latentinteraction variables together with an estimate of their latentvariable interaction. Latent interaction variables take a lotof computational resources, making it almost impossible totest more complicated models in MPLUS (to estimate latentinteraction variables, mathematical integration is necessary.Tomake themodel less computationally demanding,MPLUShas a procedure that makes this more effective, based onMonte Carlo methods. This method was used in all thepresented models including an interaction).

EIC

EIS

WIC

WIS

C

SSCSSS

JD

II0.483

0.592

0.873

0.485

0.74

0.606

−0.523

−0.417

0.302

0.499

0.685

0.42

0.159

0.098

0.245

0.433

−0.113

−0.171

0.481

−0.058

Figure 1: Final structural model showing standardized relation-ships between the latent variables EIC, experienced incivility fromcoworker; EIS, experienced incivility from supervisor; WIS, wit-nessed incivility from coworker; WIS, witnessed incivility fromsupervisor; C, control; JD, job demands; SSC, social support fromcoworker; SSS social support from supervisor; and II, instigatedincivility. Figures in italics = correlations, figures in plain text = paths(𝑁 = 2132). All figures are significant at the 𝑝 < .005 level, exceptthe relationship between EIC and II, which was on the border ofsignificance.

3. Results

3.1. Descriptive Results. Table 1 displays the correlations,means, and standard deviations for the latent variables in themodel.

3.2. Instigated Incivility. The first hypothesis concernedwhether experienced and witnessed workplace incivility wasrelated to instigated incivility in the workplace. Figure 1shows the model used for investigating this hypothesis.The full model, including all paths from experienced andwitnessed incivility to acting uncivilly, from experiencedand witnessed incivility to the organizational variables andfrom organizational variables to acting uncivilly revealeda very good fit, 𝜒2(1090) = 7601.9, RMSEA = .053, andCFI = .974. In thismodel a number of pathswere insignificant(p > .10), those paths were set to zero and the model wasreestimated. Results revealed an even better fit, 𝜒2(1107) =6614.3, RMSEA = .048, and CFI = .978. In relation to the firsthypothesis the most important paths were the ones endingat the instigated incivility latent variable. Three variableswere found to have significant paths: the largest was fromwitnessed incivility from coworkers (𝛽 = .433, p < .001),the second largest from experienced incivility from superior(𝛽 = .245; p < .001). Control (𝛽 = .159, p < .001) alsohad a significant path. On the border of significance wasexperienced incivility from coworkers (𝛽 = .098, 𝑝 = .007).We tested whether witnessed incivility from coworkers hada unique relation to instigated incivility using the MPLUSDIFFTEST. It was found that deleting this path from themodel decreased the fit significantly, Δ𝜒2(1) = 144.7, indi-cating a unique relationship between witnessed coworkerincivility and instigated incivility.


Table 1: Descriptives and correlations (𝜌) of latent variables included in the models (𝑁 = 2132).

1 2 3 4 5 6 7 8 9Experienced coworker incivility (1)Experienced supervisor incivility (2) .50Witnessed coworker incivility (3) .60 .40Witnessed supervisor incivility (4) .41 .77 .50Control (5) −.21 −.32 −.14 −.26

Job demands (6) .24 .33 .22 .31 −.17

Support coworker (7) −.41 −.26 −.29 −.20 .26 −.20

Support supervisor (8) −.33 −.61 −.28 −.54 .43 −.31 .43Negative outcomes (9) −.40 −.55 −.35 −.51 .41 −.38 .33 .56Instigated incivility (10) .39 .34 .44 .37 −.04 −.17 −.18 −.18 −.23

M (of scales) 6.87 7.71 8.44 8.40 12.72 3.02 6.02 6.33 6.80SD (of scales) 6.18 7.25 6.69 7.40 5.73 3.27 3.87 3.68 2.83Note. All correlations except the relationship between control and instigated incivility (−.04) were significant at p < .001.

Hypothesis 2 concerned mediation effects of the orga-nizational factors. Since perceived control was the onlyorganizational variable that revealed a significant path to theinstigated incivility latent variable in themodel, and incivilityfrom superior was the only variable with a significant pathto perceived control (𝛽 = −.404, 𝑝 < .001), we only testedmediation effects through this path. It was found that theindirect relationship was significant (standardized specificindirect effect was 𝛽 = −.066; 𝑝 < .001). The total effect (𝛽= .179) was slightly lower than the direct path (𝛽 = .245),suggesting that the direct relationship between control andinstigated incivility was suppressed (the indirect and directeffects were also estimated based onMPLUS bootstrap. Using1000 bootstraps the 99.5% bias corrected CI [−0.099, −0.034]for the standardized indirect effect, and [0.167, 0.323] forthe standardized direct effect). To summarize, supervisorincivility was found to predict instigated incivility throughperceived low control. In other words, perceived control hadan indirect effect on the relationship between experiencedsupervisor incivility and instigated incivility.

Next the moderation models related to hypothesis threewere tested. Social support was found to moderate therelationship between experienced and instigated incivility(see the top two panels of Figure 2). Social support fromcoworkers interacted with experienced incivility from co-workers. Participants high in both these variables tendedto report relatively higher instigated incivility (𝛽 = .148,p < .001; coefficients are raw, suggesting that instigatedincivility increases by .148 when the product of experiencedincivility and social support increases with 1). Social supportfrom supervisor interacted with experienced incivility fromsuperiors. Participants reporting higher levels of incivilityfrom their supervisors togetherwithmore support from themalso reported more instigated incivility (𝛽 = .081). Perceivedcontrol (see middle panel of Figure 2) moderated both therelationship between experienced incivility from coworkers(𝛽 = .210, 𝑝 < .001) and instigated incivility, as well as therelationship between experienced incivility from supervisors(𝛽 = .093, 𝑝 < .001) and instigated incivility.This suggest thatsubjects who perceive control and at the same time report

higher levels of experienced incivility have a tendency toreport higher levels of instigated incivility.

In addition, social support from supervisor and controlmoderated the relationship between witnessed supervisorincivility and instigated incivility (see the bottom two panelsof Figure 2). Subjects who had witnessed more incivilityfrom their superiors, who also reported relatively moresupport from their superiors, tended to report higher levelsof instigated incivility (𝛽 = .062, p < .001). Also, subjectswho reported having witnessed more incivility from theirsuperiors and experiencing higher level of control reportedhigher levels of instigated incivility (𝛽 = .077, p < .001).It is important to note that the organizational variablesmoderate the relationships between incivility and instigatedincivility. This suggests that instigated incivility is reportedby participants who describe their organization as relativelyhigh in incivility but at the same time perceive that they havesupport and/or control. Having experienced incivility fromsuperiors and coworkers and witnessed supervisor incivilityseems to increase the amount of instigated incivility theparticipants report when the support or control is perceivedas high.

3.3. Negative Outcomes. Figure 3 shows the basic model usedwhen testing hypotheses 4-5 about how experienced andwitnessed incivility is related to negative outcomes (well-being, job satisfaction, turnover intentions, and sleepingproblems). The full model, including all paths from the fourlatent incivility variables to the negative outcomes latentvariable, with the organizational variables in the middle, hada good fit to the data, 𝜒2(953) = 8350.0, RMSEA = .060,and CFI = .966. A model where all paths with standardizedcoefficients that were insignificant were set to zero hadan even better fit, 𝜒2(968) = 6639.0, RMSEA = .052, andCFI = .974. In relation to hypothesis 4, two incivility pathswere significantly related to negative outcomes, experiencedincivility from coworker (𝛽 = −.192, p < .001) and witnessedincivility from superiors (𝛽 = −.199, p < .001). Perceivedcontrol (𝛽 = .272, p < .001), job demands (𝛽 = −.238,p < .001), and social support (𝛽 = .250, p < .001) from


0.1

0.3

0.5SSC low SSC high

Late

nt in

stig

ated

inci

vilit

y

EIC lowEIC high

−0.5

−0.3

−0.1

(a) EIC × SSC

0.1

0.3

0.5SSS low SSS high

Late

nt in

stig

ated

inci

vilit

y

EIS lowEIC high

−0.5

−0.3

−0.1

(b) EIS × SSS

0.1

0.3

0.5C low C high

Late

nt in

stig

ated

inci

vilit

y

EIC lowEIC high

−0.5

−0.3

−0.1

(c) EIC × C

0.1

0.3

0.5C low C high

Late

nt in

stig

ated

inci

vilit

y

EIS lowEIS high

−0.5

−0.3

−0.1

(d) EIS × C

0.1

0.3

0.5SSS low SSS high

Late

nt in

stig

ated

inci

vilit

y

WIS lowWIS high

−0.5

−0.3

−0.1

(e) WIS × SSS

0.1

0.3

0.5C low C high

Late

nt in

stig

ated

inci

vilit

y

WIS lowWIS high

−0.5

−0.3

−0.1

(f) WIS × C

Figure 2:The significant interactions between the latent variables (EIC; experienced incivility from coworker; EIS experienced incivility fromsupervisor; WIC, witnessed incivility from coworker; WIC, witnessed incivility from coworker; WIS, witnessed incivility from supervisor;SSC, social support from coworker; SSS, social support from supervisor; and C, control) in hypothesis 3 on instigated incivility (𝑁 = 2132).

supervisor also had significant paths to negative outcomes.Experienced incivility from supervisor had significant pathsto job demands (𝛽 = .477, p < .001), control (𝛽 = −.416, p <.001), and social support from supervisor (𝛽 = −.687, p <.001). Experienced incivility from coworker had a significantpath to social support from coworker (𝛽 = −.522, p < .001),but social support from coworkers did not have a significantpath to negative outcomes. Witnessed incivility did not haveany significant paths to any of the organizational variables inthe model.

Hypothesis 5 concerned the possible mediation of orga-nizational variables on the relationships between incivilityand negative outcomes. Possible mediation effects were onlytested for experienced incivility from supervisor. This wasthe only incivility variable correlating with the organizationalvariables that also correlated with negative outcomes. Thetotal effect from experienced incivility from superiors tonegative outcomes was (𝛽 = −.399, p < .001). We firsttested a model setting the direct path between experiencedincivility from supervisor and negative outcomes to zero. In


EIC

EIS

WIC

WIS

C

SSCSSS

JD

NO0.483

0.592

0.873

0.486

0.74

0.606

−0.522

−0.416

0.306

0.501

−0.687

0.412

0.272

−0.174−0.192

−0.115

−0.199

0.477

0.25

−0.071

−0.238

Figure 3: Final structural model showing standardized relation-ships between the latent variables EIC, experienced incivility fromcoworker; EIS, experienced incivility from supervisor; WIS, wit-nessed incivility from coworker; WIS, witnessed incivility fromsupervisor; C, control; JD, job demands; SSC, social support fromcoworker; SSS social support from supervisor; and NO, Negativeoutcomes. Figures in italic = correlations, figures in plain text = paths(𝑁 = 2132). All figures are significant at the 𝑝 < .005 level.

this model, all indirect effects from experienced incivilityfrom supervisor to negative outcomes were significant; jobdemands (𝛽 = −.113, p < .001), control (𝛽 = −.113, p < .001),and social support from supervisor (𝛽=−.172, p< .001).Next,we tested whether estimating the direct path significantlydecreased model fit. Testing the difference in 𝜒2 with the socalled “DIFFTEST” inMPLUS suggested that this addition tothe model did not increase fit, Δ𝜒2(1) = 1.001, 𝑝 > .005 (thetotal indirect effect based on MPLUS bootstrap (1000) had99.5% CI [−0.467, −0.330], all CI of the single indirect effectsexcluded zero, and the direct effect was 0.00). To summarize,participants’ reported negative outcomes was directly relatedto their reported experienced incivility from coworker andwitnessed incivility from supervisor. Experienced incivilityfrom supervisor had the strongest relation to negative out-comes, but there were no direct effects when the indirecteffects through the mediating organizational variables wereincluded in the model.

In relation to the hypothesis 6, the interactionmodels thatwere tested for instigated incivility were used with negativeoutcomes as dependent variable, but none of the 16 possibleinteractions were found to be significant.

4. Discussion

The aim of the study was to examine workplace incivilityas a social process, including experienced, witnessed andinstigated incivility, and negative outcomes of workplaceincivility. The first hypothesis concerning the relationshipsbetween experienced and witnessed incivility, and instigatinguncivilized acts were partly supported. The study showed astrong and unique relationship between witnessing incivilityfrom coworkers and acting uncivilized, and to some extentbeing targeted by incivility from a supervisor was also relatedto instigating incivility. The results are in line with earlier

studies by Robinson et al. [41], who found that merely beingin a climate of deviance was shown to impact individualdeviant behaviour. The present study shows a similar patternfor workplace incivility. In the estimated model witnessedincivility from a supervisor did not have a unique significantrelationship to instigated incivility which is in line with theresearch of Ferguson and Barry [12].They found that employ-ees adapt to observed behaviours of their colleagues ratherthan their supervisors. The results expand on the currentliterature as to include how merely witnessing incivility canimpact the individual’s behaviour.

The suggestionmade by Estes andWang [4] that incivilityshould be studied in an organizational context was inves-tigated in the second hypothesis. Perceived lower controlmediated the relationship between being targeted by incivilityfrom a supervisor and instigated incivility. As organizationalfactors may come to impact the perpetration of workplaceincivility between employees, incivility should be consideredon both an individual and an organizational level.

In relation to the third hypothesis, it was striking thathaving a socially supportive and controllable environmentcoupled with high amounts of incivility was connected withmore instigated incivility. Literature has previously shownthat social support can have buffering effects on workplacebullying [26]. In the present study, however, it was foundthat high levels of social support from either coworkers orsupervisors moderated the relationship between experiencedincivility and instigation of more uncivil acts, contrary toa buffering hypothesis. This relationship could possibly bedue to a social climate in the organization. Similar aggressiveclimates in organizations has been discussed by Ramsay etal. [42], where groups with aggressive social rules are morelikely to engage in intergroup bullyingand to condone bul-lying between group members, especially if group membersstrongly identify with the group. In that way, the sociallysupportive environment and group cohesion can serve as anenhancement of current group norms in a negative or anaggressive climate. In an aggressive climate, the risk to beexcluded or victimized is higher when deviating from thenorm [43].

The fourth hypothesis was that experienced and wit-nessed incivility from supervisor or coworker related toemployees’ negative outcomes in the form of well-being, jobsatisfaction, turnover intentions, and sleeping problems.Thishypothesis was partly supported, as experienced incivilityfrom coworkers and witnessed incivility from a supervisorwere directly related to negative outcomes. The finding thatbeing targeted by incivility from a coworker directly relatesto negative outcomes is consistent with previous literatureon workplace incivility and detrimental effects on well-being[9, 16], job satisfaction [16, 17], and turnover intentions [6,19]. It is also consistent with literature on other types ofworkplace aggression and sleeping problems [21, 22]. Thedirect relationship of witnessing a supervisor acting uncivillyand negative outcomes is, however, a novel addition to theliterature. Turnover intentions have previously been morestrongly related to experienced incivility from a supervisorthan incivility from a coworker [20]. In the present study, wit-nessing supervisor incivility had a relationship with negative


outcomes, whereas witnessing coworkers acting uncivilly didnot significantly relate to negative outcomes. This gives somesupport to the notion that the relationship found by Leiterand colleagues [20] also can apply in the context of witnessedbehaviour.

Employees who are targeted by incivility from a super-visor report more job demands, lower social support andcontrol and as a result they perceive more negative outcomes,as predicted in the fifth hypothesis in the study. The fact thatthe organizational factors did not mediate the relationshipbetween any other sources of incivility and negative outcomeswas not in line with the hypothesis. However, the indirectrelationship of experienced supervisor incivility and negativeoutcomes via organizational factors illustrates the importanceof supervisors for the organizational climate and workers’health.

Contrary to hypothesis 6, none of the organizationalvariables moderated the relationships between any of theincivility variables and negative outcomes. This is not in linewith previous studies of the DCS model in organizationalresearch that assessed the buffering effects of demand controland support in relation to well-being [10].

Considering the overall findings, incivility appears linkedto a social process in the workplace for both instigatedincivility and negative outcomes.Whereas coworker incivilityhad the largest contribution to explain instigated incivility,experienced supervisor incivility contributed to explain neg-ative outcomes via organizational factors. One should notethat witnessed incivility from coworkers interestingly onlyhad a direct path to instigated incivility, and did not make asignificant contribution to any of the other tested hypotheses.The findings could be characteristic of the hospitality indus-try, as it has been pointed out as a sector that could fosteraggression [28]. This would explain the counter-intuitivemoderation of social support on the relationship betweenbeing targeted by incivility and instigated incivility.

4.1. Limitations. In relation to our models we report total,direct, mediation, and moderation effects but since thepresent work is cross-sectional there is no possibility to knowif the directions are causal. The high correlations betweenthe latent variables may have revealed one suppressed rela-tionship in the model. This can explain that the total effect(𝛽 = .179) was slightly lower than the direct path (𝛽 = .245)between control and instigated incivility.

Moreover, the low response rate could to some extenthave limited the study.The low response rate could maybe bedue to e-mail administration through the union. As the studywas conducted among members of the Hotel and RestaurantWorkers Union, largely representing unionized parties of thehospitality industry, the sample is not representative of ageneral population of the labour market.

The use of theWHO-Five scale in order to measure levelsof well-being needs to be considered. More intense testingof the Swedish version of the scale is warranted. However,previous studies have shown that using theWHO-Five ratherthan other instruments may reduce the risk of ceiling effects[36]. This factor could otherwise risk inducing a false imageof severity among the measures. The aforementioned factors

may to some extent have limited the study and need to betaken into consideration when interpreting the results.

4.2. Future Research. Based on the findings of this study,future research should consider workplace incivility as asocial phenomenon. More research is needed concerningthe different components of workplace incivility, and theirrelationships to instigated incivility and negative outcomes.Special attention should be paid tomediating andmoderatingeffects of organizational variables. The results found in thepresent study support potential indirect paths via organiza-tional variables, but these paths need to be more thoroughlyinvestigated in future research. In addition, testing the mod-eration effects of organizational factors should be particularlyconsidered in other samples, as the moderating role of asocially supportive environment is a counter-intuitive findingand may even be reversed in other sectors. Longitudinalstudies are needed to complement the cross-sectional natureof this studyand address the issue of causality in research onworkplace incivility.

5. Conclusion

The present research effort shows that workplace incivilitywas connected to both instigated incivility and negativeoutcomes in the form of reduced well-being, job satisfac-tion, turnover intentions, and sleeping problems. Witnessingcoworker incivility was the most important dimension toexplain instigated incivility. In addition, experienced incivil-ity from coworker and supervisor, as well as witnessed inci-vility from supervisor, were unexpectedly related to instigatedincivility via moderations of perceived high control and highsocial support.

Negative outcomes were to a high degree explainedby experienced supervisor incivility via mediation throughperceived low social support, low control, and high jobdemands. The results emphasize the significance of studyingworkplace incivility as a social process, considering bothexperienced andwitnessedworkplace incivility from cowork-ers and supervisors in the same model. The results alsoindicate the importance of including organizational factorsas key components in future studies of the research area.



Acknowledgment

Theproject was financed by the Swedish Council forWorkingLife and Social Research (FORTE dnr 2012-0138).

References

[1] L. M. Andersson and C. M. Pearson, “Tit for tat? the spiralingeffect of incivility in the workplace,” Academy of ManagementReview, vol. 24, no. 3, pp. 452–471, 1999.


[2] C. M. Pearson, L. M. Andersson, and J. W. Wegner, “Whenworkers flout convention: a study of workplace incivility,”Human Relations, vol. 54, no. 11, pp. 1387–1419, 2001.

[3] M. S. Hershcovis, “‘Incivility, social undermining, bullying...ohmy!’: a call to reconcile constructs within workplace aggressionresearch,” Journal of Organizational Behavior, vol. 32, no. 3, pp.499–519, 2011.

[4] B. Estes and J. Wang, “Integrative literature review: workplaceincivility: Impacts on individual and organizational perfor-mance,” Human Resource Development Review, vol. 7, no. 2, pp.218–240, 2008.

[5] B. Griffin, “Multilevel relationships between organizational-level incivility, justice and intention to stay,” Work and Stress,vol. 24, no. 4, pp. 309–323, 2010.

[6] C. M. Pearson, L. M. Andersson, and C. L. Porath, “Assessingand attacking workplace incivility,” Organizational Dynamics,vol. 29, no. 2, pp. 123–137, 2000.

[7] C. L. Porath and C. M. Pearson, “Emotional and behavioralresponses to workplace incivility and the impact of hierarchicalstatus,” Journal of Applied Social Psychology, vol. 42, no. 1, pp.E326–E357, 2012.

[8] S. G. Taylor and D. H. Kluemper, “Linking perceptions of rolestress and incivility to workplace aggression: the moderatingrole of personality,” Journal of Occupational Health Psychology,vol. 17, no. 3, pp. 316–329, 2012.

[9] S. Lim, L. M. Cortina, and V. J. Magley, “Personal and work-group incivility: impact on work and health outcomes,” Journalof Applied Psychology, vol. 93, no. 1, pp. 95–107, 2008.

[10] R. A. Karasek and T.Theorell,HealthyWork: Stress, Productivityand the Reconstruction of Working Life, Basic Books, New York,NY, USA, 1990.

[11] G. Notelaers, E. Baillien, H. de Witte, S. Einarsen, and J.K. Vermunt, “Testing the strain hypothesis of the DemandControl Model to explain severe bullying at work,” Economicand Industrial Democracy, vol. 34, no. 1, pp. 69–87, 2013.

[12] M. Ferguson and B. Barry, “I know what you did: the effects ofinterpersonal deviance on bystanders,” Journal of OccupationalHealth Psychology, vol. 16, no. 1, pp. 80–94, 2011.

[13] R. J. Martin and D. W. Hine, “Development and validationof the uncivil workplace behavior questionnaire,” Journal ofOccupational Health Psychology, vol. 10, no. 4, pp. 477–490,2005.

[14] K. Sakurai and S. M. Jex, “Coworker incivility and incivilitytargets’ work effort and counterproductive work behaviors:the moderating role of supervisor social support,” Journal ofOccupational Health Psychology, vol. 17, no. 2, pp. 150–161, 2012.

[15] P. Schilpzand, I. E. De Pater, and A. Erez, “Workplace incivility:a review of the literature and agenda for future research,” Journalof Organizational Behavior, 2014.

[16] T. G. Reio Jr. and R. Ghosh, “Antecedents and outcomes ofworkplace incivility: implications for human resource devel-opment research and practice,” Human Resource DevelopmentQuarterly, vol. 20, no. 3, pp. 237–264, 2009.

[17] L. M. Penney and P. E. Spector, “Job stress, incivility, andcounterproductive work behavior (CWB): the moderating roleof negative affectivity,” Journal of Organizational Behavior, vol.26, no. 7, pp. 777–796, 2005.

[18] P. C. Morrow, J. C. Mcelroy, and K. P. Scheibe, “Work unitincivility, job satisfaction, and total quality management amongtransportation employees,” Transportation Research Part E:Logistics andTransportationReview, vol. 47, no. 6, pp. 1210–1220,2011.

[19] L. M. Cortina, V. J. Magley, J. H. Williams, and R. D. Langhout,“Incivility in the workplace: incidence and impact,” Journal ofOccupational Health Psychology, vol. 6, no. 1, pp. 64–80, 2001.

[20] M. P. Leiter, S. L. Price, and H. K. Spence Laschinger, “Gen-erational differences in distress, attitudes and incivility amongnurses,” Journal of Nursing Management, vol. 18, no. 8, pp. 970–980, 2010.

[21] I. Niedhammer, S. David, S. Degioanni et al., “Workplacebullying and sleep disturbances: findings from a large scalecross-sectional survey in the Frenchworking population,” Sleep:Journal of Sleep and Sleep Disorders Research, vol. 32, no. 9, pp.1211–1219, 2009.

[22] D. C. Yamada, “The phenomenon of ‘workplace bullying’ andthe need for status-blind hostile work environment protection,”Georgetown Law Journal, vol. 88, no. 3, pp. 475–536, 2000.

[23] A.M.Hansen,A.Hogh, R. Persson, B.Karlson,A.H.Garde, andP.Ørbæk, “Bullying at work, health outcomes, and physiologicalstress response,” Journal of Psychosomatic Research, vol. 60, no.1, pp. 63–72, 2006.

[24] M. R. Tuckey, M. F. Dollard, P. J. Hosking, and A. H. Winefield,“Workplace bullying: the role of psychosocial work environ-ment factors,” International Journal of Stress Management, vol.16, no. 3, pp. 215–232, 2009.

[25] K. N. Miner, I. H. Settles, J. S. Pratt-Hyatt, and C. C. Brady,“Experiencing incivility in organizations: the buffering effects ofemotional and organizational support,” Journal of Applied SocialPsychology, vol. 42, no. 2, pp. 340–372, 2012.

[26] N. Djurkovic, D.McCormack, and G. Casimir, “Workplace bul-lying and intention to leave: the moderating effect of perceivedorganisational support,”Human Resource Management Journal,vol. 18, no. 4, pp. 405–422, 2008.

[27] E. Pelfrene, P. Vlerick, F. Kittel, R. P. Mak, M. Kornitzer, and G.De Backer, “Psychosocial work environment and psychologicalwell-being: assessment of the buffering effects in the jobdemand-control (-support) model in BELSTRESS,” Stress andHealth, vol. 18, no. 1, pp. 43–56, 2002.

[28] N. Johns and P. J.Menzel, “If you can’t stand the heat! . . . kitchenviolence and culinary art,” International Journal of HospitalityManagement, vol. 18, pp. 99–109, 1999.

[29] G. E.Mathisen, S. Einarsen, and R.Mykletun, “The occurrencesand correlates of bullying and harassment in the restaurantsector,” Scandinavian Journal of Psychology, vol. 49, no. 1, pp. 59–68, 2008.

[30] E. Schad, E. Torkelson, M. Backstrom, and B. Karlson, “Intro-ducing a Swedish translation of the workplace incivility scale,”Lund Psychological Reports, vol. 14, no. 1, pp. 1–15, 2014.

[31] R. J. Bennett and S. L. Robinson, “Development of a measure ofworkplace deviance,” Journal of Applied Psychology, vol. 85, no.3, pp. 349–360, 2000.

[32] L. M. Smith, M. A. Andrusyszyn, and H. K. Spence Laschinger,“Effects of workplace incivility and empowerment on newly-graduated nurses’ organizational commitment,” Journal of Nurs-ing Management, vol. 18, no. 8, pp. 1004–1015, 2010.

[33] G. Blau and L. Andersson, “Testing a measure of instigatedworkplace incivility,” Journal of Occupational and Organiza-tional Psychology, vol. 78, no. 4, pp. 595–614, 2005.

[34] J. H. Pejtersen, T. S. Kristensen, V. Borg, and J. B. Bjorner,“The second version of theCopenhagenPsychosocialQuestion-naire,” Scandinavian Journal of Public Health, vol. 38, no. 3, pp.8–24, 2010.


[35] National Research Centre for Working Environment,Copenhagen Psychosocial Questionnaire, 2015, 2014, http://www.arbejdsmiljoforskning.dk/en/publikationer/spoergeskemaer/psykisk-arbejdsmiljoe.

[36] P. Bech, L. R. Olsen, M. Kjoller, and N. K. Rasmussen, “Mea-suring well-being rather than the absence of distress symptoms:a comparison of the SF-36 mental health subscale and theWHO-five well-being scale,” International Journal of Methodsin Psychiatric Research, vol. 12, no. 2, pp. 85–91, 2003.

[37] Psychiatric Centre North Zealand, “WHO (fem) Valbefinnan-deindex,” 2015, https://www.psykiatri-regionh.dk/who-5/who-5-questionnaires/Pages/default.aspx.

[38] L.-T. Hu and P. M. Bentler, “Cutoff criteria for fit indexes incovariance structure analysis: conventional criteria versus newalternatives,” Structural Equation Modeling, vol. 6, no. 1, pp. 1–55, 1999.

[39] J. H. Steiger, “Understanding the limitations of global fitassessment in structural equation modeling,” Personality andIndividual Differences, vol. 42, no. 5, pp. 893–898, 2007.

[40] A. Klein and H. Moosbrugger, “Maximum likelihood esti-mation of latent interaction effects with the LMS method,”Psychometrika, vol. 65, no. 4, pp. 457–474, 2000.

[41] S. L. Robinson, W.Wang, and C. Kiewitz, “Coworkers behavingbadly: the impact of coworker deviant behavior upon individualemployees,” Annual Review of Organizational Psychology andOrganizational Behavior, vol. 1, no. 1, pp. 123–143, 2014.

[42] S. Ramsay, A. Troth, and S. Branch, “Work-place bullying:a group processes framework,” Journal of Occupational andOrganizational Psychology, vol. 84, no. 4, pp. 799–816, 2011.

[43] J. Escartın, J. Ullrich, D. Zapf, E. Schluter, and R. van Dick,“Individual- and group-level effects of social identification onworkplace bullying,” European Journal of Work and Organiza-tional Psychology, vol. 22, no. 2, pp. 182–193, 2013.

Research ArticleEffects of a Workplace Intervention Targeting Psychosocial RiskFactors on Safety and Health Outcomes

Leslie B. Hammer,1 Donald M. Truxillo,1 Todd Bodner,1 Jennifer Rineer,1

Amy C. Pytlovany,1 and Amy Richman2

1Portland State University, P.O. Box 751, Portland, OR 97207-0751, USA2Work Family Directions, 303 Wyman Street, Suite 300, Office No. 380, Waltham, MA 02451, USA

Correspondence should be addressed to Leslie B. Hammer; [email protected]

Received 16 January 2015; Revised 6 April 2015; Accepted 12 May 2015


Copyright © 2015 Leslie B. Hammer et al. This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.

The goal of this study was to test the effectiveness of a workplace intervention targeting work-life stress and safety-relatedpsychosocial risk factors on health and safety outcomes. Data were collected over time using a randomized control trial design with264 construction workers employed in an urban municipal department. The intervention involved family- and safety-supportivesupervisor behavior training (computer-based), followed by two weeks of behavior tracking and a four-hour, facilitated teameffectiveness session including supervisors and employees. A significant positive intervention effect was found for an objectivemeasure of blood pressure at the 12-month follow-up. However, no significant intervention results were found for self-reportedgeneral health, safety participation, or safety compliance.These findings suggest that an intervention focused on supervisor supporttraining and a teameffectiveness process for planning andproblem solving should be further refined andutilized in order to improveemployee health with additional research on the beneficial effects on worker safety.

1. Introduction

Work-life stress and poor safety communication are psy-chosocial risk factors that have been identified to contributeto decreased health and safety of workers. Workplace inter-ventions focused on increasing supervisor support for work-life balance and safety communication have proven to beeffective for reducing such risks (e.g., [1, 2]). Furthermore,it has been argued that strategies that take a Total WorkerHealth (TWH) approach may be the most effective way toimprove the health and safety of workers, addressing bothhealth promotion and health protection in an integrativefashion. The National Institute for Occupational Safety andHealth (NIOSH) defines TWH as “a strategy integratingoccupational safety and health protection with health pro-motion to prevent worker injury and illness and to advancehealth andwell-being” [3]. However, published studies on theeffectiveness of TWH programs remain scant [4].

The present study addresses this gap by assessing a TWHintervention, the Safety and Health Improvement Program

(SHIP), designed to address work-family stress and safetyrisk factors. We examine the effectiveness of SHIP using asample of construction workers, a sector and demographicgroup that the National Occupational Research Agenda [5]has targeted as understudied. Although there is recognitionthat managing work and family roles is challenging forworkers and their families and that these challenges leadto diminished worker health and safety (e.g., [6, 7]), fewworkplace interventions that specifically address work-lifestress and safety communication have been developed basedon theory and they have not been systematically tested usingscientifically sound experimental designs (for exceptions, see[2, 8–10]). Further, while the effects of supervisor behaviors,team, and organizational climate have been shown to affecta number of safety outcomes [11], relatively few studies haveexamined actual safety interventions other than Zohar andLuria [2], and no published intervention studies have takena TWH approach that integrates both work-life stress andsafety risk factors. Therefore, the present study addressesthis gap in the research by examining the effects of a TWH


http://dx.doi.org/10.1155/2015/836967


intervention that addresses work-life stress and safety onworker health, well-being, and safety outcomes using a ran-domized control design in a sample of construction workers.

The need for psychosocial workplace interventions topromote and protect construction worker health and safetywas recently illustrated by Bodner et al. [12] where work-family stress and conflict were significantly and positivelycorrelated with diastolic blood pressure, body mass index,and pain reports. Likewise, work-family stress and con-flict were significantly related to missing work in the pastsix months due to an injury. Although these relationshipsreported were correlational and, at one point in time, theysuggest that constructionworkers are a vulnerable populationand that interventions that reduce psychosocial risk factorssuch as targeting support for work-life balance and supportfor safety should be examined for their beneficial effects onworker health and safety.

2. Evaluation of Workplace Interventions:Process Evaluations and Effects Evaluation

Nielsen et al. [13] discussed the importance of process evalua-tion to intervention research and provided a discussion of theissues to consider when combining process evaluation andeffect evaluation data (i.e., evaluation of how an interventionworks versus an evaluation of what intervention works).It has been rightly argued that the context within whichintervention takes place needs to be considered to fullyunderstand the effects of an intervention. This includes theassessment of intervention fidelity as well as the complexcontextual environment that changes from group to groupandmay never be able to be controlled in a group randomizeddesign, for example. Usingmixedmethods designs is one wayof helping to triangulate the process evaluation and effectevaluation data.

Furthermore, Biron et al.’s [14] work identified severalfactors that potentially contribute to limited interventioneffects. They suggest that process evaluations should beconducted during intervention rollouts to better understandthe role of (1) organizational contextual influences such asreadiness for change; (2) the possibility that the introductionof an intervention is perceived as a job demand, leading todecreased well-being rather than to the expected beneficialeffects the researchers had hoped for; (3) low ownership bystakeholders; and (4) characteristics of the intervention suchas scope, approach, and target which may exceed existingorganizational resources.

This idea extends to the four levels of training criteriadiscussed by Kirkpatrick [15], running from thosemost easilyaffected by training (reactions) to critical outcomes (results)that are the ultimate goals of training programs. Specifically,reactions refer to training participants’ affective response tothe training; learning refers to improvement in knowledgeand skills after training; behavior refers to changes in behaviorsuch as transfer of learned knowledge and skills back to thejob; the ultimate training criterion, results, refers to changesin important organizational outcomes, such as organizationalproductivity or, in TWH terms, participant health. Using thisframework, in the present study, our focus was on the three

highest levels, specifically, changes in learning due to thecomputer-based training of supervisors; behavior in termsof transfer of knowledge and skills back to the job throughbehavior tracking and the team effectiveness process; andresults in terms of improvements in employee health andsafety.

Thus, using a combination of qualitative and quantitativemethods, we developed an evaluation of SHIP that includedboth process and effect assessments. We assessed follow-upand uptake of the intervention within workgroups at 30, 60,and 90 days after intervention through inspection of notesgenerated from check-in meetings. Furthermore, consistentwith training research, we argue that it is critical for trainingto include a design that fosters motivation to transfer thetraining content to the job (e.g., [16, 17]). As part of our train-ing design, we incorporated behavioral self-monitoring usingiPods that tracked behaviors learned during the training.Thishas been a proven method to enhance transfer of trainingin prior research (see [18]). Behavioral self-monitoring is atechnique in which individuals repeatedly observe, evaluate,and record aspects of their own behavior.

3. Work-Family Psychosocial RiskFactors and Health

Work-family/life stressors are rising for nearly every demo-graphic and occupational group in the U.S. [19]. There isgrowing recognition that work-family stressors have risenfor workers and their families across the nation, leadingto decreased health of workers and their family members[20, 21]. In addition, escalating time pressures and work-family conflict have negative business consequences suchas reduced worker productivity and turnover [22–24] andnegative long-term consequences for the economic health oforganizations and, ultimately, society. Effects of psychosocialfactors such as work-life stress on the health of workershave been documented, as have the effects of such stress andconflict on health behaviors, precursors to chronic healthoutcomes [7]. Furthermore, work-life stress has been shownto be related to worker safety outcomes [6, 25, 26]. Thus,work-life stress is a psychosocial risk factor and is identifiedas an occupational hazard by Hammer and Sauter [7]. Inturn, supervisor support for work and family is related toreductions in work-life stress [8, 9].

Work-family conflict and stress are linked to generalmental and physical health outcomes [27–32]; more chronicphysical symptoms; and higher levels of dysphoria, psycho-logical distress, and sickness absence [33, 34]. Other studiessuggest that, over time, the effects of work-family stress resultin negative health outcomes among objectively measuredindicators such as high blood pressure [35, 36] and othermental and physical health problems [37–39].

In the present study, a primary outcome of interest isblood pressure, a known risk factor for cardiovascular disease[40]. Although many factors can contribute to high bloodpressure, the impact ofwork-related psychosocial risk factors,such as job strain, on blood pressure and cardiovascular dis-ease is established [41–44]. However, there are also workplacefactors that can help buffer these effects. One mechanism


shown to help decrease the relationship between job strainand cardiovascular disease and its risk factors is socialsupport [45, 46]. For example, research has demonstrated thathigher levels of supervisor support are related to lower levelsof blood pressure and improved sleep [47, 48].

Building upon this, there are multiple types of socialsupport that may influence worker health: spousal support,coworker support, and supervisor support. For the purposesof this discussion, we will focus on the supervisor supportconstruct developed by Hammer and colleagues [49] calledfamily-supportive supervisory behaviors (FSSB). This is aform of social support focused on supervisors providingsupport specifically toworkers to assist with the integration ofwork and family, thereby reducingwork-life stress and relatedstrain outcomes. A central component of the interventionin the present study is a focus on reducing work-life stressthrough training supervisors to focus on FSSB.

This emphasis on training the supervisor is based onour earlier work that conceptualizes the supervisor as thelinking pin or the key critical organizational level that impactshealth and well-being of workers. We based the developmentof the intervention, supervisor supportive training, on theFSSB concept. FSSB is made up of 4 types of support basedon the work of Hammer and colleagues [49], and thisis represented in the training intervention that containedtraining on the FSSB dimensions of emotional support,instrumental support, work-family role modeling, and work-family creative management. In addition, the interventioninvolved a team-based work design change process thatdirectly involved the workgroup employees as described inthe Method section of the paper. The intervention consistedof two change approaches, a top-down approach that focuseson the supervisors (computer-based training) and a bottom-up approach that focuses on the workgroup members (teameffectiveness process). An advantage of this approach isthat it increases the odds of creating positive change inorganizations; a disadvantage of this approach is that wecannot attribute any intervention effects to either of the twocomponents. Thus, this intervention was designed to haveboth a top-down (i.e., supervisor-based) and bottom-up (i.e.,employee-based) components. This multilevel interventiondesign approach is expected to have a stronger effect onintervention outcomes compared to one level or another.

While few work-family interventions have been devel-oped based on theory and research and evaluated usingscientifically sound designs that integrate measurement ofthe intervention’s effects on safety and health outcomes[50, 51], there have been increasing employer interest inand experimentation with creating supportive work-familyworkplaces and flexible work arrangements, schedules, andother work-life and “family friendly” policies [42–44, 52–54]. However, most employers find it challenging to knowhow to effectively implement these new ways of working[55] and, specifically, which interventions are most effective.Moreover, rigorous evaluations of work-family programsand policies that involve longitudinal data and appropriatecomparison groups are virtually nonexistent. Identifying andtesting such workplace interventions to reduce work-familystress and conflict is an important public health issue, given

the significant effects of high work-family conflict on thehealth and well-being of workers and their families (e.g., [1,33]). Further, increasing organizational support for work andfamily through supportive managers and workplace culturesand through increasing employees’ involvement in develop-ing strategies for eliminating low-valueworkmay have signif-icant implications for the health and well-being of workers.

4. Safety-Related Psychosocial Risk Factorsand Safety Outcomes

Similarly, little is known about workplace interventions thatare implemented to improve safety communication andclimate, other than the work of Zohar and colleagues onsupervisor safety communication strategies (e.g., [2, 10, 56]).While there is a substantial literature regarding the consistenteffects of safety climate and leadership on safety behaviorsand attitudes (e.g., [11, 57–59]), few studies have addressingsafety climate training and interventions, and none havecombined work-life stress reduction with improvements insupervisor safety communication.

Safety is a critical outcome, especially in high-risk occu-pations such as construction. Safety outcomes are deter-mined bymore thanworkplace environmental and individualbehavior factors. As research is beginning to show, psychoso-cial workplace factors such as work-life stress and conflict,as well as poor safety communication and climate, also affectsafety outcomes [2, 6, 25, 26].

At least three critical meta-analyses have demonstratedthe link between safety climate (shared employee perceptionsof the safety environment at the organizational and grouplevel, as well as individual perceptions of climate) and safetyoutcomes [11, 57, 58]. It has further been argued that safetyclimate is determined by supervisory practices, communi-cations, and behaviors (e.g., [59]). For example, Zohar [60]found that transformational leadership focused on followerwelfare was important to safety outcomes. Accordingly,Zohar [61] used an intervention focused on increasing safetyinteractions in teams so that safety would be seen as havingas much priority as production. Feedback from the nextlevel of supervisors was also included. The interventionincreased safety-related interactions, improved safety culture,and decreased minor accidents. Zohar and Luria [2] alsofound thatmonitoring and providing feedback to supervisorsabout safety-related interactions improved safety behaviorand safety climate. Therefore, the development of an inter-vention that targets supervisory behaviors is likely to improvesafety climate, which has been shown meta-analytically toaffect safety behaviors by increasing worker safetymotivationand safety knowledge [11].

Griffin and Neal [62] break down the construct of safetybehaviors (performance) into two subdimensions: safetycompliance and safety participation. Safety compliance refersto engaging in core safety behaviors that are central tothe maintenance of a safe working environment, such aswearing safety goggles. Safety participation consists of morecontextual behaviors that contribute to an overall environ-ment of organizational safety, such as helping coworkers orvolunteering for safety-related activities. In the present study,


safety will be operationalized using these two dimensions[63].

5. The Present Study

The present study examines an integrated work-family andsafety support intervention, SHIP, over time within a vulner-able worker population, construction workers. Past researchhas used white collar and retail samples to assess the effectsof family support from supervisors. In contrast, in this studywe advance the literature by examining the effects of thisintervention on blood pressure, self-reported health, safetyparticipation behavior, and safety compliance behavior in aconstruction worker sample. Construction work is a safety-sensitive occupation and, as such, we focused on work-lifeconflict and safety communication training. Constructionworkers have more rigid schedules than professional-levelworkers who may experience high levels of work-life stressbut who may have more schedule flexibility to manage suchstress. Furthermore, construction work is a male-dominatedprofession and the expectation for hiding or limiting one’swork-life stress may be an additional subtle occupationalpressure. Furthermore, construction is a highly dangerous jobwithmany of the injuries in theUS occurring in construction.Thus, this is an appropriate sample for examining work-lifeand safety hazards.

Furthermore, this research is part of a larger researchprogram on supervisor support interventions and specificallyfamily-supportive supervisor behaviors (FSSB) conductedby Hammer and colleagues [8, 49]. We have extended thistraining paradigm to also include supervisor support forsafety and thus, with training focused on work-life conflictand safety, we chose theoretically relevant outcome variables.Furthermore, some of our prior research has demonstratedthat work-life conflict, specifically family-to-work conflict,was related to safety participation outcomes.

Construction workers complete physically demandingtasks on the job, but few studies have examined the negativehealth and safety effects related to psychosocial aspects oftheir work (for an exception, see [12]). To address this froma TWH perspective, SHIP was developed to improve workerhealth and safety through reductions in occupational stressvia improved supervisor support and team effectiveness.These intervention components are grounded in theory frommultiple disciplines and are partially supported by findingsfrom pilot/feasibility studies conducted by the Work, Family,and Health Network (http://www.workfamilyhealthnetwork.com/), as well as by the work of Zohar and colleagues.

The FSSB training intervention developed by Hammerand colleagues [8] as well as the supervisor-based safetytraining intervention developed by Zohar [61] was used asthe basis for the development of the integrated trainingapproach, SHIP, in the present study. Our behavior track-ing strategy was informed by Olson and Winchester’s [18]study demonstrating the effectiveness of this method forimproving training transfer. In addition, we drew on thesuggestion of Zohar and Luria [59], as well as that of Kellyand colleagues [9], to integrate work teams into the changeprocess. Thus, we used the team effectiveness process (TEP)

developed by Work Family Directions (WFD Consulting), aconsulting firm that specializes in work-family integrationpractices within organizations. The TEP process has beenused in a variety of industries with employees in many typesof jobs including hospitality, financial services, technology,engineering, manufacturing, call centers, and sales.

In sum, the purpose of this study is to examine theeffects of an integrated, theory-based work-life and safetyTWH intervention. SHIP targets the psychosocial workenvironment through training supervisors and work teammembers to decrease the psychosocial risk factors of work-life stress and poor safety communication, with expectedeffects on the health (e.g., blood pressure), well-being, andsafety of workers. Specifically, the components of SHIPare (1) family-supportive supervisor behavior (FSSB) andsupervisor-based safety (SBS) training, and (2) posttrainingtracking of learned behaviors, (3) team effectiveness processfor planning and problem solving (TEP), and (4) monthlyposttraining check-ins to revisit goals and assess progress.Overall, SHIPwas designed to increase work-life support andimprove safety and health. We examined the effectiveness ofSHIP using a randomized control design with constructionworkers employed in a municipal city utility department.In the present study, we test whether the SHIP interventionto increase workplace support and decrease stress improvesemployees’ safety, health, andwell-being.Wehypothesize thatSHIP will lead to improvements in worker safety (H1) andhealth (H2) over time.

6. Method

6.1. Participants and Design. Participants were constructionand utility workers in a municipal public works department.Job roles of participants included, but were not limited to,utility worker, electrician, plumber, carpenter, heavy equip-ment operator, and sidewalk repair person. Employees wereorganized into 8 divisions which were further divided into atotal 21 functional workgroups, each led by a supervisor. Halfof these workgroups (groups: 𝑘 = 11; employees: 𝑁 = 167)were randomly assigned to receive the SHIP intervention; theother half represented a control condition that received nointervention (groups: 𝑘 = 10; employees: 𝑁 = 125). Surveysand health assessments were administered prior to the inter-vention time period (baseline) and then again 12months later(follow-up). Of the 292 employees in the organization, 264(90%) participated in either the preintervention (𝑁 = 227) orpostintervention (𝑁 = 198) data collection periods, and 167participated in both baseline and follow-up. Figure 1 providesa Consort diagram for the study.

Participants were predominantly male (90%) and white(79%; 2% Hispanic or Latino) with 97% having completedhigh school and 54% with college experience. The averageage of participants was 45.13 years (SD = 9.60). Manyparticipants were married (60%) or living with a significantother (12%); 55% indicated having children at home, and33% indicated that they care for an adult relative. Participantshad worked in their current job on average 11.4 years (SD =8.5 years), and most (80%) reported working 40 hours perweek. Participants’ organizational roles were self-identified


Enrollment

AllocationUsual practice

10 work groups125 employees

Intervention11 work groups167 employees

Randomization

BaselineSurveys and health assessments

n = 90rr = 72%

BaselineSurveys and health assessments

n = 137rr = 82%

12 monthsSurveys and health assessments

n = 83rr = 66%

12 monthsSurveys and health assessments

n = 115rr = 69%

Analyzedn = 116 (participated at baseline

or 12 months)rr = 93%

Analyzedn =148 (participated at baseline

or 12 months)rr = 89%

Usual practice

Ship intervention

Analysis

Follow-up

Employees eligible to participate21 workgroups292 employees

Figure 1: Consort diagram for SHIP randomized control trial. Rr:retention rate.

as supervisor (8.3%), crew leader (13.4%), crew member(70.5%), and others (5.1%).

6.2. Intervention Description. The intervention examinedhere targets the entire work group, that is, both supervisorsand workgroup members. First, we will describe the twocomponents targeted at supervisors only. These includedcomputer-based training focused on FSSB and SBS using thecTRAIN platform, followed by tracking of trained behaviors(completed December 2012). Second, we will describe theworkgroup planning and problem solving team effectivenessprocess (TEP; completed January-February 2013) and thesubsequent 30-, 60-, and 90-day post-TEP check-ins. Figure 2depicts these intervention phases.

6.2.1. FSSB and SBS Computer-Based Supervisor Training.Team supervisors in the intervention condition first com-pleted a 1-hour computer-based training program usingthe cTRAIN training platform. The cTRAIN platform wasdesigned using proven behavioral training principles and isa self-paced, interactive training, with frequent quizzes andinformative feedback [64]. Content was based on Hammerand colleagues [8] FSSB training and Zohar and Luria’s [2]SBS training. Our preliminary research has demonstratedthat the effectiveness of the FSSB training program to improve

worker health using a computer-based trainingmethodology[8]. The SBS training module integrates methods used inthe only published safety intervention that has focused onimproving safety climate by improving communication skillsbetween supervisors and teammembers around safety issues[2]. Training included lessons on supervisor behaviors inthe following areas: (1) emotional support, (2) daily joband personal problem solving, (3) family-supportive rolemodeling, (4) creative work-family management, (5) safetycommunication, (6) feedback/reinforcement and coaching,(7) providing resources, and (8) safety role modeling.

6.2.2. Behavior Tracking. Next, supervisors chose specifictraining-related behaviors, based on the lessons noted above,that they wanted to improve. These behaviors were self-monitored and tracked for two weeks using HabiTrak track-ing software, which had been preloaded onto an iPod Touch.HabiTrak technology was designed based on decades ofresearch showing that transfer of knowledge is greater whenindividuals set goals and observe, evaluate, and record aspectsof their own behavior (e.g., [18, 65–69]). The HabiTrakprogram guided supervisors through the behavior changeprocess; users could access a detailed history of their behav-iors and receive support through a help tab that providedbehavioral definitions and video instructions for the specificlearned behaviors. This program has been proven effective totransfer training in past research [18].

6.2.3. Team Effectiveness Process (TEP). We drew on the sug-gestions of Zohar and Luria [59] and Kelly and colleagues [9],to integrate work teams into the change process. Accordingly,we used a modification of an existing team intervention, theteam effectiveness process (TEP; [70–72]) developed byWorkFamily Directions, a consulting firm that specializes in work-family integration practices within organizations. The TEPprocess involves an initial team assessment (brief paper-and-pencil survey of team practices, work-life effectiveness, andsources of overwork and inefficiency), followed by a four-hour team session led by aWFD-trained facilitator.These ses-sions apply social support and locus of control theory [73] toimprove team planning and problem solving and to encour-age supportive behaviors related to safety, health, and work-life balance within teams, including their supervisors. Duringthe TEP session, team members utilized a variety of groupproblem solving methods including a review of assessmentresults, root cause analysis, brainstorming solutions, smallgroup discussion, and voting on key issues. Conversationsfocused on designing newways of performing essential tasks,identifying and eliminating low-value work and increasingfocus on safety and positive work-familymanagement behav-iors. Each teamdeveloped an action plan outliningwhat stepsthey would take to make improvements, who was going tobe responsible for those steps, and when those actions wouldbe completed. Teams also developed operating principlessummarizing teamagreements regarding effectiveness, safety,and work-life balance supportive behaviors. Some examplesfrom the teams include “We will have regular crew meetingson the job site,” “We will encourage questions from newworkers,” and “We will respect each other’s personal issues.”


TEP(January-February 2013)

(i) 4-hour facilitated teamsession

(ii) Presessionquestionnaires

(iii) Planning, and problem-solving, supportivebehaviors

(iv) Action plans

Check-ins

(i) 30 days after TEP(ii) 60 days after TEP(iii) 90 days after TEP

∗100% of check-inscompleted

Behavior tracking(December 2012)

(i) 2-week tracking(ii) Self-set goals based

on training(iii) Definitions and videos

of learned behaviorseasily accessible oniPods used fortracking

Supervisor training(December 2012)

(i) 1-hour CBT(ii) Family supportive

behaviors(iii) Safety-supportive

behaviors

∗Average posttrainingquiz score: 85%

Figure 2: SHIP components. CBT: computer-based training, TEP: team effectiveness process.

6.2.4. Check-Ins. Supervisors met with their teams 30, 60,and 90 days following the TEP session to review the team’soperating principles, assess their progress, and update theiraction plans. The supervisor followed a Check-In MeetingGuide which included assessing change in morale and workattitudes, efficient use of time and resources, focus on safetypractices, and communication within the team. Successstories, “win-wins,” and best practices were also noted.

6.3. Data Collection Procedure. Baseline data were collectedOctober–December 2012 and the 12-month postinterventionassessments were conducted October–December 2013. Datacollection occurred at the work site during company time.Employees were informed that they were being invitedto participate in a research study about factors affectingemployees’ safety, health, andwork experiences. Participationwas voluntary, and each employee received a $25 gift cardfrom the researchers for completion at each data collectionsession. Tomatch surveys across time points, employees wereassigned unique identification codes based on an employeeroster provided by the organization.

Assessments included a paper-and-pencil survey andobjective health measures including blood pressure. Concur-rent with survey completion, employees were called one at atime to complete the health assessments. A copy of all healthmeasureswas provided to the participants, as well as informa-tion for how to read and interpret the results. Protocols werein place for occurrences of high blood pressure (>160/90mmHg), and additional information including the phone numberto a medical doctor was given to those participants.

7. Measures

7.1. Process Evaluation: Follow-Up Sessions at 30, 60, and90 Days. All teams completed the 30-, 60-, and 90-daycheck-in meetings which were led by the supervisor withassistance from the trained facilitator(s). After each check-in meeting, the supervisor completed a form which includedupdates to their action plan and ratings of changes in theteam on a 3-point scale from “No improvement” to “Greatimprovement.” These measures included morale and workclimate, efficient use of time and resources, focus on safetypractices, and communication within the team. Supervisors

also made qualitative notes about success stories, “win-wins”and “best practices.”

7.2. Effects Evaluation: Reactions, Learning, Behavior, andResults. Measures of supervisor computer-based quiz results(learning), safety behaviors (behaviors), and perceived health(results), as well as onsite physical health assessments(results), were taken at baseline and twelve months afterintervention from both intervention and control teams.

7.2.1. Supervisor Computer-Based Quiz Scores. Learning wasassessed as part of the computer-based training throughembedded quiz questions at the end of each subsection. Afinal overall score was computed for each manager based onthe posttest quiz score.

7.2.2. Safety Behaviors. Safety compliance and safety partic-ipation behaviors were each measured with three-item, self-report scales [63]. A sample item from the safety compliancemeasure is “I use the correct safety procedures for carryingout my job.” A sample item from the safety participationmeasure is “I voluntarily carry out tasks or activities thathelp to improve workplace safety.” Responses to the itemswere on a 5-point scale with options ranging from 1 =“Strongly Disagree,” through 3 = “Neutral,” to 5 = “StronglyAgree.” Scale scores were computed as the mean numericitem response with higher scores indicating higher levelsof safety compliance and participation. Scale scores for thesafety compliance (coefficient-alpha = .92 at baseline and.91 at follow-up) and safety participation (coefficient-alpha =.86 at baseline and .89 at follow-up) measures demonstratedacceptable levels of measurement reliability at both assess-ment periods.

7.2.3. Self-Reported Health. Physical health was measured bythe physical health composite score from the SF-12 [74], a12-item self-report inventory. A sample item is “In general,would you say your health is:” with item response optionsranging from 1 = “Poor” to 5 = “Excellent.” The physicalhealth composite score is a weighted composite of the 12-item responses with higher scores indicating higher levelsof physical health; physical health composite scores arepopulation normed to have a mean of 50 and a standarddeviation of 10. Scale scores for the physical health composite


Table 1: Estimated means, standard deviations, and correlations among study variables by intervention condition.

Variable(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

M M 47.51 48.10 95.78 94.41 3.63 3.75 4.12 4.11 44.87 0.25SD 6.67 7.07 11.97 10.18 0.82 0.81 0.67 0.56 9.11 0.44

(1) Health (baseline) 46.52 6.68 (.76) .61∗ −.16∗ −.21∗ .08 .08 .08 −.05 −.23∗ −.31∗

(2) Health (12m) 48.14 6.73 .53∗ (.81) −.31∗ −.30∗ −.03 .08 −.06 .11 −.23∗ −.35∗

(3) Blood pressure (baseline) 95.45 9.86 −.05 .18 1.00 .83∗ −.03 .07 .01 −.09 .14 .09(4) Blood pressure (12m) 95.81 9.89 −.06 −.15 .58∗ 1.00 −.15 −.07 −.06 −.08 .09 .10(5) Safety participation (baseline) 3.65 0.81 .02 .05 .02 .13 (.86) .60∗ .53∗ .40∗ .15 .05(6) Safety participation (12m) 3.67 0.84 −.18 −.03 .07 .11 .82∗ (.89) .37∗ .64∗ −.04 −.13(7) Safety compliance (baseline) 4.19 0.62 .09 −.10 .05 .25∗ .51∗ .42∗ (.92) .49∗ −.05 .15(8) Safety compliance (12m) 4.17 0.70 .14 .11 .00 .07 .52∗ .62∗ .62∗ (.91) −.05 −.02(9) Age in Years 45.50 10.13 −.14 −.27∗ .10 .03 .13 .13 .14 .06 1.00 .38∗

(10) Blood pressure medication 0.27 0.45 −.29∗ −.42∗ −.06 −.10 −.08 .07 .01 .21∗ .31∗ 1.00Notes: ∗𝑝 < .05. Intervention𝑁 = 148; Control𝑁 = 116. Intervention condition information above main diagonal; Control condition information below maindiagonal; Blood Pressure Medication (Yes = 1, No = 0). Estimates are based on full-information maximum likelihood estimation to account for missing datavalues. Diagonal entries in parentheses are Cronbach’s alpha reliability coefficients.

(coefficient-alpha = .76 at baseline and .81 at follow-up)demonstrated acceptable levels of measurement reliability atboth assessment periods.

7.2.4. Blood Pressure. Blood pressure was measured usingan Omron HEM-907EL machine, with an arm cuff. Threeconsecutive readings (with a one-minute rest in between)and an overall average were recorded. Mean blood pres-sure, defined as 1/3 systolic blood pressure + 2/3 diastolicblood pressure, was calculated for each participant’s averagereading. This measure of blood pressure has been shownto predict cardiovascular disease and death and may be thebest predictor of these health outcomes when single bloodpressure parameters are used [75]. In addition to these bloodpressure measurements, participants were asked whetherthey were currently taking blood pressure medication. Use ofsuchmedicationwas used as a control variable in the analysesexamining blood pressure as a dependent variable.

8. Results

Table 1 provides descriptive statistics for the various studyvariables at baseline and at the 12-month follow-up byintervention condition. No significant differences acrossintervention conditions were observed at baseline for bloodpressure (𝐵 = 0.31, 𝑝 = .84), SF-12 physical health compositescores (𝐵 = 1.23, 𝑝 = .18), safety compliance (𝐵 = −0.08,𝑝 = .42), and safety participation (𝐵 = −0.29, 𝑝 = .99). Giventhe lack of significant differences across groups at baseline,we next interpret some overall patterns in the data ignoringintervention group membership.

In addition to expected large correlations for the samevariables over time, we observe, as might also be expected,negative correlations between age and SF-12 physical healthcomposite scores at both time points, a negative correlationbetween taking blood pressuremedication and SF-12 physicalhealth composite scores at both time points, and a positivecorrelation between taking blood pressure medication and

age.We also observe significantly highermean blood pressurelevels at baseline (𝑝 = .001) and 12 months (𝑝 = .01) thanwhat would be considered normal (i.e., with a 120/80mmHgreading as “normal,” themean blood pressure should be 93.2).Furthermore, mean SF-12 physical health composite scoreswere significantly lower at baseline (𝑝 < .001) and at 12months (𝑝 = .002) than the population-normed mean valueof 50. Thus, this sample appears to be less healthy than whatis considered normative on these metrics.

8.1. Missing Data and Analytic Strategy. Of the 264 partic-ipants, 61 (33 intervention; 28 control) participated only atbaseline, 36 (11 intervention; 25 control) participated onlyat the 12-month follow-up, and 167 (104 intervention; 63control) participated at both baseline and follow-up (seeFigure 1 for the Consort diagram). Thus, there is a notableamount of missing data. Several analyses were conductedto explore patterns in the missing data using demographicvariables and the safety and health variables under inves-tigation. Missing demographic variables assumed invariantover time (e.g., ethnicity and gender) and logically structuredover time (e.g., age) were imputed based on the availablevariable value at the observed assessment wave. Those whoparticipated only at baseline were on average significantlyolder (𝑀 = 48.60) than those who participated only atfollow-up (𝑀 = 42.67) and at both assessments waves (𝑀 =44.41), 𝐹(2, 259) = 5.78, 𝑝 = .004. Furthermore, thosewho participated at baseline only were more likely to takeblood pressuremedication at baseline (36.1%) than thosewhoparticipated at both baseline and follow-up (22.3%), 𝜒2(1) =4.40, 𝑝 = .036. No other variables varied significantly acrossthese three participant groups (i.e., those who participatedonly at baseline, thosewho participated only at follow-up, andthose who participated at both time points).

In light of the amount of missing data and the notedpatterns across participation groups, we used the full-information maximum likelihood routine available in Mplus4.2 to estimate intervention effects. The advantage of this


Table 2: Model results for intervention effects on safety and health outcomes.

Predictor12-month safety outcomes 12-month health outcomes

DV: safety participation DV: safety compliance DV: blood pressure DV: physical healthCoefficient (SE) Coefficient (SE) Coefficient (SE) Coefficient (SE)

Intercept 1.37∗ (0.28) 2.13∗ (0.31) 32.87∗ (4.95) 24.03∗ (3.20)Age −0.01 (0.01) −0.00 (0.01) −0.03 (0.06) −0.06 (0.05)Blood pressure medication ⋅ ⋅ 0.29 (1.33) −3.44∗ (1.09)DV at baseline 0.69∗ (0.05) 0.51∗ (0.06) 0.68∗ (0.05) 0.55∗ (0.07)Intervention 0.14 (0.09) −0.02 (0.08) −2.15∗ (1.03) −0.32 (0.82)Residual variance 0.35∗ (0.04) 0.27∗ (0.03) 45.98∗ (4.95) 28.93∗ (3.20)Model 𝑅2 .49∗ .30∗ .55∗ .42∗

Notes: ∗𝑝 < .05. N = 264. Intervention (intervention = 1, control = 0); blood pressure medication (yes = 1, no = 0). Models use full-information maximumlikelihood routines to estimate parameters accounting for missing data.

missing data approach over typical software default options,such as listwise deletion, is that the full-information approachprovides more appropriate parameter estimates and stan-dard errors when the data are missing-at-random (MAR).Although there are no available tests for the MAR assump-tion, the plausibility of this assumption increases if observedvariables related to both the likelihood ofmissingness and thevalues of other observed variables without missing data areused in the analytical model [76].Thus, we include as controlvariables in each model the value of the safety and healthoutcome variable at baseline as well as participant age; for themodels testing for intervention effects on blood pressure andphysical health, we also include an indicator of blood pressuremedication use at baseline.

Finally, the lack of independence due to the nesting ofemployees within divisions or workgroups was assessed. Weused the intraclass correlation (ICC) to quantify the degreeof nonindependence for the four safety and health outcomes.For all four outcomes, Mplus estimated ICCs near zero (maxICC = .003). Thus, in the presented models, we do notestimate division- or workgroup-level random effects; theparameter estimates from the models that included theserandom effects were substantively identical to those reported.

8.2. Evaluation of Intervention Implementation. All supervi-sors in the intervention condition completed the computer-based training. To do so, supervisors needed to answer cor-rectly each of the periodic quizzes in the training to continueto the next training topic; incorrect responses required thesupervisor to repeat that training section before continuing.The average score on the final training knowledge test was85% indicating an adequate knowledge training outcome [15].All supervisors in the intervention condition reported usingthe behavioral self-monitoring tools. Although this data wasnot collected from the supervisors, poststudy interviewsyielded some insights on the ease and difficulty of this task.Supervisors found it easier to provide emotional support(e.g., taking time to talk to employees) and role modeling(e.g., leaving work on time and avoiding coming into workon weekends); supervisors found it more difficult to provideresources to manage conflicts (e.g., due to budget and staffingconstraints).

All workgroups assigned to the intervention conditioncompleted the TEP sessions. Furthermore, all workgroupssuccessfully completed the 30-, 60-, and 90-day check-intasks following these TEP sessions. At 60-day check-ins, 90%of teams reported some or great improvement to morale andwork attitudes; 70% of teams reported some improvementin more efficient use of time and resources; 100% of teamsreported some or great improvement to increased focuson safety practices within the team; and 100% reportedsome or great improvement to communication within theteam. Examples of topics that these workgroups identifiedand worked on include institute end of day jobs reviewmeeting, implementing new job priority system, institute“ride alongs” for Traffic Control to educate them to safetyrisks, and conducting preconstruction meetings on specificprojects to avoid emergencies and inefficiencies. When askedabout success stories, supervisors made comments such as“More communication between crew members,” “One staffmember wearing safety vest more,” and “Sharing liningprocess with other sections gets them what they need andremoves misunderstanding for crews.”

8.3. Tests of Intervention Effects. Table 2 provides the resultsof the analysis of intervention effects on the safety and healthoutcomes. For the safety outcomes, no significant interven-tion effects were observed. Despite trending in the expecteddirection, mean safety participation scores at the 12-monthpostintervention assessment period were not significantlyhigher in the intervention workgroups than in the controlworkgroups after controlling for baseline safety participationscores and age (𝐵 = 0.14, 𝑝 = .12, Δ𝑅2 = .014). Similarly,mean safety compliance scores at the 12-month postinterven-tion assessment period did not differ significantly among theintervention and control workgroups controlling for baselinesafety compliance scores and age (𝐵 = −0.02, 𝑝 = .83,Δ𝑅2

= .001). Thus, hypothesis 1 was not supported.For the health outcomes, mean blood pressure scores

at the 12-month postintervention assessment period weresignificantly lower in the interventionworkgroups than in thecontrol workgroups controlling for baseline blood pressurescores, age, and use of blood pressure medication (𝐵 = −2.15,𝑝 = .038, Δ𝑅2 = .015). In contrast, mean SF-12 physical


health composite scores at the 12-month postinterventionassessment period did not differ significantly among theintervention and control workgroups controlling for baselineSF-12 scores, age, and use of blood pressure medication (𝐵 =−0.32, 𝑝 = .69, Δ𝑅2 < .001). Thus, hypothesis 2 was partlysupported.

9. Discussion

Thepresent study found evidence of SHIP intervention effectson one of the two health indicators (i.e., blood pressure) andneither of the two safety indicators. Descriptively, the size ofthe intervention effect on blood pressure—considering thevalue of the Δ𝑅2 statistic and a difference of 2.15mmHg—would be considered small in size. While this decrease is inthe lower end of the range of systolic BP losses found in ameta-analysis of the impact of weight reduction on bloodpressure [77], it is well above the lowest. Furthermore, in thehealth domain, small effects can be important. Indeed clinicaltrials have been stopped on ethical grounds with even smallereffect sizes (e.g., [78–80]). Thus, we consider this effect asimportant because of the profound effects of elevated bloodpressure over hundreds of workers.

SHIP was an intervention designed to reduce psychoso-cial risk factors of work-life stress and poor safety commu-nication. While the specific risk factors were not assessedas outcomes, we can conclude that SHIP was successful inimproving blood pressure, a health outcome more closelyaligned theoretically with a reduction in work-life stressrather than with an improvement in safety communication.Therefore, we believe that there is evidence to suggest thatthe SHIP components of FSSB, SBS, and TEP, together, led toimprovements in blood pressure over time. Future researchshould examine ways of improving SHIP to more directlytarget safety outcomes and to examine in more detail theprocesses, both psychological and physical, through whichSHIP operates.

A limitation of the integrated intervention examined inthe present study is that the two intervention components(supervisor training andTEP), while reinforcing one another,cannot be teased apart as to whether the intervention effectsare due to the supervisor training component, the team-based component, or both. In the design of this study, weerred on the side of creating an intervention that wouldbe successful on the belief that programs in organizationsare most successful that involve top-down (i.e., supervisor-based) and bottom-up (i.e., employee-based) components.

Although the apparent lack of intervention effects on self-reported safety behaviors is disappointing, inspection of thebaselinemean scores for these variables shown in Table 1 sug-gests one possible explanation for the lack of safety-specificintervention effects, namely, a ceiling effect. Indeed, themeanbaseline score for safety compliance (𝑀baseline = 4.14) isclose to its maximum possible value of 5. The mean baselinescore for safety participation (𝑀baseline = 3.64, again with amaximumof 5) is not as high as that for safety compliance butis still well above the theoretical scalemidpoint. Both indicategeneral agreement that participants on average comply withand participate in safety-related considerations. Thus, there

was not much room for potential improvement, at least asmeasured through these instruments. These baseline meanslikely reflect some awareness of the importance of safe workroutines and the effects of prior safety training, critical issuesin an industry with a high injury risk. Additionally, based onour qualitative data from the 60-day check-ins, 100% of teamsreported some or great improvement to increased focus onsafety practices within the team; and 100% reported some orgreat improvement to communication within the team.

We view the nonsignificant but trending interventioneffect on safety participation, where there is likely less ofa ceiling effect due to the somewhat lower baseline mean,as promising and worth exploring in future research with alarger sample. Indeed, safety participationmay be a key factorin maintaining a safe workplace because it suggests a concernfor the safety of coworkers and the team. Further, it may bethat the nature of the intervention, focused on changing thesafety climate through supervisors, could have changed thestandard participants used to assess their safety behaviors.Such “beta change” among participants can actually makethe detection of change more difficult even with successfulinterventions [81].

In sum, we suggest that this study offers a first look at anintervention that integrates both safety and work-life balancesupervisor training, along with a team-based approach thatis focused on team effectiveness process for planning andproblem solving in an effort at reducing psychosocial riskfactors and in turn improving safety and health of workers.We believe that while this is a first step, it is an importantcontribution to the literature because there are few workplaceinterventions aimed at the reduction of psychosocial riskfactors that have been evaluated using scientifically soundresearch designs. These findings add to the existing work-lifestress reduction intervention research on the effectiveness ofFSSB training [8, 9] and extend this work to the examinationof team-based methods using the TEP approach, as wellas integrating safety communication training drawing onmethods used by Zohar and colleagues [2, 10, 59–61]. Wesuggest that while the findings for safety were not significant,this could be due to particularly strong safety behaviors inplace at baseline leading to possible ceiling effects. Thus,we would not abandon the safety communication trainingportion of SHIP but rather suggest that future research isneeded.

We believe that to effectively develop workplace inter-ventions that lead to improved health and safety of workers,we need to replicate, customize for different industries, andbetter understand the processes that are at play. This is justthe start of a research program on the effectiveness of SHIP.Questions related to strength of intervention, psychologicalprocesses, and behaviors impacted are all part of our futureresearch program as we currently work towards extendingthis research to additional industries and populations. Asstated by Biron et al. [82] “. . . information on how to developeffective strategies to reduce or eliminate psychosocial risksin the workplace is much more scarce, ambiguous andinconclusive.” We believe that the findings from this researchsuggest that an intervention focused on supervisor supporttraining and a team effectiveness process for planning and


problem solving can improve a critical area of employeehealth and this positive effect suggests that a continuedstrengthening and targeting of the intervention could expandthe impact to further improve employee health and safety.

In the end, we believe that our intervention offers impor-tant insight into ways that the psychosocial workplace riskfactors, at least that are associated with work-life stress, canbe impacted leading to improved health ofworkers as demon-strated by significant reductions in blood pressure. This typeof preventative approach, combined with more awarenessaround health promotion activities and behaviors, is whatNIOSH is working to advance with the focus on TWH.



Acknowledgments

Funding for this project was through the Oregon HealthyWorkforceCenter (http://www.ohsu.edu/ohwc) andNationalInstitute for Occupational Safety and Health Total WorkerHealth Center of Excellence (Grant no. U19OH010154).

References

[1] L. B. Hammer and K. L. Zimmerman, “Quality of work life,”in American Psychological Association Handbook of IndustrialOrganizational Psychology, S. Zedeck, Ed., pp. 399–431, Ameri-can Psychological Association, Washington, DC, USA, 2011.

[2] D. Zohar and G. Luria, “The use of supervisory practices asleverage to improve safety behavior: a cross-level interventionmodel,” Journal of Safety Research, vol. 34, no. 5, pp. 567–577,2003.

[3] Center for Disease Control and Prevention, Total WorkerHealth, 2014, http://www.cdc.gov/niosh/twh/.

[4] W. K. Anger, D. L. Elliot, T. Bodner et al., “Effectiveness of totalworker health interventions,” Journal of Occupational HealthPsychology, vol. 20, no. 2, pp. 226–247, 2015.

[5] NORA Construction Sector Council, National constructionagenda: for occupational safety and health research and practicein the U.S. construction sector, 2008, http://www.cdc.gov/niosh/nora/comment/agendas/construction/.

[6] J. C. Cullen and L. B. Hammer, “Developing and testing atheoretical model linking work-family conflict to employeesafety,” Journal of Occupational Health Psychology, vol. 12, no.3, pp. 266–278, 2007.

[7] L. B. Hammer and S. Sauter, “Total worker health and work-life stress,” Journal of Occupational and EnvironmentalMedicine,vol. 55, no. 12, pp. S25–S29, 2013.

[8] L. B. Hammer, E. E. Kossek, W. K. Anger, T. Bodner, and K. L.Zimmerman, “Clarifying work-family intervention processes:the roles of work-family conflict and family-supportive super-visor behaviors,” Journal of Applied Psychology, vol. 96, no. 1, pp.134–150, 2011.

[9] E. L. Kelly, P. Moen, J. M. Oakes et al., “Changing work andwork-family conflict: evidence from the Work, Family, andHealth Network,” American Sociological Review, vol. 79, no. 3,pp. 485–516, 2014.

[10] D. Zohar and T. Polachek, “Discourse-based intervention formodifying supervisory communication as leverage for safetyclimate and performance improvement: a randomized fieldstudy,” Journal of Applied Psychology, vol. 99, no. 1, pp. 113–124,2014.

[11] M. S. Christian, J. C. Bradley, J. C. Wallace, and M. J. Burke,“Workplace safety: a meta-analysis of the roles of person andsituation factors,” Journal of Applied Psychology, vol. 94, no. 5,pp. 1103–1127, 2009.

[12] T. Bodner, M. Kraner, B. Bradford, L. Hammer, and D. Truxillo,“Safety, health, and well-being of municipal utility and con-struction workers,” Journal of Occupational and EnvironmentalMedicine, vol. 56, no. 7, pp. 771–778, 2014.

[13] K. Nielsen, R. Randall, and R. K. B. Christensen, “Developingnew ways of evaluating organizational-level interventions,” inContemporary Occupational Health Psychology: Global Perspec-tives on Research and Practice, J. Houdmont and S. Leka, Eds.,vol. 1, pp. 21–45, John Wiley & Sons, London, UK, 2010.

[14] C. Biron, C. Gatrell, and C. L. Cooper, “Autopsy of a failure:evaluating process and contextual issues in an organizational-level work stress intervention,” International Journal of StressManagement, vol. 17, no. 2, pp. 135–158, 2010.

[15] D. Kirkpatrick, “Techniques for evaluating training programs,”Journal of the American Society for Training and Development,vol. 13, no. 11, pp. 3–9, 1959.

[16] M. J. Burke and R. R. Day, “A cumulative study of the effective-ness of managerial training,” Journal of Applied Psychology, vol.71, no. 2, pp. 232–245, 1986.

[17] J. K. Ford, S. W. J. Kozlowski, K. Kraiger, E. Salas, andM. S. Tea-chout, Improving Training Effectiveness in Work Organizations,Erlbaum, Mahwah, NJ, USA, 1997.

[18] R. Olson and J. Winchester, “Behavioral self-monitoring ofsafety and productivity in the workplace: a methodologicalprimer and quantitative literature review,” Journal of Organiza-tional Behavior Management, vol. 28, no. 1, pp. 9–75, 2008.

[19] E. E. Kossek and B. Distelberg, “Work and family employmentpolicy for a transformedwork force: current trends and themes,”in Work-Life Policies, A. Crouter and A. Booth, Eds., pp. 3–49,Urban Institute Press, Washington, DC, USA, 2014.

[20] K. L. Zimmerman and L. B. Hammer, “Positive Spillover: wherehave we been and where are we going?” in Contemporary Occu-pational Health Psychology: Global Perspectives on Research,Education, and Practice, J. Houdmont and S. Leka, Eds., vol. 1,pp. 272–295, Wiley-Blackwell, Chichester, UK, 2010.

[21] E. E. Kossek, B. A. Lautsch, and S. C. Eaton, “Telecommuting,control, and boundary management: correlates of policy useand practice, job control, and work-family effectiveness,” Jour-nal of Vocational Behavior, vol. 68, no. 2, pp. 347–367, 2006.

[22] A. A. Grandey and R. Cropanzano, “The conservation ofresources model applied to work-family conflict and strain,”Journal of Vocational Behavior, vol. 54, no. 2, pp. 350–370, 1999.

[23] E. L. Kelly, E. E. Kossek, L. B. Hammer et al., “Getting therefrom here: research on the effects of work–family initiatives onwork–family conflict and business outcomes,” The Academy ofManagement Annals, vol. 2, no. 1, pp. 305–349, 2008.

[24] R. G. Netemeyer, J. S. Boles, and R. McMurrian, “Developmentand validation of work-family conflict and family-work conflictscales,” Journal of Applied Psychology, vol. 81, no. 4, pp. 400–410,1996.

[25] T. D. Smith and D. M. Dejoy, “Occupational Injury in America:an analysis of risk factors using data from the General Social


Survey (GSS),” Journal of Safety Research, vol. 43, no. 1, pp. 67–74, 2012.

[26] L. Turner, A. Siow, and G. Kambourov, “Relationship skills inthe labor and marriage markets,” in Proceedings of the AnnualMeeting of the Society for EconomicDynamics (SED ’14),MeetingPapers no. 155, 2014.

[27] M. R. Frone, M. Russell, and M. L. Cooper, “Relation of work-family conflict to health outcomes: a four-year longitudinalstudy of employed parents,” Journal of Occupational and Orga-nizational Psychology, vol. 70, no. 4, pp. 325–335, 1997.

[28] M. R. Frone, “Work-family conflict and employee psychiatricdisorders: the national comorbidity survey,” Journal of AppliedPsychology, vol. 85, no. 6, pp. 888–895, 2000.

[29] J. H. Greenhaus, T. D. Allen, and P. E. Spector, “Health con-sequences of work–family conflict: the dark side of the work–family interface,”Research inOccupational Stress andWell Being,vol. 5, pp. 61–98, 2006.

[30] J. G. Grzywacz and B. L. Bass, “Work, family, andmental health:testing different models of work-family fit,” Journal of Marriageand Family, vol. 65, no. 1, pp. 248–262, 2003.

[31] L. B. Hammer, M. B. Neal, J. T. Newsom, K. J. Brockwood, andC. L. Colton, “A longitudinal study of the effects of dual-earnercouples’ utilization of family-friendly workplace supports onwork and family outcomes,” Journal of Applied Psychology, vol.90, no. 4, pp. 799–810, 2005.

[32] M. Melchior, A. Caspi, B. J. Milne, A. Danese, R. Poulton, andT. E.Moffitt, “Work stress precipitates depression and anxiety inyoung, working women and men,” Psychological Medicine, vol.37, no. 8, pp. 1119–1129, 2007.

[33] J. G. Grzywacz, “Work-family spillover and health duringmidlife: is managing conflict everything?” American Journal ofHealth Promotion, vol. 14, no. 4, pp. 236–243, 2000.

[34] A.Vaananen, R.Kalimo, S. Toppinen-Tanner et al., “Role clarity,fairness, and organizational climate as predictors of sicknessabsence: a prospective study in the private sector,” ScandinavianJournal of Public Health, vol. 32, no. 6, pp. 426–434, 2004.

[35] K. L. Belkic, P. A. Landsbergis, P. L. Schnall, and D. Baker, “Isjob strain a major source of cardiovascular disease risk?” Scan-dinavian Journal of Work, Environment & Health, vol. 30, no. 2,pp. 85–128, 2004.

[36] P. A. Landsbergis, P. L. Schnall, K. L. Belkic, D. Baker, J.Schwartz, and T. G. Pickering, “Work stressors and cardiovas-cular disease,”Work, vol. 17, no. 3, pp. 191–208, 2001.

[37] D. C. Ganster and J. Schaubroeck, “Work stress and employeehealth,” Journal of Management, vol. 17, no. 2, pp. 235–271, 1991.

[38] D. C. Ganster, M. L. Fox, and D. J. Dwyer, “Explaining employ-ees’ health care costs: a prospective examination of stressfuljob demands, personal control, and physiological reactivity,”Journal of Applied Psychology, vol. 86, no. 5, pp. 954–964, 2001.

[39] M. Macik-Frey, J. C. Quick, and D. L. Nelson, “Advances inoccupational health: from a stressful beginning to a positivefuture,” Journal ofManagement, vol. 33, no. 6, pp. 809–840, 2007.

[40] K. Steenland, L. Fine, K. Belkic et al., “Research findings linkingworkplace factors to CVD outcomes,” Occupational Medicine,vol. 15, no. 1, pp. 7–68, 2000.

[41] P. L. Schnall, P. A. Landsbergis, and D. Baker, “Job strain andcardiovascular disease,” Annual Review of Public Health, vol. 15,no. 1, pp. 381–411, 1994.

[42] J. Bond, E. Galinsky, S. Kim, and E. Brownfield, NationalStudy of Employers: Highlights of Findings, Families and WorkInstitute, New York, NY, USA, 2005.

[43] H. J. Holzer, “Work and family life: the perspective of employ-ers,” in Work, Family, Health, and Well-Being, S. M. Bianchi, L.M. Casper, and R. B. King, Eds., pp. 81–95, Lawrence ErlbaumAssociates, Mahwah, NJ, USA, 2005.

[44] Corporate Voices for Working Families and WFD Consulting,TheBusiness Impacts of Flexibility: An Imperative For Expansion,2005.

[45] C. Brisson, S. Montreuil, and L. Punnett, “Effects of an ergo-nomic training program on workers with video display units,”Scandinavian Journal of Work, Environment & Health, vol. 25,no. 3, pp. 255–263, 1999.

[46] B. N. Uchino, J. T. Cacioppo, and J. K. Kiecolt-Glaser, “Therelationship between social support and physiological pro-cesses: a review with emphasis on underlying mechanisms andimplications for health,” Psychological Bulletin, vol. 119, no. 3, pp.488–531, 1996.

[47] L. F. Berkman, O. Buxton, K. Ertel, and C. Okechukwu,“Managers’ practices related to work–family balance predictemployee cardiovascular risk and sleep duration in extendedcare settings,” Journal of Occupational Health Psychology, vol. 15,no. 3, pp. 316–329, 2010.

[48] N. Wager, G. Fieldman, and T. Hussey, “The effect on ambu-latory blood pressure of working under favourably and unfa-vourably perceived supervisors,” Occupational and Environ-mental Medicine, vol. 60, no. 7, pp. 468–474, 2003.

[49] L. B. Hammer, E. E. Kossek, K. Zimmerman, and R. Daniels,“Clarifying the construct of family-supportive supervisorybehaviors (FSSB): a multilevel perspective,” Research in Occu-pational Stress and Well Being, vol. 6, pp. 165–204, 2006.

[50] P. Brough and M. P. O’Driscoll, “Organizational interventionsfor balancing work and home demands: an overview,” Work &Stress, vol. 24, no. 3, pp. 280–297, 2010.

[51] L. B. Hammer, C. A. Demsky, E. E. Kossek, and J. Bray, “Work-family intervention research,” inOxford Handbook of Work andFamily, T. D. Allen and L. T. Eby, Eds., Oxford University Press,2015.

[52] E. L. Kelly, “The strange history of employer-sponsored childcare: interested actors, uncertainty, and the transformation oflaw in organizational fields,” American Journal of Sociology, vol.109, no. 3, pp. 606–649, 2003.

[53] E. E. Kossek, “Workplace policies and practices to support workand families,” in Work, Family, Health, and Well-Being, S. M.Bianchi, L.M.Casper, andR. B. King, Eds., pp. 95–115, LawrenceErlbaum Associates, Mahwah, NJ, USA, 2005.

[54] A. L. Richman, J. T. Civian, L. L. Shannon, E. Jeffrey Hill, and R.T. Brennan, “The relationship of perceived flexibility, supportivework-life policies, and use of formal flexible arrangements andoccasional flexibility to employee engagement and expectedretention,” Community, Work and Family, vol. 11, no. 2, pp. 183–197, 2008.

[55] A.M. Ryan and E. E. Kossek, “Work-life policy implementation:breaking down or creating barriers to inclusiveness?” HumanResource Management, vol. 47, no. 2, pp. 295–310, 2008.

[56] G. Luria, D. Zohar, and I. Erev, “The effect of workers’ visibilityon effectiveness of intervention programs: supervisory-basedsafety interventions,” Journal of Safety Research, vol. 39, no. 3,pp. 273–280, 2008.

[57] J. M. Beus, S. C. Payne, M. E. Bergman, and W. Arthur, “Safetyclimate and injuries: an examination of theoretical and empi-rical relationships,” Journal of Applied Psychology, vol. 95, no. 4,pp. 713–727, 2010.


[58] S. Clarke, “The relationship between safety climate and safetyperformance: a meta-analytic review,” Journal of OccupationalHealth Psychology, vol. 11, no. 4, pp. 315–327, 2006.

[59] D. Zohar and G. Luria, “A multilevel model of safety climate:cross-level relationships between organization and group-levelclimates,” Journal of Applied Psychology, vol. 90, no. 4, pp. 616–628, 2005.

[60] D. Zohar, “The effects of leadership dimensions, safety climate,and assigned priorities on minor injuries in work groups,”Journal of Organizational Behavior, vol. 23, no. 1, pp. 75–92,2003.

[61] D. Zohar, “Modifying supervisory practices to improve sub-unit safety: a leadership-based intervention model,” Journal ofApplied Psychology, vol. 87, no. 1, pp. 156–163, 2002.

[62] M. A. Griffin and A. Neal, “Perceptions of safety at work: aframework for linking safety climate to safety performance,knowledge, and motivation,” Journal of Occupational HealthPsychology, vol. 5, no. 3, pp. 347–358, 2000.

[63] A. Neal, M. A. Griffin, and P. M. Hart, “The impact of organiza-tional climate on safety climate and individual behavior,” SafetyScience, vol. 34, no. 1–3, pp. 99–109, 2000.

[64] W. K. Anger, D. S. Rohlman, J. Kirkpatrick, R. R. Reed, C. A.Lundeen, and D. A. Eckerman, “cTRAIN: a computer-aidedtraining system developed in SuperCard for teaching skillsusing behavioral education principles,”Behavior ResearchMeth-ods, Instruments, & Computers, vol. 33, no. 2, pp. 277–281, 2001.

[65] J. S. Hickman and E. S. Geller, “A safety self-management inter-vention for mining operations,” Journal of Safety Research, vol.34, no. 3, pp. 299–308, 2003.

[66] J. S. Hickman and E. S. Geller, “Self-management to increasesafe driving among short-haul truck drivers,” Journal of Organi-zational Behavior Management, vol. 23, no. 4, pp. 1–20, 2003.

[67] A. Jackson, Effectiveness of self-monitoring interventions: theeffects on safe bus operation [M.S. thesis], Electronic Theses,Treatises and Dissertations, paper 3834, 2003.

[68] T. R. Krause, “Self-observation,” in The Behavior-Based SafetyProcess: Managing Involvement for an Injury-Free Culture, pp.283–288, Van Nostrand Reinhold, New York, NY, USA, 1997.

[69] R. Olson and J. Austin, “Behavior-based safety and workingalone: the effects of a self-monitoring package on the safe per-formance of bus operators,” Journal of Organizational BehaviorManagement, vol. 21, no. 3, pp. 5–43, 2001.

[70] B. Munck, “Changing a culture of face time,” Harvard BusinessReview, vol. 79, no. 10, pp. 125–132, 2001.

[71] S. Shellenbarger, “Taking back the weekend: companies helpemployees cut back on overwork,”TheWall Street Journal, 2006,http://www.wsj.com/articles/SB114790837783355992.

[72] D. Burrus, “The team approach to tackling overwork,” TheAlliance, 2005, https://www.wfd.com/PDFS/Team%20Approach%20Burrus%20WFD%20The%20Alliance%20Aug%202005.pdf.

[73] B. H. Gottlieb and A. E. Bergen, “Social support concepts andmeasures,” Journal of Psychosomatic Research, vol. 69, no. 5, pp.511–520, 2010.

[74] J. E. Ware Jr., M. Kosinski, and S. D. Keller, “A 12-item short-form health survey: construction of scales and preliminary testsof reliability and validity,”Medical Care, vol. 34, no. 3, pp. 220–233, 1996.

[75] L. Palaniappan, L. A. Simons, J. Simons, Y. Friedlander, andJ. McCallum, “Comparison of usefulness of systolic, diastolic,and mean blood pressure and pulse pressure as predictors ofcardiovascular death in patients ≥60 years of age (The Dubbo

Study),”The American Journal of Cardiology, vol. 90, no. 12, pp.1398–1401, 2002.

[76] J. L. Schafer, Analysis of Incomplete Multivariate Data, vol. 72 ofMonographs on Statistics and Applied Probability, Chapman &Hall, London, UK, 1997.

[77] J. E. Neter, B. E. Stam, F. J. Kok, D. E. Grobbee, and J. M.Geleijnse, “Influence of weight reduction on blood pressure: ameta-analysis of randomized controlled trials,” Hypertension,vol. 42, no. 5, pp. 878–884, 2003.

[78] D. L. DeMets, R. Hardy, L.M. Friedman, and K. K. Gordon Lan,“Statistical aspects of early termination in the beta-blocker heartattack trial,” Controlled Clinical Trials, vol. 5, no. 4, pp. 362–372,1984.

[79] G. B. Kolata, “Fetal alcohol advisory debated,” Science, vol. 214,no. 4521, pp. 642–645, 1981.

[80] Steering Committee of the Physicians’ Health Study ResearchGroup, “Preliminary report: findings from the aspirin compo-nent of the ongoing Physicians’ Health Study,”TheNew EnglandJournal of Medicine, vol. 318, no. 4, pp. 262–264, 1988.

[81] R. E. Millsap and S. B. Hartog, “Alpha, beta, and gamma changein evaluation research: a structural equation approach,” Journalof Applied Psychology, vol. 73, no. 3, pp. 574–584, 1988.

[82] J. C. Biron,M.Karanika-Murray, andC. Cooper,OrganizationalInterventions: Addressing Process and Context, Routledge, Lon-don, UK, 2012.

Review ArticleAn Evaluation of the Policy Context on PsychosocialRisks and Mental Health in the Workplace in the EuropeanUnion: Achievements, Challenges, and the Future

Stavroula Leka,1 Aditya Jain,2 Sergio Iavicoli,3 and Cristina Di Tecco3

1Centre for Organizational Health & Development, School of Medicine, University of Nottingham, Jubilee Campus,Wollaton Road, Nottingham NG8 1BB, UK2Nottingham University Business School, University of Nottingham, Jubilee Campus, Wollaton Road, Nottingham NG8 1BB, UK3Italian Workers’ Compensation Authority (INAIL), Department of Occupational and Environmental Medicine,Epidemiology and Hygiene, Via Fontana Candida 1, Monteporzio Catone, 00040 Rome, Italy

Correspondence should be addressed to Stavroula Leka; [email protected]

Received 5 April 2015; Revised 13 August 2015; Accepted 1 September 2015

Academic Editor: Barthelemy Kuate Defo

Copyright © 2015 Stavroula Leka et al. This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Despite the developments both in hard and soft law policies in the European Union in relation to mental health and psychosocialrisks in the workplace, a review of these policies at EU level has not been conducted to identify strengths, weaknesses, and gapsto be addressed in the future. Keeping in mind that the aim should be to engage employers in good practice, ideally such policiesshould include key definitions and elements of the psychosocial risk management process, covering risk factors, mental healthoutcomes, risk assessment and preventive actions, or interventions. The current paper aims to fill this gap by reviewing hard andsoft law policies on mental health in the workplace and psychosocial risks applicable at EU level and conducting a gap analysisaccording to a set of dimensions identified in models of good practice in this area. Our review of ninety-four policies in totalrevealed several gaps, especially in relation to binding in comparison to nonbinding policies. These are discussed in light of thecontext of policy-making in the EU, and recommendations are offered for future actions in this area.

1. Introduction

It is generally accepted that “work is good for you,” contribut-ing to personal fulfillment and financial and social prosperity[1].There are economic, social, andmoral arguments that, forthose who are able to work, “work is the best form of welfare”[2–4] and is the most effective way to improve the well-beingof these individuals, their families, and their communities.Moreover, for people who have experienced poor mentalhealth, maintaining, or returning to, employment can alsobe a vital element in the recovery process, helping to buildself-esteem, confidence, and social inclusion [5]. A betterworking environment can help improve employment rates ofpeople who develop mental health problems. Not doing thisputs additional costs on governments that have to providesocial welfare support for people who would prefer to be inemployment.

There is also growing awareness that (long-term) unem-ployment is harmful to physical and mental health, so itcould be assumed that the opposite must be true that workis beneficial for health. However, that does not necessarilyfollow [1]. Work is generally good for your health and well-being, provided you have “a good job” [1, 6]. Good jobs areobviously better than bad jobs, but bad jobs might be eitherless beneficial or even harmful. In fact, a study byWesterlundet al. [7] shows an improvement in fatigue and depressivesymptoms associated with the retirement event, especially forthose exposed to the worst work environment.

This paper focuses on mental health in the workplaceand adopts a comprehensive approach and an inclusivedefinition of mental health with a focus not only on (theabsence of) mental health disorders but also on positive stateof psychological well-being. This approach underlines theneed to address mental health in its totality by recognising


http://dx.doi.org/10.1155/2015/213089


interrelationships among risks tomental health, subthresholdconditions of poor psychological health and well-being (suchas stress), which may have not yet resulted in a diagnosedmental health disorder but may severely affect their expres-sion, and diagnosed mental health disorders. According tothis perspective, efforts to tackle mental ill health should notfocus on particular problems in isolation, such as depression,for example, but they should seek to put in place policiesand practices that will tackle a wider range of risk factorsto mental health by appropriate interventions. These shouldprioritise prevention and tackling problems at source whilealso developing awareness and facilitating treatment.

This paper focuses on the workplace where one ofthe key states of suboptimal mental health that can havesevere consequences is work-related stress. Work-relatedstress is the response people may have when presentedwith work demands and pressures that are not matchedto their knowledge and abilities and which challenge theirability to cope [8]. The European Commission [9] definedstress as a pattern of emotional, cognitive, behavioural, andphysiological reactions to adverse and noxious aspects ofwork content, work organisation, and work environment.In the framework agreement on work-related stress [10],stress is defined as a state, which is accompanied by physical,psychological, or social complaints or dysfunctions andwhich results from individuals feeling unable to bridge a gapwith the requirements or expectations placed on them.

A substantial body of evidence is now available onwork-related risks that can negatively affect both mentaland physical health with an associated negative effect onbusiness performance and society [11]. Although risks in thephysical work environment can have a direct negative effecton mental health that is accentuated by their interactionwith risks in the psychosocial work environment. In addition,psychosocial hazards (also often termed work organisationcharacteristics or organisational stressors) have been shownto pose significant risk and have a negative impact on mentalhealth, mainly through the experience of work-related stress[11, 12].These hazards are closely associatedwith the changingnature of work.

1.1. The Prevalence and Impact of Work-Related PsychosocialRisks and Mental Ill Health in the EU. In 2005 and againin 2010, every fourth participant of the European WorkingConditions survey believed that their health is at risk dueto work-related stress [13]. Even from early 2000, studiessuggested that between 50 and 60% of all lost workingdays have some link with work-related stress [14] leadingto significant financial costs to companies as well as societyin terms of both human distress and impaired economicperformance. In 2002, the European Commission reportedthat the yearly cost of work-related stress and related mentalhealth problems in 15 Member States of the pre-2004 EU wasestimated to be on average between 3 and 4% of the grossnational product, amounting to C265 billion annually [15].

In addition, the estimates for the proportion of theworkforce in Europe that may be living with a mental healthproblem at any one time range from one in five [16] to two infive [17], with a lifetime risk of at least two in five [16]. In the

EU-27, it was found that 15% of citizens had sought help for apsychological or emotional problem, with 72% having takenantidepressants [18].

A report by EU-OSHA summarized the economic costsof work-related stress illnesses. It reported that, in France,between 220,500 and 335,000 (1–1.4%) people were affectedby a stress-related illnesswhich cost the society between C830and C1.656 million; in Germany, the cost of psychologicaldisorders was estimated to be EUR 3,000 million [19]. Eachcase of stress-related ill health has been reported to lead to anaverage of 30.9 working days lost [20]. Estimates from theUKLabour Force Survey indicate that self-reported work-relatedstress, depression, or anxiety accounted for an estimated11.4 million lost working days in Britain in 2008/09 [21].This was an increase from earlier estimates, which indicatedthat stress-related diseases are responsible for the loss of 6.5million working days each year in the UK, costing employersaround C571 million and society as a whole as much as C5.7billion. A recent study concluded that the “social cost” ofjust one aspect of work-related stress (job strain) in Franceamounts to at least 2-3 billion euros, taking into accounthealthcare expenditure related to absenteeism, people givingup work, and premature deaths [22].

1.2. Policies on Psychosocial Risks and Mental Health in theWorkplace. Psychosocial risks and their management areamong employers’ responsibilities as stipulated in the Frame-work Directive 89/391/EEC on Safety and Health of Workersat Work as it obliges employers to address and manage alltypes of risk in a preventive manner and to establish healthand safety procedures and systems to do so. In addition tothe Framework Directive, a number of policies and guidanceof relevance to mental health have been developed and areapplicable to the European level. These include both legallybinding instruments (such as EU regulations, Directives,decisions, and national pieces of legislation) and other “hard”policies (such as ILO conventions) developed by recognisednational, European, and international organisations as wellas nonbinding/voluntary policies (or “soft” policies) whichmay take the form of recommendations, resolutions, opin-ions, proposals, conclusions of EU institutions (Commission,Council, and Parliament), the Committee of the Regions, andthe European Economic and Social Committee, as well associal partner agreements and frameworks of actions, andspecifications, guidance, campaigns, and so forth initiated byrecognised European and international committees, agencies,and organisations.

Regulatory instruments of relevance tomental health andpsychosocial risks are applicable to all EU member states.However, even though each of these regulations addressescertain aspects of mental health and/or the psychosocialwork environment, it should be noted that the terms “mentalhealth,” “stress,” and “psychosocial risks” are not mentionedexplicitly inmost pieces of legislation [23].Themain examplein this respect is the Framework Directive 89/391/EEC onSafety and Health of Workers at Work. Even though theDirective asks employers to ensure workers’ health and safetyin every aspect related to work, “addressing all types of riskat source,” it does not include the terms “psychosocial risk”


or “work-related stress.” However, it does require employersto “adapt the work to the individual, especially as regardsthe design of workplaces, the choice of work equipment, andthe choice of working and production methods, with a view,in particular, to alleviating monotonous work and work ata predetermined work rate, developing a coherent overallprevention policy which covers technology, organizationof work, working conditions, social relationships, and theinfluence of factors related to the working environment.”

The Directive further specifies that “health surveillanceshould be provided for workers according to national sys-tems. Particularly sensitive risk groups must be protectedagainst the dangers which specifically affect them.” In thissense, there is an indirect reference to, and provision for, risksrelated to mental health at work. This is also the case for theDirective on organisation of working time (93/104/EC), whilethe Council Directive onwork with display screen equipment(90/270/EEC) actually refers to “problems ofmental stress” inthe context of risk assessment. It should be noted here that, insome EUmember states, the national regulatory frameworksare more specific than the key EU occupational health andsafety Directives and do make reference to psychosocial risksand work-related stress.

A debate has been taking place in the scientific andpolicy literatures about the lack of clarity in regulatoryframeworks and related guidance on mental health at workand the management of psychosocial risks (e.g., [24–26]). Arecent European Survey of Enterprises on New & EmergingRisks (ESENER) which covered over 28,000 enterprises in 31countries across Europe has revealed that even though work-related stress was reported among the key OSH concerns forEuropean enterprises, only about half of the establishmentssurveyed reported that they inform their employees aboutpsychosocial risks and their effects on health and safetyand less than a third had procedures in place to deal withwork-related stress. The findings of the survey also showedthat 42% of management representatives consider it moredifficult to tackle psychosocial risks, compared with othersafety and health issues. The most important factors thatmake psychosocial risks particularly difficult to deal withwere reported to be “the sensitivity of the issue,” “lack ofawareness,” “lack of resources,” and “lack of training” [27].The second edition of EU-OSHA’s ESENER collected similarinformation on OSH management and workplace risks, witha particular focus on psychosocial risks, from almost 50,000enterprises in 36 countries across Europe. Recently published,first findings have revealed that psychosocial risk factors arereported as more challenging than other risks. The mostimportant factors that make psychosocial risks particularlydifficult to deal are “lack of information” and “lack ofadequate tools to deal with the risk effectively” as perceivedby almost one in five establishments reporting “dealingwith difficult customers” or “experiencing time pressure”[28].

Similar findings have also been found in stakeholdersurveys, which report that many stakeholders still perceiveworkplace hazards as primarily relating to physical aspects ofthe work environment. Furthermore, where issues relating to

mental health are reported to be important OSH concerns,there are significant differences among the perception ofstakeholders in different countries in the EU [29, 30]. Thesedifferences in perception (in terms of perspectives, priorities,and interests) of mental health at work between socialactors, particularly between employers’ organisations andtrade unions, are a challenge for effective social dialogue onthese issues and for the effective implementation of recentlyintroduced voluntary policy initiatives for the managementof psychosocial risks such as the European framework agree-ments onwork-related stress and on harassment and violenceat work [31].

In addition to the regulatory instruments, a significantlylarger number of “soft” policy initiatives of relevance tomental health and psychosocial risks in the workplace havebeen developed and implemented at the EU level. An EU-OSHA report on workplace mental health promotion citessome of the recent policy documents and initiatives withinthe EU relevant to mental health at work [32]:

(1) Lisbon Strategy: EU goal for economic growth andcompetitiveness;

(2) Community Strategy on Health and Safety at Work,2007–2012;

(3) Commission White Paper “Together for Health”;(4) Framework Agreement on Work-related Stress;(5) Framework Agreement on Harassment and Violence

at Work;(6) The Mental Health Pact.

The EU-OSHA report highlights the wide scope of policies inthis area, which range from broad EU strategies and publichealth policies to social dialogue initiatives. In additionto these, other policy initiatives of relevance to mentalhealth and psychosocial risks in the workplace include thesetting-up of formalised stakeholder committees, EU levelcampaigns, policies on managing disability, and initiativesby organisations such as the WHO and ILO. Many of thesesoft law initiatives and policies are directly relevant to mentalhealth in the workplace, psychosocial risks, work-relatedstress, and their management. However, very little evaluationhas been conducted on hard and soft law policies in Europe.

An evaluation of the implementation of the FrameworkDirective conducted a decade ago indicated that the tasksof risk assessment, documentation, and supervision are notuniversally spread, even in member states with a traditionbased on prevention [33]. The report also highlighted that,where procedures were in place in organisations, they gener-ally focused on obvious risks where long-term effects (e.g.,mental health) as well as risks that are not easily observedwere being neglected. There was also hardly any consid-eration of psychosocial risk factors, and risk assessmentswere often being considered to be a one-time obligationlacking continuity where the efficiency of the measures wasnot sufficiently monitored by employers. The findings of theevaluation indicated that much still needed to be done asregards psychosocial risks such as work control and work


organisation, preventing unreasonably intense work pace,and repetitivework.This suggested an insufficient applicationof some of the general principles of the prevention foreseenin the Framework Directive 89/391 [25].

Concerning the evaluation of the framework agreementsfor work-related stress and for harassment and violence atwork, the main activities that followed the signing of theagreements were their translation in national languages [34,35]; however, they did act as catalysts for the implementationof new or updated legislation in some countries (e.g., theCzech republic and Italy). It should be noted that there is arather mixed picture regarding the state of European socialdialogue in the area of psychosocial risks at work and, asa result, serious questions have been raised in the literatureas to the appropriateness and effectiveness of “autonomous,or voluntary, agreements” [36]. Indeed, Ertel and colleagues[31] call for focused activities at European level to harmonizestakeholder perspectives on the issue of psychosocial riskfactors and work-related stress.

As discussed before, in some EU member states (e.g.,Sweden, Belgium, Italy, Germany, the Czech Republic), leg-islation is even more specific than EU law and makes directreference to work-related stress, bullying and harassment, orpsychosocial risks [6], although in very few countries stress-related diseases are included in official lists of occupationaldiseases. In addition, good practice examples in this area existin a number of member states. Some examples include theManagement Standards in the UK and Italy, Work Positive inIreland, the Work and Health Covenants and Catalogues inthe Netherlands, ISTAS in Spain, SOBANE in Belgium, thetools developed by INRS and ANACT in France, and EU-OSHA’s online simple risk assessment tool for SMEs, OiRA[37]. Indeed, Iavicoli et al. [38] have called for a criticalevaluation of efforts employed so far to address psychosocialrisks and mental health in the workplace to be conductedin order to develop an approach at European level that willallow both flexibility at national level and a certain level ofbenchmarking across members states in terms of relevantdata and good practices applied.

1.3. The Current Study. Since policies are an importantstarting point in addressing key issues, it is first important toidentify the key elements policies in this area should address.Keeping in mind that the aim should be to engage employersin good practice, ideally such policies should include ele-ments of the psychosocial riskmanagement process, coveringrisk factors,mental health outcomes, risk assessment and pre-ventive actions, or interventions. However, a review of hardand soft law policies at EU level along these dimensions hasnot been conducted to identify strengths and weaknesses andgaps to be addressed in the future. The current paper aims tofill this gap by reviewing hard and soft law policies on mentalhealth in the workplace and psychosocial risks applicable atEU level and conducting a gap analysis according to a setof dimensions identified in models of good practice in thisarea. In particular, the review and gap analysis has used thePRIMA-EF model as a guide [11] which highlights the keysteps and principles of the psychosocial risk managementprocess.

2. Method

The first step in the process was to identify all relevanthard and soft law policies of relevance to mental health inthe workplace and psychosocial risks. This was based onreviews previously conducted by themembers of the researchteam (see [23]). This review was updated to include sectoralDirectives as well as policies of relevance to mental healthin the workplace more broadly speaking (and not solelypsychosocial risks andwork-related stress).The review there-fore included not only general and specific health and safetypolicies but also policies relating to working hours, part-time work, temporary work, parental leave, discrimination,organizational restructuring (job insecurity), and so forth.

On the basis of a set of defined criteria in the form of apolicy scorecard (see Table 1), a gap analysis was carried out toexamine the extent towhich the current EUpolicy frameworkcovered issues relating to mental health in the workplace.Each policy (regulatory or nonbinding) was scored on a scaleof “0–5” on the basis of its relevance/applicability to and/orcoverage of dimensions relating tomental health at work.Thefive dimensions were chosen on the basis of good practiceguidance [11] and according to the comprehensive definitionon mental health in the workplace adopted in this study.The five dimensions were reference to mental health to inthe objectives and scope of the policy, coverage of exposurefactors, mental health problems/disorders at work and relatedoutcomes, risk assessment aspects, and preventive actions inrelation to mental health in the workplace. Policies whichdid not cover or refer to mental health at work were givena score of 0 while policies which were directly relevant andcomprehensively covered each dimension were given a scoreof 5.

Each policy was reviewed by four researchers working inpairs to analyze the policy content and “assign scores” on theestablished criteria. To ensure interrater reliability, a methodfor qualitative data analysis for applied policy research pro-posed by Ritchie and Spencer [39] was used where bothpairs of researchers reviewed the policy text for hard andsoft law policies independently. The assigned scores werethen discussed and reflected upon by all four researchers.Where disagreement arose, an independent expert reviewedthe policy in question. The final assigned score on eachdimension was established by consensus in terms of themajority.

3. Results

Table 2 presents the policy scorecard of regulatory instru-ments of relevance to mental health and psychosocial risksapplicable to the EU member states. These include Euro-pean Union Directives and ILO conventions. These reg-ulations address certain aspects of mental health and/orthe psychosocial work environment; however, most policiesscored 5 or below across the five dimensions highlightinga lack of coverage and specificity. Directive 89/391/EEC,the European Framework Directive on Safety and Healthof Workers at Work, received the highest score [13] alongwith a Directive 2010/32/EU, implementing the Framework


Table 1: Policy scorecard: key dimensions and scoring criteria.

Key dimensions 0 1 2 3 4 5Mental health inthe workplacereferred to in theobjectives andscope of the policy

Not covered by thegeneral objectivesor scope of the

policy

Covered inprinciple butnot effectivelyaddressed

Only implicitlycovered by the

objectives/scope ofthe policy

Partially coveredby the

objectives/scope ofthe policy

Sufficient coveragebut lack of

definitions of keyterms within the

policy

Comprehensivelycovered by the

general objective orscope of the policy

Coverage ofexposure factors inrelation to mentalhealth in theworkplace

No reference to oracknowledge-

ment/coverage ofexposure factors inrelation to mental

health in theworkplace


Only implicitacknowledge-

ment/coverage ofsome exposure

factors in relationto mental health inthe workplace

Partial acknowl-edgement/coverageof exposure factors

in relation tomental health inthe workplace


specificity onexposure factors inrelation to mental


Comprehensivecoverage of

exposure factors inrelation to mental


Coverage of mentalhealthproblems/disordersat work and relatedoutcomes

No reference oracknowledge-

ment/coverage ofmental health

problems/disordersat work and related

outcomes



ment/coverage ofmental health


outcomes

Partial acknowl-edgement/coverageof mental health


outcomes


specificity onmental health


outcomes

Comprehensivecoverage of mental

healthproblems/disordersat work and related

outcomes

Coverage of riskassessment aspectsin relation tomental health inthe workplace


ment/coverage ofrisk assessment

aspects in relationto mental health inthe workplace



ment/coverage ofrisk assessment

aspects in relationto mental health inthe workplace

Partial acknowl-edgement/coverageof risk assessmentaspects in relationto mental health inthe workplace


specificity on riskassessment aspects


Comprehensivecoverage of risk

assessment aspectsin relation to

mental health inthe workplace

Coverage ofpreventive actionsin relation tomental health inthe workplace


ment/coverage ofpreventive actions




ment/coverage ofpreventive actions


Partial acknowl-edgement/coverage

of preventiveactions in relationto mental health inthe workplace


specificity onpreventive actions


Comprehensivecoverage of

preventive actionsin relation to

mental health inthe workplace

Agreement on prevention from sharp injuries in the hospitaland healthcare sector concluded by HOSPEEM and EPSU.Directive 2010/32/EU is, however, applicable only to thehealthcare sector.

Table 3 presents the policy scorecard of voluntary policyinitiatives, which directly address mental health and psy-chosocial risks in the workplace. These policy initiativeswere scored much more favourably as compared to bind-ing/regulatory policies. Eleven policy initiatives had overallscores of 20 or more, indicating that many policy initiativesexplicitly referred to mental health and psychosocial risks inthe workplace in the objectives and scope of the policy andsufficiently or comprehensively covered aspects relating toexposure factors, mental health problems at work and relatedoutcomes, aspects of risk assessment and preventive actions.

Further analysis explored the average coverage of eachof the review dimensions across binding and nonbindingpolicies (see Figure 1). The solid lines in Figure 1 depictaverage scores of all binding/regulatory policies and all non-binding/voluntary policy initiatives, while the dotted linesplot the scores of the highest scored binding policy (Directive89/391/EEC) and nonbinding policy (PRIMA-EF guidance)

on each dimension. It is clear that nonbinding/voluntarypolicy initiatives aremore explicit in their reference tomentalhealth and psychosocial risks in the workplace in the objec-tives and scope of the policy and cover aspects relating toexposure factors, mental health problems at work and relatedoutcomes, aspects of risk assessment and preventive actions,in more detail as compared to binding/regulatory policiesoverall and in each of the five dimensions. A comparisonof the highest scored binding and nonbinding policies alsoindicate the same finding.

4. Discussion

From the review and gap analysis presented on regulatoryand voluntary policy initiatives, it is possible to make someobservations. Keeping in mind that the policies reviewed arethose that apply at European Union level alone (and notmember state policies), it is encouraging to see that a largenumber of relevant policies exist both of a binding and avoluntary nature. Our review covered thirty-four regulatoryand sixty voluntary policy initiatives and, in the case of thelatter, the number is likely to steadily increase year on year,


Table2:Po

licyscorecard:regu

latory

instr

umentsof

relevancetomentalh

ealth

andpsycho

socialris

ksin

thew

orkplace

attheE

urop

eanlevel.

Instr

ument

Mentalh

ealth

inthe

workplace

referred

toin

theo

bjectiv

es/scope

ofthep

olicy

Coverageo

fexp

osure

factorsinrelationto

mentalh

ealth

inthe

workplace

Coverageo

fmental

health

prob

lems/d

isordersa

tworkandrelated

outcom

es

Coverageo

frisk

assessmentaspectsin

relatio

nto

mentalh

ealth

inthew

orkplace

Coverageo

fpreventive

actio

nsin

relatio

nto

mentalh

ealth

inthe

workplace

Overall

(max.25)

(1)D

irectiv

e89/39

1/EE

CtheE

urop

ean

Fram

eworkDire

ctiveo

nSafetyand

Health

atWork

23

04

413

(2)D

irectiv

e2010/32/EUim

plem

entin

gtheframew

orkagreem

ento

npreventio

nfro

msharpinjurie

sintheh

ospitaland

healthcare

sector

concludedby

HOSP

EEM

andEP

SU

05

15

213

(3)D

irectiv

e200

3/88

/ECconcerning

certainaspectso

fthe

organisatio

nof

working

time(consolidates

andrepeals

Dire

ctive9

3/104/EC

)

13

23

312

(4)D

irectiv

e90/270/EE

Cthem

inim

umsafetyandhealth

requ

irementsforw

ork

with

displayscreen

equipm

ent(fift

hindividu

alDire

ctivew

ithin

them

eaning

ofArticle16

(1)o

fDire

ctive8

9/391/E

EC)

33

03

211

(5)D

irectiv

e92/85/ECon

pregnant

workersandwom

enwho

have

recently

givenbirthor

areb

reast-feeding

33

03

110

(6)D

irectiv

e94/33/ECon

thep

rotection

ofyoun

gpeop

leatwork

32

02

18

(7)C

155Occup

ationalSafetyandHealth

Con

vention(ILO

),1981

32

01

17

(8)D

irectiv

e200

0/78

/ECestablish

inga

generalframew

orkfore

qualtre

atmentin

employmentand

occupatio

n0

20

23

7

(9)C

183MaternityProtectio

nCon

vention(ILO

),2000

02

02

37

(10)C

181P

rivateE

mploymentA

gencies

Con

vention(ILO

),1997

04

02

17

(11)D

irectiv

e200

6/54

/ECon

the

implem

entatio

nof

thep

rincipleo

fequ

alop

portun

ities

andequaltreatmento

fmen

andwom

enin

mattersof

employment

andoccupatio

n

02

02

37


Table2:Con

tinued.

Instr

ument

Mentalh

ealth

inthe

workplace

referred

toin

theo

bjectiv

es/scope

ofthep

olicy

Coverageo

fexp

osure

factorsinrelationto

mentalh

ealth

inthe

workplace

Coverageo

fmental

health

prob

lems/d

isordersa

tworkandrelated

outcom

es

Coverageo

frisk

assessmentaspectsin

relatio

nto

mentalh

ealth

inthew

orkplace

Coverageo

fpreventive

actio

nsin

relatio

nto

mentalh

ealth

inthe

workplace

Overall

(max.25)

(12)D

irectiv

e200

2/14/ECestablish

inga

generalframew

orkforinformingand

consultin

gem

ployeesintheE

urop

ean

Com

mun

ity

01

02

25

(13)D

irectiv

e200

2/15/ECon

the

organisatio

nof

workingtim

eofp

ersons

perfo

rmingmob

ileroad

transport

activ

ities

01

11

25

(14)C

187Prom

otionalFramew

orkfor

Occup

ationalSafetyandHealth

Con

vention(ILO

),2006

01

11

25

(15)D

irectiv

e96/34

/ECon

the

fram

eworkagreem

ento

nparentalleave

01

00

34

(16)D

irectiv

e200

0/43

/EC

implem

entin

gthep

rincipleo

fequ

altre

atmentb

etweenperson

sirrespectiveo

fracialor

ethn

icorigin

01

01

24

(17)D

irectiv

e200

9/104/EC

concerning

them

inim

umsafetyandhealth

requ

irementsforthe

useo

fwork

equipm

entb

yworkersatwo

rk(secon

dindividu

alDire

ctivew

ithin

them

eaning

ofArticle16

(1)o

fDire

ctive8

9/391/E

EC)

[replacin

gDire

ctive8

9/655/EE

C]

01

11

14

(18)D

irectiv

e200

8/94

/ECon

the

protectio

nof

employeesinthee

vent

oftheinsolvencyof

theire

mployer

(repealin

gDire

ctive2

002/74/ECand

Cou

ncilDire

ctive8

0/987/EE

C)

01

01

13

(19)D

irectiv

e98/59

/ECon

the

approxim

ationof

thelaw

softhe

mem

ber

states

relatingto

collectiver

edun

dancies

01

01

13

(20)D

irectiv

e92/91/EEC

concerning

the

minim

umrequ

irementsforimproving

thes

afetyandhealth

protectio

nof

workersin

them

ineral-extracting

indu

strie

sthrou

ghdrilling(eleventh

individu

alDire

ctivew

ithin

them

eaning

ofArticle16

(1)o

fDire

ctive8

9/391/E

EC)

01

01

13


Table2:Con

tinued.

Instr

ument

Mentalh

ealth

inthe

workplace

referred

toin

theo

bjectiv

es/scope

ofthep

olicy

Coverageo

fexp

osure

factorsinrelationto

mentalh

ealth

inthe

workplace

Coverageo

fmental

health

prob

lems/d

isordersa

tworkandrelated

outcom

es

Coverageo

frisk

assessmentaspectsin

relatio

nto

mentalh

ealth

inthew

orkplace

Coverageo

fpreventive

actio

nsin

relatio

nto

mentalh

ealth

inthe

workplace

Overall

(max.25)

(21)D

irectiv

e92/104/EE

Con

the

minim

umrequ

irementsforimproving

thes

afetyandhealth

protectio

nof

workersin

surfa

ceandun

dergroun

dmineral-extractingindu

strie

s(twelfth

individu

alDire

ctivew

ithin

them

eaning

ofArticle16

(1)o

fDire

ctive8

9/391/E

EC)

01

01

13

(22)Directiv

e89/65

4/EE

Cconcerning

them

inim

umsafetyandhealth

requ

irementsforthe

workplace

(first

individu

aldirectivew

ithin

them

eaning

ofArticle16

(1)o

fDire

ctive8

9/391/E

EC)

01

01

02

(23)D

irectiv

e89/65

6/EE

Con

the

minim

umhealth

andsafetyrequ

irements

forthe

useb

yworkersof

person

alprotectiv

eequ

ipmentatthe

workplace

(third

individu

aldirectivew

ithin

the

meaning

ofArticle16

(1)o

fDire

ctive

89/391/EEC

)

01

01

02

(24)D

irectiv

e90/26

9/EE

Con

the

minim

umhealth

andsafetyrequ

irements

forthe

manualh

andlingof

loadsw

here

thereisa

riskparticularlyof

back

injury

toworkers(fo

urth

individu

alDire

ctive

with

inthem

eaning

ofArticle16

(1)o

fDire

ctive8

9/391/E

EC)

01

01

02

(25)C

175Part-timeW

orkCon

vention

(ILO

),1994

01

01

02

(26)D

irectiv

e97/81/ECconcerning

the

fram

eworkagreem

ento

npart-timew

ork

01

00

12

(27)D

irectiv

e99/70

/ECconcerning

the

fram

eworkagreem

ento

nfixed-te

rmwork

01

01

02

(28)D

irectiv

e200

0/79

/ECconcerning

theE

urop

eanAgreemento

nthe

Organisa

tionof

Working

Timeo

fMob

ileWorkersin

CivilA

viation

01

00

12


Table2:Con

tinued.

Instr

ument

Mentalh

ealth

inthe

workplace

referred

toin

theo

bjectiv

es/scope

ofthep

olicy

Coverageo

fexp

osure

factorsinrelationto

mentalh

ealth

inthe

workplace

Coverageo

fmental

health

prob

lems/d

isordersa

tworkandrelated

outcom

es

Coverageo

frisk

assessmentaspectsin

relatio

nto

mentalh

ealth

inthew

orkplace

Coverageo

fpreventive

actio

nsin

relatio

nto

mentalh

ealth

inthe

workplace

Overall

(max.25)

(29)C

ouncilDirectiv

e200

1/23/ECon

thea

pproximationof

thelaw

softhe

mem

berstatesrela

tingto

the

safeguarding

ofem

ployees’rig

htsinthe

evento

ftransfersof

undertakings,

busin

esseso

rpartsof

undertakings

orbu

sinesses

01

00

12

(30)D

irectiv

e200

2/73/ECon

equal

treatmentfor

men

andwom

enas

regards

accessto

employment,vocatio

naltraining

andprom

otion,

andworking

cond

ition

s(amending

Dire

ctive7

6/207/EE

C)

01

00

12

(31)D

irectiv

e200

9/38

/ECon

the

establish

mento

faEu

ropean

Works

Cou

ncilor

aprocedu

rein

Com

mun

ity-scaleun

dertakings

and

Com

mun

ity-scalegrou

psof

undertakings

forthe

purposes

ofinform

ingand

consultin

gem

ployees(recast)

01

00

12

(32)D

irectiv

e93/103/EC

concerning

the

minim

umsafetyandhealth

requ

irements

forw

orkon

boardfishing

vessels

(thirteenthindividu

alDire

ctivew

ithin

them

eaning

ofArticle16

(1)o

fDire

ctive

89/391/EEC

)

01

00

12

(33)D

irectiv

e92/57/EEC

onthe

implem

entatio

nof

minim

umsafetyand

health

requ

irementsattempo

rary

ormob

ileconstructio

nsites

(eighth

individu

alDire

ctivew

ithin

them

eaning

ofArticle16

(1)o

fDire

ctive8

9/391/E

EC)

01

00

12

(34)D

irectiv

e91/38

3/EE

Csupp

lementin

gthem

easuresto

encourageimprovem

entsin

thes

afety

andhealth

atworkof

workerswith

afixed-durationem

ploymentrelationship

oratem

porary

employmentrela

tionship

01

01

02


Table3:Po

licyscorecard:voluntarypo

licyinitiatives

ofrelevancetomentalh

ealth

andpsycho

socialris

ksin

thew

orkplace.

Docum

ent

Mentalh

ealth

inthew

orkplace

referred

toin

the

objectives

and

scop

eofthe

policy

Coverageo

fexpo

sure

factorsin

relatio

nto

mental

health

inthe

workplace

Coverageo

fmental

health

prob

lemsa

tworkandrelated

outcom

es

Coverageo

frisk

assessmentaspects

inrelationto

mentalh

ealth

inthew

orkplace

Coverageo

fpreventiv

eactions

inrelationto

mentalh

ealth

inthew

orkplace

Overall

(max.25)

(1)G

uida

nce:EC

,199

9Guidanceo

nWork-Re

lated

Stress—Spiceo

fLife

orLissof

Death?

45

55

524

(2)G

uida

nce:EU

-OSH

A,200

2How

toTackle

Psycho

socialIssues

andRe

duce

Work-Re

lated

Stress

45

55

524

(3)G

uida

nce:WHO,200

8PR

IMA-

EF:G

uidanceo

ntheE

urop

eanFram

eworkforP

sychosocialR

iskManagem

ent:ARe

source

forE

mployersa

ndWorker

Representativ

es

45

55

524

(4)G

uida

nce:ILO,198

6Psycho

socialFactorsa

tWork:

Recogn

ition

andCon

trol

45

55

524

(5)G

uida

nce:ILO,2012SO

LVEAp

proach

45

55

524

(6)G

uida

nce:WHO,200

3,WorkOrganizationand

Stress

45

54

523

(7)W

HOHealth

yWorkp

lacesF

ramew

ork,20

10,

Health

yWorkplaces:AMod

elforA

ction:

For

Employers,Workers,Policym

akersa

ndPractitioners

45

44

522

(8)W

HOMentalh

ealth

declarationforE

urop

e,20

05,and

MentalH

ealth

Actio

nPlan

forE

urop

e5

44

44

21

(9)W

HOEu

rope

anMentalH

ealth

Actio

nPlan

,2013

45

53

421

(10)G

uida

nce:ILO,2012,StressPreventio

natWork

Checkp

oints:Practic

alIm

provem

entsforS

tress

Preventio

nin

theW

orkplace

45

44

421

(11)E

UHigh-levelC

onference,Br

ussels,

2010,

Investinginto

well-b

eing

atWork:Managing

Psycho

socialRisksinTimes

ofCh

ange

45

34

420

(12)C

ommittee

ofSenior

Labo

urInspectors

(SLIC)

,20

12,C

ampaignon

Psycho

socialRisksa

tWork

44

34

318

(13)C

ommun

icationfrom

theC

ommiss

ionCOM

(2014)

332on

anEU

StrategicF

ramew

orkon

Health

andSafetyatWork2014–2020

44

33

418

(14)F

ramew

orkAgreemento

nWork-relatedStress,

2004

European

SocialPartners—ET

UC,

UNICE

(BUSINES

SEURO

PE),UEA

PME,

andCE

EP3

43

34

17

(15)C

ommun

icationfrom

theC

ommiss

ion{SE

C(200

7)214–

216}

Improvingqu

ality

andProd

uctiv

ityat

Work:Com

mun

ityStrategy

2007–2012on

Health

and

SafetyatWork

43

33

417


Table3:Con

tinued.

Docum

ent

Mentalh

ealth

inthew

orkplace

referred

toin

the

objectives

and

scop

eofthe

policy

Coverageo

fexpo

sure

factorsin

relatio

nto

mental

health

inthe

workplace

Coverageo

fmental

health

prob

lemsa

tworkandrelated

outcom

es

Coverageo

frisk

assessmentaspects

inrelationto

mentalh

ealth

inthew

orkplace

Coverageo

fpreventiv

eactions

inrelationto

mentalh

ealth

inthew

orkplace

Overall

(max.25)

(16)E

U-C

onference,Be

rlin,

2011-PromotingMental

Health

andWell-B

eing

inWorkplaces

44

33

317

(17)E

NISO1007

5-1:1991

Ergono

micprinciples

Related

toWork-Lo

ad–G

eneralTerm

sand

Definitio

ns2

43

34

16

(18)R

194revisedan

nex,ILO20

10,R

ecom

mendatio

nconcerning

theL

istof

Occup

ationalD

iseases

andthe

RecordingandNotificatio

nof

Occup

ationalA

ccidents

andDise

ases

44

43

N/A

15

(19)E

NISO1007

5-2:1996

Ergono

micPrinciples

Related

toWork-Lo

ad–D

esignPrinciples

23

33

415

(20)O

pinion

oftheE

urop

eanEc

onom

ican

dSo

cial

Com

mittee,2013,on

theE

urop

eanYear

ofMental

Health

—Be

tterW

ork,Be

tterQ

ualityof

Life(2013/C

44/06)

44

13

315

(21)C

ouncilof

theE

urop

eanUnion

Con

clusions,

2002,oncombatin

gstr

essa

nddepressio

n-related

prob

lems

32

41

414

(22)E

urop

eanPa

ctforM

entalH

ealth

and

Well-b

eing

,200

8,To

gether

forM

entalH

ealth

and

Well-b

eing

33

23

314

(23)E

urop

eanPa

rliam

entresolutionT6

-006

3/20

09on

MentalH

ealth

,Reference

2008/2209(INI),

Non

legisla

tiveR

esolution

33

23

314

(24)G

reen

paper–

EC,200

5,Im

provingtheM

ental

Health

oftheP

opulation:

Towards

aStrategyon

Mental

Health

forthe

European

Union

32

32

313

(25)E

urop

eanPa

rliam

entR

esolution20

06/205

8(INI)on

ImprovingtheM

entalH

ealth

ofthe

Popu

latio

n:To

wards

aStrategyon

MentalH

ealth

for

theE

urop

eanUnion

33

22

313

(26)G

uida

nce:WHO,200

7Ra

ising

Awarenesso

fStressatWorkin

Develo

ping

Cou

ntrie

s:aM

odern

Hazardin

aTraditio

nalW

orking

Environm

ent:Ad

vice

toEm

ployersa

ndWorkerR

epresentatives

33

22

212

(27)G

uida

nce:EU

-OSH

A,2011W

orkplace

Violence

andHarassm

ent:aE

urop

eanPicture

22

23

312

12 BioMed Research InternationalTa

ble3:Con

tinued.

Docum

ent

Mentalh

ealth

inthew

orkplace

referred

toin

the

objectives

and

scop

eofthe

policy

Coverageo

fexpo

sure

factorsin

relatio

nto

mental

health

inthe

workplace

Coverageo

fmental

health

prob

lemsa

tworkandrelated

outcom

es

Coverageo

frisk

assessmentaspects

inrelationto

mentalh

ealth

inthew

orkplace

Coverageo

fpreventiv

eactions

inrelationto

mentalh

ealth

inthew

orkplace

Overall

(max.25)

(28)G

uida

nce:WHO,200

3Ra

ising

Awarenessto

Psycho

logicalH

arassm

entatW

ork

22

22

311

(29)C

harter

ofFu

ndam

entalR

ightso

fthe

Europe

anUnion

(200

0/C36

4/01)

24

10

411

(30)G

uida

nce:ILO,200

6Violence

atWork

22

22

311

(31)C

ommun

icationfrom

theC

ommiss

ionCOM

(2010)

682AnAgend

afor

New

Skillsa

ndJobs:A

European

Con

tributiontowards

FullEm

ployment

43

12

111

(32)W

HOAc

tionPlan

,2012forImplem

entatio

nof

theE

urop

eanStrategy

forthe

Preventio

nandCon

trol

ofNon

commun

icableDise

ases

2012–2016

13

13

311

(33)C

ouncilof

theE

urop

eanUnion

Con

clusions,

2003

onMentalh

ealth

–Con

ferenceo

nMentalIlln

ess

andStigmainEu

rope:FacingUptheC

hallenges

ofSo

cialInclu

sionandEq

uity

32

21

210

(34)C

ouncilof

theE

urop

eanUnion

Con

clusions,

2005

onaC

ommun

ityMentalH

ealth

Actio

n–Outcome

ofProceeding

s3

21

13

10

(35)O

pinion

oftheE

urop

eanEc

onom

ican

dSo

cial

Com

mittee,200

5on

theG

reen

PaperImprovingthe

MentalH

ealth

oftheP

opulation—

Towards

aStrategy

onMentalH

ealth

forthe

European

Union

(2006/C

195/11)

31

11

28

(36)E

C20

07-W

hitepa

per-To

gether

forH

ealth

—A

StrategicA

pproachforthe

EU2008–2013

21

13

18

(37)F

ramew

orkAgreemento

nHarassm

enta

ndViolence

atWork,20

07,Eu

ropean

Social

Partners-ETU

C,BU

SINES

SEURO

PE,U

EAPM

E,and

CEEP

13

11

28

(38)Th

eStand

ingCom

mittee

ofEu

rope

anDoctors

(CPM

E)Po

sitionPa

per,20

09,M

entalH

ealth

inWorkplace

Setting

s“Fitand

Health

yatWork”

32

10

28

(39)C

ouncilRe

solutio

n20

00/C

86/01o

nTh

eProm

otionof

MentalH

ealth

30

20

27

(40)M

entaland

PhysicalHealth

Platform

(MPH

P)20

09,Th

eMentaland

PhysicalHealth

Chartera

ndCa

llforA

ction

31

10

16

(41)R

ecom

mendatio

nsfrom

MentalH

ealth

Europe

(MHE),200

9WorkProgrammeo

fthe

Spanish

-Belg

ian-Hun

garia

nTrio

Presidency

ofthe

Cou

ncilof

theE

U(2010-2011)

31

10

16


Table3:Con

tinued.

Docum

ent

Mentalh

ealth

inthew

orkplace

referred

toin

the

objectives

and

scop

eofthe

policy

Coverageo

fexpo

sure

factorsin

relatio

nto

mental

health

inthe

workplace

Coverageo

fmental

health

prob

lemsa

tworkandrelated

outcom

es

Coverageo

frisk

assessmentaspects

inrelationto

mentalh

ealth

inthew

orkplace

Coverageo

fpreventiv

eactions

inrelationto

mentalh

ealth

inthew

orkplace

Overall

(max.25)

(42)R

ecom

mendatio

nsof

theE

urop

eanPa

rliam

ent

andof

theC

ouncil,

2006

,onkeycompetences

for

lifelo

nglearning

12

10

15

(43)C

ouncilof

theE

urop

eanUnion

Con

clusions,

2011,on“Th

eEurop

eanPactforM

entalH

ealth

and

Wellbeing

-ResultsandFu

ture

Actio

n”2

00

12

5

(44)C

ouncilDecision

2003/C

218/01,on

Setting

Up

anAd

visory

Com

mittee

onSafetyandHealth

atWork

11

01

14

(45)C

ouncilRe

solutio

n20

00/C

218/03,onactio

non

health

determ

inants

11

00

13

(46)C

ouncilof

theE

urop

eanUnion

,200

0,Lisbon

Strategy:tobecomethe

mostcom

petitivea

nddynamic

know

ledge-basedecon

omyin

thew

orld

capableo

fsustainablee

cono

micgrow

thwith

morea

ndbette

rjob

sandgreatersocialcoh

esion

01

01

13

(47)C

ouncilof

theE

urop

eanUnion

Con

clusions,

2001,onaC

ommun

itystr

ategyto

redu

cealcoho

l-related

harm

11

00

13

(48)F

ramew

orkAgreemento

nTelewo

rk,200

2,Eu

ropean

socialpartners—ET

UC,

UNICE

(BUSINES

SEURO

PE),UEA

PME,

andCE

EP0

10

11

3

(49)O

pinion

oftheC

ommittee

oftheR

egions,200

6,on

theP

ropo

salfor

aRecom

mendatio

nof

the

European

Parliam

entand

oftheC

ouncilon

Key

Com

petences

forL

ifelong

Learning

02

00

13

(50)C

ommiss

ionRe

commendatio

n20

08/867

/ECon

thea

ctiveinclusio

nof

peop

leexclu

dedfro

mthelabou

rmarket

01

10

13

(51)G

uida

nce:Eu

rope

anCom

miss

ion,

2009

,Report

ofAd

Hoc

ExpertGroup

ontheT

ransition

from

Institu

tionaltoCom

mun

ity-based

Care

11

00

13

(52)F

ramew

orkAgreemento

nInclusiveL

abou

rMarkets,2010Eu

ropean

socialpartners—ET

UC,

UNICE(BUSINES

SEURO

PE),UEA

PME,

andCE

EP0

20

01

3

(53)C

ommun

icationfrom

theC

ommiss

ionCOM

(2010)

2020

EURO

PE2020:A

Strategy

forS

mart,

Susta

inable,

andInclu

siveG

rowth

02

00

13


Table3:Con

tinued.

Docum

ent

Mentalh

ealth

inthew

orkplace

referred

toin

the

objectives

and

scop

eofthe

policy

Coverageo

fexpo

sure

factorsin

relatio

nto

mental

health

inthe

workplace

Coverageo

fmental

health

prob

lemsa

tworkandrelated

outcom

es

Coverageo

frisk

assessmentaspects

inrelationto

mentalh

ealth

inthew

orkplace

Coverageo

fpreventiv

eactions

inrelationto

mentalh

ealth

inthew

orkplace

Overall

(max.25)

(54)C

ouncilof

theE

urop

eanUnion

Con

clusions,

2011,onclo

singhealth

gaps

with

intheE

Uthroug

hconcertedactio

nto

prom

oteh

ealth

ylifesty

lebehaviou

rs1

10

01

3

(55)C

ouncilRe

solutio

n20

00/C

218/02,onthe

balanced

participationof

wom

enandmen

infamily

andworking

life

01

00

12

(56)F

ramew

orkof

Actio

nsforthe

Lifelong

Develop

mento

fCom

petenciesa

ndQua

lificatio

ns,

2002

European

socialpartners—ET

UC,

BUSINES

SEURO

PE,U

EAPM

E,andCE

EP

01

00

12

(57)F

ramew

orkof

Actio

nson

GenderE

quality

,200

5,Eu

ropean

socialpartners—ET

UC,

UNICE

(BUSINES

SEURO

PE),UEA

PME,

andCE

EP0

10

01

2

(58)E

C20

07-W

hitepa

pero

naS

trategyforE

urop

eon

Nutrition,

Overw

eightand

ObesityRe

lated

Health

Issues

01

00

12

(59)O

pinion

oftheC

ommittee

oftheR

egions

2008

onFlexicurity

01

00

12

(60)W

HOEu

rope

anMentalH

ealth

Strategy,2011

10

01

02


0.45 1.52

0.18

1.31.39

0.97

2 3

0

4

4

2.6

2.31

2.5

1.84

1.92.45

2.2

4

5

5

5

5

4.8

Mental health in the workplacereferred to in the objectives/scope

of the policy

Coverage of exposure factors inrelation to mental health in the

workplace

Coverage of mental healthproblems/disorders at work and

related outcomes

Coverage of risk assessmentaspects in relation to mental health

in the workplace

Coverage of preventive actions inrelation to mental health in the

workplace

Overall

Binding/regulatory policy instrumentsDirective 89/391/EEC the European Framework Directive on Safety and Health at WorkNonbinding/voluntary policy initiativesPRIMA-EF: Guidance on the European framework for psychosocial risk management∗

Figure 1: Gap analysis on coverage dimensions across binding and nonbinding policies. ∗Note: The EC 1999 Guidance on work-relatedstress: Spice of life or kiss of death? EU-OSHA 2002 Guidance on How to Tackle Psychosocial Issues and Reduce Work-Related Stress; ILO1986 Guidance on Psychosocial Factors at Work: Recognition and Control; and the ILO, 2012, SOLVE Guidance had the same score as thePRIMA-EF guidance on each dimension.

since mental health and psychosocial risks in the workplacerepresent a constant priority in Europe and other countries.The review and gap analysis also shows that higher scoreshave been assigned to nonbinding (or soft law) policies.Indeed no binding policy achieved a score higher than 2.5,while several voluntary policies achieved scores of 4.5 andhigher.This certainly reflects the focus of the specific policiesas well as their development process and regulatory nature.

Binding policies are the outcome of lengthy negotiationsamong various stakeholders. Depending on the issue at handand the extent to which it is considered controversial, thetext of the policy will reflect this. It is not surprising to seeless coverage of the review dimensions in binding regulationdue to the lack of agreement on psychosocial issues amongsocial partners and their perceived “sensitivity,” however gapsin terms of definitions and terminology cannot be ignored.As discussed previously, these issues have been raised in theliterature and there are several calls for clarifying the textof binding policies further through the inclusion of specificterms (such as work-related stress, psychosocial risks, andmental health at work). While from our review it can be

seen that there is more coverage of exposure factors, riskassessment, aspects and preventive action, this is still limitedin comparison to nonbinding policies.

On the other hand, voluntary policies are often developedby experts alone and usually do not involve negotiationbut rather a review process (which could still involve allrelevant stakeholders). They are more focused in terms ofaddressing specific issues andoften aimat providing guidanceon implementing good practice. As a result, terminology inthese policies is more specific and inclusive and coverage ofkey elements is more extensive (as also shown in Figure 1).

It is important to note that this review provides anoverview on the basis of the content of policies in this area.However, it does not draw any conclusions on the uptake andimpact of these policies in practice. Two key issues concernthe extent to which these policies offer specific guidance onmanaging risks in relation to mental health in the workplaceto enable organisations (and especially small and medium-sized enterprises) to implement a preventive framework ofaction and whether existing policies have actually fulfilledexpectations in practice in the area of mental health in


the workplace. Naturally, one would expect that the bindingnature of regulation means that they would be adopted morein practice. However, recent findings suggest that althoughoccupational health and safety legislation is seen by Europeanemployers as a key driver to address health and safety issues,it has been less effective for the management of psychosocialrisks and the promotion of mental health in the workplace[27, 40].

In relation to voluntary policy instruments, there is thequestion of whether they have been effective in supportingthe implementation of existing legislation and in guaran-teeing quality with regard to the “essential requirements”established by European binding policies. Unfortunately, verylittle evaluation exists in this area and it is difficult to drawany meaningful conclusions. A meaningful example in thisdirection comes from the last report on the implementationof the European framework agreement onwork-related stresssigned by the representatives of European social partners[41]. This agreement has had important positive effects,accelerating social dialogue and the development of policieson work-related stress in most of the EU countries. After10 years from signing, it has been implemented in most ofthe countries of the EU in different ways: being translatedin 8 countries, leading to the signing of national agreementswith social partners in 9 countries, being implemented innational legislation in 9 countries. In addition, an evaluationof soft lawwould not be sufficient unless national policies andrelevant initiatives were also taken into account. Traditionsof national level research into occupational health and safetyin general and specifically in relation to psychosocial risksand their management, national discourses on health andsafety definitions and priorities socially and politically, andthe practical application of research knowledge to workplacepractice are also important determinants of action in thisarea [42]. However, ESENER results do indicate low actionof European organisations and further need for guidanceand support [43]. Given the number of voluntary initiatives,one would expect further uptake at company level, andquestions in relation to effectively communicating theseto organisations/employers in a user-friendly manner, orhighlighting positive benefits, are relevant in order to engagethem in action.

According to the findings of our review and the widercontext of policy-making in Europe, if the status quo con-cerning the policy context to mental health in the workplaceis maintained, it is likely that a number of initiatives willcontinue to take place across the EU in this area, given theimpact of mental ill health on individuals, organisations,and society. However, there is uncertainty as to whetherthey will achieve the desired outcomes. Although there havebeen a number of policy initiatives for more than ten yearsin the EU, awareness in relation to mental health in theworkplace and the importance of preventive action still seemsto be lacking on the whole and especially among SMEs. Thisis despite the available data that map the prevalence andimpact both of risk factors and mental ill health outcomes.In addition, despite the fact that the Framework Directive89/391/EEC covers all types of risk to workers’ health andas the framework agreement on work-related stress clarifies,

including work-related stress, there still appears to be limitedawareness of this provision both by employers and otherkey stakeholders such as policy makers and inspectors indifferent countries. Limited awareness and expertise on howto conduct inspections on psychosocial risks associated withmental ill healthwere among the key drivers for the 2012 SLICcampaign [44]. However, with widespread budget cuts in thepublic sector, inspections in many countries are becomingmore reactive in nature [37].

In light of this, it would be advisable to revisit thecontent of the Framework Directive in relation to psychoso-cial risks and mental health in the workplace to providefurther clarity and harmonize terminology across other keypieces of legislation accordingly. In absence of this, a clearinterpretation of the legal provisions in this area by theEuropean Commission would be needed. There is also morescope for better and closer collaboration and coordination toachieve maximum impact in a cost-effective manner at EUinstitutional level since several policy initiatives and studieshave been implemented in this area, for example, from differ-ent EC Directorate Generals, the European Parliament, andthe European Agency for Safety & Health at Work. Finally, itis important that employer responsibility is strengthened andawareness is further developed both in relation to the policyframework on mental health in the workplace and specificpreventive measures that should be introduced to promotemental health, and the promotion of soft law initiatives isessential towards this end.

5. Conclusions

Mental health and psychosocial risks in the workplace havebeen recognised as priorities in occupational health andsafety in the European Union for at least two decades. Anumber of hard and soft law policies of relevance to themhave been developed over the years that have promotedawareness and action among policy makers, social partners,organisations, and indeed individual workers. This paperaimed to provide a review and gap analysis of hard andsoft law policies applicable at EU level in this area and offerrecommendations for the future. Our review of ninety-fourpolicies across five key dimensions revealed several gaps,especially in relation to binding in comparison to voluntarypolicies. According to the findings of our review and thewider context of policy-making in Europe, if the statusquo as concerns the policy context to mental health in theworkplace is maintained, it is uncertain whether desiredoutcomes will be achieved in practice since awareness inrelation tomental health in theworkplace and the importanceof preventive action still seems to be lacking. It is thereforerecommended that key EU legislation is made clearer in thisarea by either including specific terminology and harmo-nizing it across other key pieces of legislation accordinglyor by the development of a clear interpretation of the legalprovisions in this area by the European Commission. It isalso recommended that there is a better coordination at EUinstitutional level to achieve maximum impact and not iso-lated and indeed competitive and non-cost-effective efforts.Finally, it is important that soft law initiatives continue to be


promoted to strengthen employer awareness, responsibility,and engagement in preventive actions.



Acknowledgment

This work was completed with the support of a EuropeanCommission grant by the Directorate General for Employ-ment, Social Affairs and Inclusion.

References

[1] G.Waddell and A. K. Burton, Is Work Good for Your Health andWell-Being?The Stationary Office, London, UK, 2006.

[2] A. Deacon, From Welfare to Work: Lessons from America,Institute of Economic Affairs, 1997.

[3] D. King and M. Wickham-Jones, “From Clinton to Blair:the Democratic (Party) origins of welfare to work,” PoliticalQuarterly, vol. 70, no. 1, pp. 62–74, 1999.

[4] L. M. Mead, The New Paternalism: Supervisory Approaches toPoverty, Brookings Institution, Washington, DC, USA, 1997.

[5] R. Perkins, P. Farmer, and P. Litchfield, Realising Ambitions:Better Employment Support for People with a Mental HealthCondition, Department of Work and Pensions, London, UK,2009.

[6] M. K. Langenhan, S. Leka, and A. Jain, “Psychosocial risks:is risk management strategic enough in business and policymaking?” Safety & Health atWork, vol. 4, no. 2, pp. 87–94, 2013.

[7] H.Westerlund, J. Vahtera, J. E. Ferrie et al., “Effect of retirementon major chronic conditions and fatigue: French GAZELOccupational Cohort Study,” British Medical Journal, vol. 341,Article ID c6149, 2010.

[8] WHO—World Health Organization, Work Organization andStress, Protecting Workers’ Health Series, No. 3, World HealthOrganization, Geneva, Switzerland, 2003.

[9] European Commission (EC), Guidance on Work-RelatedStress—Spice of Life or Kiss of Death? Office for OfficialPublications of the European Communities, Luxembourg City,Luxembourgh, 2002.

[10] European Social Partners, Framework Agreement onWork-Related Stress, European Social Partners—ETUC,UNICE(BUSINESSEUROPE), UEAPME and CEEP, Brussels,Belgium, 2004.

[11] WHO—World Health Organization, PRIMA-EF: Guidance onthe European Framework for Psychosocial Risk Management: AResource for Employers and Worker Representatives, ProtectingWorkers’ Health Series No. 9, World Health Organization,Geneva, Switzerland, 2008.

[12] WorldHealthOrganization (WHO),Health Impact of Psychoso-cial Hazards atWork: AnOverview,WorldHealthOrganization,Geneva, Switzerland, 2010.

[13] European Foundation for the Improvement of Living andWorking Conditions (Eurofound), Fifth European WorkingConditions Survey, Publications Office of the European Union,Luxembourg City, Luxembourgh, 2012.

[14] European Agency for Safety and Health at Work (EU-OSHA),Research onWork-Related Stress, Office for Official Publicationsof the European Communities, Luxembourg City, Luxem-bourgh, 2000.

[15] L. Levi, “Spice of life or kiss of death,” in Working on Stress,Magazine of the European Agency of Safety and Health atWork No. 5, Office for Official Publications of the EuropeanCommunities, Luxembourg City, Luxembourg, 2002.

[16] Organisation for Economic Co-Operation and Development(OECD), Sick on the Job? Myths and Realities about MentalHealth and Work, OECD Publishing, Paris, France, 2012.

[17] H. U. Wittchen, F. Jacobi, J. Rehm et al., “The size and burdenof mental disorders and other disorders of the brain in Europe2010,” European Neuropsychopharmacology, vol. 21, no. 9, pp.655–679, 2011.

[18] European Commission (EC), “Eurobarometer on mentalhealth,” 2010, http://ec.europa.eu/health/mental health/euro-barometers/index en.htm.

[19] EU-OSHA—European Agency for Safety and Health at Work,OSH in Figures: Stress at Work—Facts and Figures, Office forOfficial Publications of the European Communities, Luxem-bourg City, Luxembourg, 2009.

[20] MentalHealth Foundation,TheFundamental Facts, 2007, http://www.mentalhealth.org.uk/publications/fundamental-facts/.

[21] Health and Safety Executive (HSE), Self-ReportedWork-RelatedIllness and Workplace Injuries in 2008/09: Results from theLabour Force Survey, HSE Books, Sudbury, Canada, 2010.

[22] C. Trontin, M. Lassagne, S. Boini, and S. Rinal, Le cout du stressprofessionnel en France en 2007, INRS, Paris Tech, Paris, France,2007.

[23] S. Leka, A. Jain, S. Iavicoli, M. Vartia, and M. Ertel, “Therole of policy for the management of psychosocial risks at theworkplace in the European Union,” Safety Science, vol. 49, no. 4,pp. 558–564, 2011.

[24] L. Levi, “Working life and mental health—a challenge topsychiatry?”World Psychiatry, vol. 4, no. 1, pp. 53–57, 2005.

[25] S. Leka, A. Jain, G. Wetsloot, and T. Cox, “Policy-level inter-ventions andwork-related psychosocial riskmanagement in theEuropean Union,” Work & Stress, vol. 24, no. 3, pp. 298–307,2010.

[26] T.W.Taris, I. van derWal, andM.A. J. Kompier, “Large-scale jobstress interventions: the Dutch experience,” in ContemporaryOccupational Health Psychology: Global Perspectives on Researchand Practice, J. Houdmont and S. Leka, Eds., vol. 1, pp. 77–97,Wiley-Blackwell, Chichester, UK, 2010.

[27] EU-OSHA—European Agency for Safety and Health at Work,European Survey of Enterprises on New and Emerging Risks:Managing Safety and Health at Work. European Risk Observa-tory Report, Office for Official Publications of the EuropeanCommunities, Luxembourg City, Luxembourg, 2010.

[28] European Agency for Safety and Health at Work (EU-OSHA),Second European Survey of Enterprises on New and EmergingRisks (ESENER-2), Publications Office of the European Union,Luxembourg City, Luxembourgh, 2015.

[29] S. Iavicoli, P. Deitinger, C. Grandi, M. Lupoli, A. Pera, and B.Rondinone, “Fact-finding survey on the perception of work-related stress in EU candidate countries,” in Stress at Work inEnlarging Europe, S. Iavicoli, P. Deitinger, C. Grandi, M. Lupoli,A. Pera, and M. Petyx, Eds., pp. 81–97, ISPESL, Rome, Italy,2004.


[30] S. Iavicoli, E. Natali, P. Deitinger et al., “Occupational healthand safety policy and psychosocial risks in Europe: the role ofstakeholders’ perceptions,” Health Policy, vol. 101, no. 1, pp. 87–94, 2011.

[31] M. Ertel, U. Stilijanow, S. Iavicoli, E. Natali, A. Jain, andS. Leka, “European social dialogue on psychosocial risks atwork: benefits and challenges,” European Journal of IndustrialRelations, vol. 16, no. 2, pp. 169–183, 2010.

[32] EU-OSHA—European Agency for Safety and Health at Work,Mental Health Promotion in the Workplace—A Good PracticeReport, Office for Official Publications of the European Com-munities, Luxembourg City, Luxembourg, 2011.

[33] European Commission (EC), “Communication from the Com-mission to the European Parliament, the Council, the Euro-pean Economic and Social Committee and the Committeeof Regions on the practical implementation of the provi-sions of the Health and Safety at Work Directives 89/391(Framework), 89/654 (Workplaces), 89/655 (Work Equipment),89/656 (Personal Protective Equipment), 90/269 (Manual Han-dling of Loads) and 90/270 (Display Screen Equipment),”2004, http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=CELEX:52004DC0062:EN:HTML.

[34] EuropeanCommission (EC), “Report on the implementation ofthe European social partners—framework agreement on work-related stress,” Commission Staff Working Paper SEC(2011) 241Final, European Commission (EC), Brussels, Belgium, 2011.

[35] European Social Partners, Implementation of the EuropeanAutonomous Framework Agreement on Harassment and Vio-lence at Work—Final Joint Report by the European Social Part-ners, European Social Partners—ETUC, BUSINESSEUROPE,UEAPME and CEEP, Brussels, Belgium, 2011.

[36] A. Branch, “The evolution of the European social dialoguetowards greater autonomy: challenges and potential benefits,”International Journal of Comparative Law and Industrial Rela-tions, vol. 21, no. 2, pp. 321–346, 2005.

[37] S. Leka, W. Van Wassenhove, and A. Jain, “Is psychosocial riskprevention possible? Deconstructing common presumptions,”Safety Science, vol. 71, pp. 61–67, 2015.

[38] S. Iavicoli, S. Leka, A. Jain et al., “Hard and soft law approachesto addressing psychosocial risks in Europe: lessons learnedin the development of the Italian approach,” Journal of RiskResearch, vol. 17, no. 7, pp. 855–869, 2014.

[39] J. Ritchie and J. Spencer, “Qualitative data analysis for appliedpolicy research,” in Analysing Qualitative Data, A. Bryman andR. G. Burgess, Eds., pp. 173–194, Routledge, London, UK, 1994.

[40] E. Natali, P. Deitinger, B. Rondinone, and I. Iavicoli, “Exploringstakeholders’ perceptions on social policies, infrastructures andsocial dialogue in relation to psychosocial risks,” in The Euro-pean Framework for Psychosocial Risk Management: PRIMA-EF, S. Leka and T. Cox, Eds., pp. 79–95, I-WHO Publications,Nottingham, UK, 2008.

[41] European Parliament, Occupational Health Concerns: Stress-Related and Psychological Problems Associated with Work,European Parliament, Directorate-General for Internal Policies,2013.

[42] European Agency for Safety and Health at Work (EU-OSHA),Analysis of the Determinants of Workplace Occupational Safetyand Health Practice in a Selection of EU Member States,Publications Office of the European Union, Luxembourg City,Luxembourgh, 2013.

[43] European Agency for Safety and Health at Work (EU-OSHA),Drivers and Barriers for Psychosocial Risk Management: An

Analysis of Findings of the European Survey of Enterprises onNewand Emerging Risks, Publications Office of the European Union,Luxembourg City, Luxembourgh, 2012.

[44] SLIC—The Committee of Senior Labour Inspectors,Psychosocial Risk Assessments—SLIC Inspection Campaign 2012,2012, http://www.av.se/dokument/inenglish/European Work/Slic 2012/SLIC2012 Final report.pdf.

Research ArticleWorkplace Bullying as a Risk Factor forMusculoskeletal Disorders: The Mediating Role ofJob-Related Psychological Strain

Michela Vignoli,1 Dina Guglielmi,2 Cristian Balducci,3 and Roberta Bonfiglioli4

1Department of Psychology, Alma Mater Studiorum-University of Bologna, Viale Berti Pichat 5, 40127 Bologna, Italy2Department of Educational Science, Alma Mater Studiorum-University of Bologna, Via Filippo Re 6, 40126 Bologna, Italy3Department of Political and Social Sciences, Alma Mater Studiorum-University of Bologna, Via dei Bersaglieri 6/c,40125 Bologna, Italy4Occupational Medicine, Department of Medical and Surgical Sciences, Alma Mater Studiorum-University of Bologna,Via Palagi 9, 40138 Bologna, Italy

Correspondence should be addressed to Michela Vignoli; [email protected]

Received 16 January 2015; Accepted 29 March 2015

Academic Editor: Steven L. Sauter

Copyright © 2015 Michela Vignoli et al.This is an open access article distributed under the Creative CommonsAttribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Workplace bullying is considered by the European Agency for Safety and Health at Work one of the emerging psychosocial riskfactors that could negatively affect workers’ health.Thus, the aim of this study was to analyze the process that leads from bullying tonegative health (such as musculoskeletal disorders (MSDs)), testing the mediating role of job-related strain. Data were collected on512 workers (62.9% female; mean age = 43.6 years) of a retail chain who filled in a self-report questionnaire after a one-hour trainingsession on work-related stress. Data analyses were performed controlling for potentially confounding variables (i.e., gender, age,organizational role, type of contract, and perceived physical job demands). Preacher andHayes analytical approach was used to testthe indirect relationship between bullying and MSDs. Results showed that work-related strain mediates the relationship betweenbullying and MSDs considered (low back, upper back, and neck) except for MSDs of the shoulders. Our study confirms the roleplayed by bullying and job-related strain in determining workers’ MSDs.

1. Introduction

Increasing attention has been paid in the past 15 to 20 yearsto the phenomenon of workplace bullying; in some countriesit is also called mobbing [1]. Workplace bullying refers toa series of negative behaviours carried out frequently andover a prolonged period of time, usually against an individualemployee by his or her colleagues or superior [2]. Examplesof such negative behaviours are as follows: excessive criticismof one’s work; withholding of information, which affectsperformance; being assigned an unmanageable workload;spreading of rumours; and social isolation.

Bullying is an escalating process in the course of whichthe person confronted ends up in an inferior position andbecomes the target of systematic negative social acts. There-fore, a conflict cannot be called bullying if the incident is an

isolated event or if it involves two parties of approximatelyequal strength [2]. The consequences of exposure to bullyingmay be traumatic for the affected individual [3, 4]. Prevalenceestimates of bullying are difficult due to a lack of an agreedupon definition of the phenomenon. A recent Europeansurvey [5] estimated a prevalence of 4% among Europeanworkers. However, in the same survey, 11% of workersreported they were the subject of verbal abuse at work, whichmay also be considered a form of bullying. According toothers, the prevalence of bulling may be even higher: 15%of workers may be affected at any point in time [6]. Despitethis lack of convergence on prevalence estimates, there issubstantial agreement that workplace bullying is an emergingpsychosocial risk with the potential to adversely affect thesafety and health of working people [7].


http://dx.doi.org/10.1155/2015/712642


Most studies in this area have investigated psychologicalhealth outcomes of exposure to bullying, documenting asignificant relationship between bullying and psychosocialstress, leading to anxiety and depression, including the onsetof major depressive episodes [8–12]. It is now quite clearthat exposure to bullying can lead to a profound deteriora-tion of the person’s psychological health, mainly via stressexperiences [13]. Few studies, however, have investigatedthe potential impact of bullying on outcomes other thanpsychological ones. Thus, it remains to be seen whetherbullying has the same far-reaching health effects as those, forexample, of well-established psychosocial factors, such as jobstrain or effort-reward imbalance, which have been foundto deteriorate not only to psychological but also to physicalhealth conditions [14]. Furthermore, researchers have notedthat studying the relationship between psychosocial factors,such as bullying, which are usually assessed through self-reports, and psychological outcomes, may be particularlysubjected to common method bias due to personal factorssuch as negative affectivity, which may act as a critical con-founding variable [15].This further strengthens the relevanceof assessing the potential effect of bullying on different kindsof health-related outcomes.

To address the gap in the literature presented above, inthe present study we investigate the relationship betweenexposure to bullying and very commonwork-related physicalhealth problems, namely, musculoskeletal disorders (MSDs).MSDs are dysfunctions affecting muscles, bones, nerves,tendons, ligaments, joints, cartilages, and spinal discs; theyare defined by sprains, strains, tears, soreness, pain, periph-eral nerve disorders, and connective tissue injuries of thestructures previously mentioned [16]. MSDs are the mostoften reported health problem by workers in the EuropeanUnion: 24.7% of them report back pain and 22.8% reportmuscular pain in shoulders, neck, upper or lower limbs, orcombinations of any or all of these. In theUnited States,MSDsare one of themain reasons for short- and long-termdisabilityand early retirement [17, 18].

The most common antecedents of MSDs are biome-chanical factors, such as repetitive motion, excessive force,awkward postures, and prolonged sitting and standing [16].However, psychosocial factors are also believed to be impor-tant for both the initial development of MSDs and thelong-term disability that may follow [18–22]. While theprecise mechanisms (e.g., cognitive, neuroendocrine, andmusculoskeletal) through which psychosocial factors mayaffect MSDs have not been fully elucidated, an acceptedhypothesis [23] is that psychosocial factors may operateindirectly. They may, for example, influence muscle tensionor other physiological processes and decreasing micropausesinmuscle activity and, as a consequence, affect the perceptionof pain. Plausibly, such indirect effect is exerted through theexperience of work-related stress.

Most research on the impact of psychosocial factors onMSDs has focused on factors, such as psychological jobdemands and job control [24]. A review of the availableevidence suggests that such factors (i.e., high demands andlow control) are indeed related to MSDs, specifically ofthe neck, shoulder, and back [25]. As far as exposure to

bullying is concerned, we traced two studies exploring itsrelationship with MSDs. A study on 370 Lithuanian seafarerspublished as a conference abstract revealed that exposure tobullying was significantly associated with an overall measureof upper limb MSDs [26]. Another study conducted on1024 employees of a Norwegian bus company revealed anassociation between exposure to bullying and a measure ofmusculoskeletal complaints including headache, backache,neck ache, and hand and foot pain [27]. However, the latterstudy did not control for potentially confounding factors,such as physical load factors. Furthermore, neither study fol-lowed recent recommendations emphasizing the importanceof investigating specific forms of MSDs [25].

Thus, in the present study, we further investigate therelationship between exposure to bullying andMSDs by con-trolling for potentially confounding factors and focusing onspecific musculoskeletal problems. Furthermore, we explorewhether job-related strain may indeed act as a mediator inthe relationship between exposure to bullying and MSDs,as Silverstein and Evanoff [23] hypothesized and, indeed, asSprigg et al. [24] found for other psychosocial risk factors.

2. Methods

2.1. Study Design and Sample. A cross-sectional survey wasconducted in a large retail company in Italy. A total of 553workers voluntarily participated in the study, after researchersobtained a randomized sample from the organization’s 812workers (68.1% was the response rate). All participantsworked in grocery stores belonging to the same organization;therefore all of them have the same procedures and companyregulations. The sample was composed of both supervisorsand employees. Participants worked in different departmentsof the supermarkets (e.g., gastronomy, fruit and vegetables,butchery, fish, bakery, cashiers, and nonfood); thus they allperform job activities with high physical demands.

Workerswere assembled in different groups and, after onehour of training on work-related stress, they completed ananonymous, self-administered questionnaire. The contentsof this brief training session were the main European andnational regulations about work-related stress and the maindefinitions of work-related stress used in the literature. Thistraining hour was included before filling the questionnairein order to explain to the workers that the aim of the studywas not to define how much they were stressed, but only tounderstand which psychosocial risk factors could contributeto enhancing strain and decreasing workers’ health.

2.2. Measures. Workplace bullying is normally assessedeither by using the respondents’ feeling of being victimizedby bullying (e.g., [9]), usually according to a given specificdefinition of the phenomenon, or according to the respon-dents’ perception of being exposed to a range of specificbullying behaviours described without explicit reference tothe term bullying (e.g., [28]).The first method is the so-calledself-labelling approach, which, however, is very subjectiveand strongly influenced by personality and emotional and


cognitive factors, including possible misperception. The sec-ond method is the behavioural experience method, which isgenerally believed to be more objective because it is relativelyless exposed to the effect of personal factors. Thus, in thepresent study, we used the latter approach and assessedbullying with the Italian version of the Short Negative ActsQuestionnaire (S-NAQ) [29], which has been validated inItaly with an ad hoc study [30]. The scale consists of 9items investigating howoften the respondent has experienceda variety of negative behaviours at work during the lastsix months. One example item is “Someone withholdinginformation, which affects your performance” and workerscould answer on a 5-point Likert scale ranging from 1 (never)to 5 (daily). Items were then averaged.The S-NAQ has shownpsychometric properties using Italian data, which are entirelycomparable to those of the original and longer (i.e., 22-item)version, for example, in terms of associations with variablesof mental health and well-being [30].

Job-related strain was measured through the dimensionof emotional exhaustion of the Maslach Burnout InventoryGeneral Survey (MBI-GS: [31]; Italian version [32]). The 5-item scale was scored on a 7-point frequency Likert scale (0= never to 6 = every day). One example item is as follows:“I feel emotionally drained from my work.” Items were thenaveraged.

Musculoskeletal disorders were measured through 4items related to four different parts of the body: low back,upper back, neck, and shoulders. The question was, “Duringthe past 12 months have you had pain, aching, stiffness,burning, numbness, or tingling (“pins and needles”) in anyareas of the following that occurred more than three times orat least more than a week?” The possible answers were eitheryes or no.

In addition to those variables, possible confoundingvariables were included: gender, age, organizational role, typeof contract. Furthermore, as participants were working ina large retail company, we introduced physical job demandmeasured with the Italian version [33] of Karasek’s [34]Job Content Questionnaire as a control variable. The scaleconsists of 5 items with response options ranging from 1(strongly disagree) to 4 (strongly agree). One example item is“I am often required to move or lift very heavy loads on myjob.” Items were then averaged.

2.3. Statistical Analysis. Logistic regression models werefitted to the data by using the software SPSS version 20.0.The risk factor was bullying, while the outcome variableswere four specific MSDs of the low back, upper back, neck,and shoulders. To test for the mediating role played by job-related strain (i.e., emotional exhaustion) in the relationshipbetween exposure to bullying and MSDs, we adopted thePreacher and Hayes [35] analytical approach. This approachtests the indirect relationship between an exposure factorand an outcome through a mediator by using a bootstrap(i.e., resampling) procedure that addresses some weaknessesassociated with the Sobel test [35]. To compute the directand indirect effects, all path coefficients in the model were

estimated concurrently. Furthermore, the bootstrapping pro-cedure was used to compute formal statistical tests of thespecific indirect effects.Thismethod can produce an estimateof the indirect effect, including a 95% confidence interval.When 95% confidence interval does not include zero, theindirect effect is significantly different between the levelof zero and 0.05. Four different mediation analyses wereperformed, one for each specific MSD, that is, for the lowback, upper back, neck, and shoulders.

3. Results

3.1. Demographic and Working Characteristics of Subjects.Due to missing data, 41 cases were deleted; thus, the finalsample consisted of 512 Italian workers. Most of them (322workers, 62.9%) were female and the mean age was 43.64years (SD=7.8).Themeanoccupational tenurewas 16.15 years(SD = 8.46). Concerning the type of contract, 52.3% had apart-time contract, while all other workers had a full-timecontract. Concerning the organizational role, 94 workers(18.4%) were supervisors, while 418 were employees (81.6%).

3.2. Descriptive Statistics, Correlations, and Job-Related StrainMediation Effect between Bullying and MSDs. Means, stan-dard deviations, percentages, internal consistencies, and cor-relations were computed for all the study variables (Table 1).Internal consistencies (Cronbach’s 𝛼) of the used scales weregood, as all the values exceeded the threshold of 0.70 [36].Exposure to bullying behaviours was relatively low, meaningthat, on average, employees only occasionally experiencedthose negative acts that are the essence of bullying (Table 1).The obtained value of 1.67 at the bullying measure is similarto that commonly found in organizational research in thisarea in which the same operationalization of bullying is used[37, 38]. A closer inspection of the distribution of the bullyingvariable revealed that 3.51% of employees (not reported inTable 1) reported a score indicating an exposure on a weeklyor daily basis to the bullying behaviours investigated.

On the contrary, job-related strain and physical demandwere relatively more prevalent, with their average levels (i.e.,17.30 and 2.71, resp.) being above the central point of theadopted response scale. For example, a score of 2.71 at thephysical demand scale meant that all the five investigatedaspects describing a high physical demand tended to bereported by most of participants. As far as musculoskeletalproblems are concerned, in general they were highly preva-lent among participants, with the highest prevalence being forthe low back problems.

Furthermore, results, presented in Table 1, showed that,among the confounding variables (age, gender, organisationalrole, type of contract, and physical demands), all of themwererelated to at least one of the outcome variables considered(MSDs of low back, upper back, neck, and shoulders).Thus, these confounding variables have been included in themediation analysis.

In order to test our hypothesis, which postulates thatstrain mediates between bullying and MSDs, four mediationanalyses have been performed. As mentioned before, the


Table1:Means,stand

arddeviations,percentages,C

ronb

ach’s𝛼,and

correlations

(𝑁=512).

Varia

bles

𝑁item

M(SD)

%𝛼

12

34

56

78

910

(1)A

ge+

—43.64(7.8)

——

—(2)G

ender(1=

male)

+—

—37.1

—−0.022

—(3)R

ole(1=

employee)+

——

81.6

—−0.109∗−0.315∗∗∗

—(4)C

ontract(1=

part-time)

+—

—52.3

—−0.256∗∗∗

−0.465∗∗∗

0.40

6∗∗∗

—(5)P

hysic

aldemand+

52.71

(0.73)

—0.77

0.019−0.095∗

0.069

0.06

4—

(6)B

ullying

91.6

7(0.69)

—0.87

0.06

60.108∗

−0.087−0.111∗

0.238∗∗∗

—(7)Job

-related

strain

517.30

(7.29)

—0.78

0.157∗∗∗

−0.105∗

0.04

80.011

0.44

7∗∗∗

0.327∗∗∗

—(8)L

owback

(1=yes)

——

74.6

—0.015−0.165∗∗∗

0.129∗∗

0.081

0.347∗∗∗

0.169∗∗∗

0.285∗∗∗

—(9)U

pper

back

(1=yes)

——

53.1

—0.090∗−0.162∗∗

0.111∗

0.099∗

0.308∗∗∗

0.214∗∗∗

0.294∗∗∗

0.387∗∗∗

—(10)N

eck(1=yes)

——

60.5

—0.084−0.273∗∗∗

0.113∗

0.110∗

0.266∗∗∗

0.148∗∗

0.263∗∗∗

0.374∗∗∗

0.403∗∗∗

—(11)S

houlders(1=yes)

——

56.6

—0.192∗∗∗

−0.160∗∗∗

0.053

0.049

0.351∗∗∗

0.141∗∗

0.253∗∗∗

0.259∗∗∗

0.363∗∗∗

0.44

7∗∗∗

Notes:+confou

ndingvaria

bles;∗∗∗

𝑃<0.001;∗∗

𝑃<0.01;∗𝑃<0.05.


Table 2: Direct effects (𝑁 = 512).

Mediator Outcome variables MSDStrain Low back Upper back Neck Shoulders

Bullying 2.539∗∗∗ 0.474∗ 0.534∗∗ 0.371∗ 0.209Strain — 0.046∗∗ 0.042∗∗ 0.040∗ 0.021Physical demand 3.795∗∗∗ 0.888∗∗∗ 0.634∗∗∗ 0.524∗∗∗ 0.959∗∗∗

Age 0.126∗∗∗ −0.003 0.022 0.017 0.056∗∗∗

Gender (1 = male) −1.261 −0.671∗ −0.506∗ −1.169∗∗∗ −0.571∗

Role (1 = employee) 0.604 0.535 0.366 0.224 0.066Contract (1 = part-time) −0.062 −0.132 0.222 −0.042 0.108Notes: coefficients are not standardized. ∗∗∗𝑃 < 0.001; ∗∗𝑃 < 0.01; ∗𝑃 < 0.05.

Preacher and Hayes [35] analytical approach allowed us totest the direct and indirect effects of the variables considered.Thus, we provided estimates of all the path coefficients(Table 2), as well as indirect effects (Table 3) along withthe 95% bias-corrected, bootstrapped confidence intervalsfor the four different musculoskeletal disorders (low back,upper back, neck, and shoulders). Specifically, in Table 2 bothresults concerning the direct effects of the antecedent andconfounding variables on the mediator (job-related strain)and results concerning the direct effects of the antecedents,confounding variables, and the mediator on the outcomes(MSDs of low back, upper back, neck, and shoulders) arepresented.

Thus, concerning the direct effects, bullying has a positiveeffect on strain and on all the MSDs considered, except forMSD of the shoulders. This means that the more workersare exposed to bullying, the more they report MSDs of thelow back, upper back, and neck. Also, work-related strain isdirectly related to all MSDs, except for shoulders. Lookingat the possible confounding variables, perceived physicaldemand has an effect both on strain and on all MSDs, whileage affects strain and only MSD of the shoulders. Regardinggender, females report more MSDs but not higher strain.Organizational role and type of contract seem to not have aneffect on either strain or MSDs.

Results concerning the indirect effects between theindependent variable (bullying) and the outcomes variables(MSDs of low back, upper back, neck, and shoulders) arepresented in Table 3. Results show that job-related strainmediates the relationship between bullying and all MSDs,except for MSDs of the shoulders. Those results mean that,except for the MSD of shoulders, strain helps in under-standing the process between bullying and musculoskeletaldisorders, as results presented in Table 3 show that bullyingaffects strain which in turn affects MSDs (low back, upperback, and neck).

4. Discussion

Even though psychosocial risk factors have been found to beimplicated in the development of MSDs (see, for a review,[20]), most studies in this area have been inspired by Karaseket al.’s [39] psychosocial model and have investigated the roleof psychological job demand (i.e., workload) and decision

latitude (i.e., job control) on MSDs [24]. Having to dowith the tasks performed by the worker, job demands anddecision latitude are typical job content factors (see EuropeanAgency for Safety and Health at Work [40]). Psychosocialcontextual factors, such as those describing the quality ofrelationships at work, have rarely been examined in detail.As far as workplace bullying is specifically concerned, only afew studies have explored the relationship between exposureto such contextual factors and MSDs [26, 27]. However, suchstudies have not adopted a fine-grained approach on MSDsor included an overall index of MSDs, which is less infor-mative and generally not recommended [25]. Furthermore,there is a substantial lack of knowledge about the possiblemechanisms for explaining the link between psychosocialfactors and MSDs. The experience of psychological strainhas been hypothesised as one such mechanism [23], but itsinvolvement has rarely been directly explored.

Our results confirm that exposure to bullying behaviouris linked to MSDs (in the low back, upper back, and neckregions). Only the shoulders do not seem affected by thismediation. The results suggest that, along with the directeffect between bullying and MSDs (low back, upper back,and neck), there is a process which comprises job-relatedstrain between workplace bullying and MSDs. Therefore thisrelationship ought to be explained by both the direct effectof bullying as a psychosocial factor and the indirect effectof psychological strain manifesting as MSDs. Furthermore,despite physical demands remaining the main predictor ofMSDs, when strain is considered, the effect of bullying onMSDs is quite similar (especially on the basis of the upperback and neck).

Seeing that exposure to bullying can lead to a profounddeterioration of the victim’s psychological health mainly viathe experience of stress [13], the same mechanism seems toalso influence physical health, specifically MSDs. FormerlyVie et al. [27] found both positive and negative emotionsmediate the relationship between exposure to bullying andmusculoskeletal complaints, even if it seems that negativeemotion, namely, stress, is the main mediator. In line withthis study, to our knowledge, this is the first direct evi-dence of job-related strain as a mediator between bullyingand MSDs. Therefore, the strain process, which notoriouslymay affect the body, for example, by producing tensionin the musculature, is one of the elements to consider as


Table 3: Indirect mediation effects of work-related strain between bullying and MSDs (𝑁 = 512).

Low back Upper back Neck ShouldersEffect C.I. Effect C.I. Effect C.I. Effect C.I.

Mediation 0.117 0.025; 0.236 0.106 0.027; 0.209 0.102 0.024; 0.206 0.053 −0.030; 0.143Notes: bootstrap confidence intervals were constructed using 5000 samples. When 95% confidence interval did not include zero, the indirect effect issignificantly different between the level of zero and 0.05.

we comprehend the detrimental effects of bullying on thevictims’ health. Note that we only found evidence for a partialmediation by psychological strain, since in the three caseshad psychological strain acted as a mediator (i.e., of pain inthe low back, upper back, and neck), bullying would haveremained a significant risk factor for the investigated MSDin the final model.

One explanation for the direct effect between bullying andMSDs could be that we operationalized psychological strainin terms of emotional exhaustion, which mainly taps low-arousal symptoms, such as feelings of fatigue and depression,thus capturing only certain kinds of manifestation of psycho-logical strain. High-arousal symptoms such as anxiety andirritability, which are not well represented in the emotionalexhaustion construct, may be evenmore critical in mediatingthe effect of bullying on MSDs. This is because bullyinghas been shown to generate strong feelings of anxiety and,eventually, disorders in those who are exposed [3]; at thesame time, anxiety has been found to be one of the strongeraffective mediators of the relationship between psychosocialaspects of work and MSDs [41]. In brief, there is room tobelieve that the psychological strain generated by exposureto bullying may have an even more important role in theoccurrence ofMSDs than that found in the present study.Thissuggests the need for more research in this area.

One of the main strengths of this study is the focuson workplace bullying as a psychosocial risk factor forMSDs. Even though NIOSH [16] considers these healthcomplaints an important occupational disease, relative toother psychosocial risk factors, they are still understudied.Another strong point is represented by the fact that workcharacteristics, workplace bullying, stress, and MSDs arestudied together. Usually, the relationships between workcharacteristics, bullying, and stress find evidence in stress orpsychological literature, whereas the relationships betweenwork characteristics and MSDs are predominantly foundwithin the medical, ergonomic, and epidemiological fields[24].

These strong points, however, do have some limitationsthat should be mentioned. First, the sample was not repre-sentative of a working population or of workers in the retailsector, which might decrease the opportunity to generalizethe obtained results. A second limitation of the present studyis that it is cross-sectional, so we cannot strengthen the basisfor causal inference regarding MSDs. Therefore, adoptinga rigorous longitudinal research design would reduce thelikelihood of the findings having arisen due to chance andwould allowus to investigate the effective impact that bullyinghas on workers who develop MSDs. Moreover, the adoptedmeasures were paper-and-pencil reports, which can lead

to biased responses from the subjects. Although adoptingMSDs self-report represents a limitation, evidence suggeststhat questionnaires are more sensitive indicators of MSDproblems than preexisting data sources [42]. However, in thisstudy, objectivemeasures would be suitable only for assessingthe MSDs, for instance, by medical evaluation. On the otherhand, attempting to collect objectivemeasures of the presenceof bullying in the workplace would not be feasible, due toproblems linked to the measures of negative activities, suchas bullying, which are subjective and difficult to identify[43]. Furthermore, it is not possible to state whether thetraining session could have partly impacted the workers’response rate but that session was considered necessary alsofrom the company management as workers had to answerto questions concerning their health and potential issuesconcerning bullying at work. A final limitation is that theadopted measure of workplace bullying insisted exclusivelyon repetitive and prolonged exposure to negative workplacebehaviours, thus ignoring other important defining elementsof the bullying definition such as the perceived imbalance ofpower between target and perpetrator(s). Althoughmeasuresinsisting on exposure to negative acts are often used inthe literature and they are also recommended when theaim of the study is to look at the relationship betweenbullying and other variables [44], such measures are far frombeing a perfect operationalization of bullying. Despite theselimitations, the current findings have implications for futureresearch directions and for practical implications. Indeed, forfuture studies on psychosocial risk factors and MSDs it maybe interesting to investigate not only job demands, specificallyworkload and lack of autonomy, which are often studiedas psychosocial risk factors associated with MSDs [45, 46],but also perceptions of work life quality and relationshipswithin the workplace. In this study, initial outcomes of suchrelationships have been reported, although further study isneeded not only pertaining toworkplace bullying, but relativeto the wider category of psychosocial contextual factors (i.e.,role clarity, work-family conflict). Until now, these have notbeen studied in relation toMSDs, yet they are known to affecthealth. Moreover, future research should also investigate thereciprocal relationship between bullying, job-related strain,and MSDs.

Regarding practical implications, our results underlinethat, in addition to more traditional prevention strategiesused to diminish biomechanical risk factors, establishing pre-vention strategies to reduce the presence of psychosocial riskfactors, in particular, workplace bullying, in the organizationof work should also be considered. Also, themediating role ofjob-related strain suggests that the good practices mentionedabove relative to ergonomic characteristics in the workplace


cannot be decisive in solving the issue. When addressingMSDs, both biomechanical and psychological sources shouldbe included. Our results, therefore, show that bullying can bethe initiator of the process which could lead to an increase ofMSDs, indicating the need to promote primary preventionintervention in the workplace to reduce bullying and, asa concequence, decrease perceived job-related strain andMSDs. Diverse studies have confirmed the role of organiza-tional factors affecting bullying, such as perceived cognitive,emotional and behavioral social support from colleagues[47], perceived organisational support [48] and psychologicalsafety climate [49]. Therefore our findings are in line with aprevention perspective, in which the contextual factors havethe most potential for broad impacts in reducing bullyingand its effects as they can be implemented in the workplace[50, 51]. Acting directly on the bullying prevention can help toreduce negative health outcomes, such as theMSDs presentedhere.



References

[1] H. Leymann, “The content and development of mobbing atwork,” European Journal of Work and Organizational Psychol-ogy, vol. 5, no. 2, pp. 165–184, 1996.

[2] S. Einarsen, H. Hoel, D. Zapf, and C. L. Cooper, “The conceptof bullying and harassment at work: the European tradition,”in Bullying and Harassment in the Workplace: Developments inTheory, Research, and Practice, pp. 3–39, 2011.

[3] H. Leymann and A. Gustafsson, “Mobbing at work and thedevelopment of post-traumatic stress disorders,”European Jour-nal ofWork andOrganizational Psychology, vol. 5, no. 2, pp. 251–275, 1996.

[4] E. G. E. Mikkelsen and S. Einarsen, “Basic assumptions andsymptoms of post-traumatic stress among victims of bullyingat work,” European Journal of Work and Organizational Psychol-ogy, vol. 11, no. 1, pp. 87–111, 2002.

[5] Eurofound, Fifth European Working Conditions Survey—Overview Report, Publications Office of the European Union,Luxembourg City, Luxembourg, 2012.

[6] M. B. Nielsen, S. B. Matthiesen, and S. Einarsen, “The impactof methodological moderators on prevalence rates of workplacebullying. Ameta-analysis,” Journal of Occupational and Organi-zational Psychology, vol. 83, no. 4, pp. 955–979, 2010.

[7] EuropeanAgency for Safety andHealth atWork,Annual Report,Office for Official Publications of the European Communities,2007.

[8] L. B. Finne, S. Knardahl, and B. Lau, “Workplace bullying andmental distress—a prospective study of Norwegian employees,”Scandinavian Journal of Work, Environment & Health, vol. 37,no. 4, pp. 276–287, 2011.

[9] M. Kivimaki, M. Elovainio, J. Vahtera, M. Virtanen, andS. A. Stansfeld, “Association between organizational inequityand incidence of psychiatric disorders in female employees,”Psychological Medicine, vol. 33, no. 2, pp. 319–326, 2003.

[10] M. B. Nielsen, J. Hetland, S. B. Matthiesen, and S. Einarsen,“Longitudinal relationships between workplace bullying andpsychological distress,” Scandinavian Journal of Work, Environ-ment & Health, vol. 38, no. 1, pp. 38–46, 2012.

[11] I. Niedhammer, J.-F. Chastang, S. David, L. Barouhiel, andG. Barrandon, “Psychosocial work environment and mentalhealth: job-strain and effort-reward imbalance models in acontext of major organizational changes,” International Journalof Occupational and Environmental Health, vol. 12, no. 2, pp. 111–119, 2006.

[12] R. Rugulies, “Studying the effect of the psychosocial work envi-ronment on risk of ill-health: towards a more comprehensiveassessment of working conditions,” Scandinavian Journal ofWork, Environment & Health, vol. 38, no. 3, pp. 187–191, 2012.

[13] M. B. Nielsen and S. Einarsen, “Outcomes of exposure toworkplace bullying: a meta-analytic review,”Work & Stress, vol.26, no. 4, pp. 309–332, 2012.

[14] J. Siegrist and T. Theorell, “Socioeconomic position and health.The role of work and employment,” in Social Inequalities inHealth: New Evidence and Policy Implications, J. Siegrist and M.Marmot, Eds., Oxford University Press, Oxford, UK, 2006.

[15] T. Theorell and H. M. Hasselhorn, “On cross-sectional ques-tionnaire studies of relationships between psychosocial con-ditions at work and health—are they reliable?” InternationalArchives of Occupational and Environmental Health, vol. 78, no.7, pp. 517–522, 2005.

[16] NIOSH, Musculoskeletal Disorders and Workplace Factors ACritical Review of Epidemiologic Evidence for Work-RelatedMusculoskeletal Disorders of the Neck, Upper Extremity, and LowBack, U.S. Departement of Health and Human Service, 1997.

[17] H. Brenner and W. Ahern, “Sickness absence and early retire-ment onhealth grounds in the construction industry in Ireland,”Occupational and Environmental Medicine, vol. 57, no. 9, pp.615–620, 2000.

[18] National Research Council and the Institute of Medicine,Musculoskeletal Disorders and the Workplace: Low Back andUpper Extremities. Panel on Musculoskeletal Disorders and theWorkplace, Commission on Behavioral and Social Sciences andEducation, National Academy Press, Washington, DC, USA,2001.

[19] S. D. Moon and S. L. Sauter, Beyond Biomechanics: PsychosocialAspects of Musculoskeletal Disorders in Office Work, Taylor &Francis, London, UK, 1996.

[20] P. M. Bongers, A. M. Kremer, and J. T. Laak, “Are psychosocialfactors, risk factors for symptoms and signs of the shoulder,elbow, or hand/wrist?: a review of the epidemiological litera-ture,” The American Journal of Industrial Medicine, vol. 41, no.5, pp. 315–342, 2002.

[21] G. A. M. Ariens, W. vanMechelen, P. M. Bongers, L. M. Bouter,and G. van der Wal, “Psychosocial risk factors for neck pain: asystematic review,”American Journal of IndustrialMedicine, vol.39, no. 2, pp. 180–193, 2001.

[22] P. M. Bongers, S. Ijmker, S. Van den Heuvel, and B. M. Blatter,“Epidemiology of work related neck and upper limb problems:psychosocial and personal risk factors (Part I) and effectiveinterventions from a bio behavioural perspective (Part II),”Journal of Occupational Rehabilitation, vol. 16, no. 3, pp. 272–295, 2006.

[23] B. Silverstein and B. Evanoff, “Musculoskeletal disorders,” inOccupational and Environmental Health, B. S. Levy, D. H.Wegman, S. L. Baron, and R. K. Sokas, Eds., Oxford UniversityPress, Ontario, Canada, 6th edition, 2011.


[24] C. A. Sprigg, C. B. Stride, T. D. Wall, D. J. Holman, and P. R.Smith, “Work characteristics, musculoskeletal disorders, andthe mediating role of psychological strain: a study of call centeremployees,” Journal of Applied Psychology, vol. 92, no. 5, pp.1456–1466, 2007.

[25] B. R. Da Costa and E. R. Vieira, “Risk factors for work-related musculoskeletal disorders: a systematic review of recentlongitudinal studies,” American Journal of Industrial Medicine,vol. 53, no. 3, pp. 285–323, 2010.

[26] V. Malinauskiene and I. Jonutyte, “Bullying in the workplaceand musculoskeletal disorders of the upper limb among seafar-ers,” Epidemiology, vol. 19, no. 6, p. S96, 2008.

[27] T. L. Vie, L. Glasø, and S. Einarsen, “Howdoes it feel?Workplacebullying, emotions and musculoskeletal complaints,” Scandina-vian Journal of Psychology, vol. 53, no. 2, pp. 165–173, 2012.

[28] I. Niedhammer, S. David, S. Degioanni, A. Drummond, andP. Philip, “Workplace bullying and psychotropic drug use: themediating role of physical and mental health status,” Annals ofOccupational Hygiene, vol. 55, no. 2, pp. 152–163, 2011.

[29] G. Notelaers and S. Einarsen, “The construction and validationof the short-negative acts questionnaire,” in Comunicacion enel 6th International Conference on Workplace Bullying, pp. 4–6,2008.

[30] C. Balducci, P. Spagnoli, V. Alfano, M. Barattucci, G. Note-laers, and F. Fraccaroli, “Valutare il rischio mobbing nelleorganizzazioni. Contributo alla validazione italiana dello ShortNegative Acts Questionnaire (S-NAQ),” Psicologia Sociale, vol.1, pp. 147–167, 2010.

[31] W. B. Schaufeli, M. P. Leiter, C. Maslach, and S. E. Jackson,Maslach Burnout Inventory—General Survey, 1996.

[32] L. Borgogni, D. Galati, L. Petitta, and Centro FormazioneSchweitzer, “Il questionario Checkup organizzativo,” in Man-uale dell'adattamento italiano, O.S. Organizzazioni Speciali,Florence, Italy, 2005.

[33] P. Cenni and F. Barbieri, “Karasek’s job content questionnaire:una versione italiana tradotta e adattata,” Report di Ricerca NonPubblicato, 1997.

[34] R. Karasek, Job ContentQuestionnaire andUser's Guide, Depart-ment of Industrial and System Engineering, University ofSouthern California, Los Angeles, Calif, USA, 1985.

[35] K. J. Preacher and A. F. Hayes, “SPSS and SAS proceduresfor estimating indirect effects in simple mediation models,”Behavior Research Methods, Instruments, & Computers, vol. 36,no. 4, pp. 717–731, 2004.

[36] J. C. Nunnally and I. H. Bernstein, Psychometric Theory, vol. 3,McGraw-Hill, New York, NY, USA, 1994.

[37] C. Balducci, M. Cecchin, and F. Fraccaroli, “The impact ofrole stressors on workplace bullying in both victims andperpetrators, controlling for personal vulnerability factors: alongitudinal analysis,”Work & Stress, vol. 26, no. 3, pp. 195–212,2012.

[38] M. R. Tuckey and A. M. Neall, “Workplace bullying erodes joband personal resources: between- and within-person perspec-tives,” Journal of Occupational Health Psychology, vol. 19, no. 4,pp. 413–424, 2014.

[39] R. Karasek, C. Brisson, N. Kawakami, I. Houtman, P. Bongers,and B. Amick, “The Job content questionnaire (JCQ): aninstrument for internationally comparative assessments of psy-chosocial job characteristics.,” Journal of Occupational HealthPsychology, vol. 3, no. 4, pp. 322–355, 1998.

[40] European Agency for Safety and Health at Work,Work-RelatedLow Back Disorders, Office for Official Publications of theEuropeanCommunities, Luxembourg City, Luxembourg, 2000.

[41] K. R. Parkes, S. Carnell, and E. Farmer, Musculoskeletal Disor-ders, Mental Health and the Work Environment, HSE, Sudbury,Canada, 2005.

[42] B. A. Silverstein, D. S. Stetson, W. M. Keyserling, and L. J.Fine, “Work-related musculoskeletal disorders: comparison ofdata sources for surveillance,” American Journal of IndustrialMedicine, vol. 31, no. 5, pp. 600–608, 1997.

[43] I. Coyne, P. Chong, E. Seigne, and P. Randall, “Self and peernominations of bullying: an analysis of incident rates, individualdifferences, and perceptions of the working environment,”European Journal of Work and Organizational Psychology, vol.12, no. 3, pp. 209–228, 2003.

[44] M. B. Nielsen, G. Notelaers, and S. Einarsen, “Measuringexposure to workplace bullying,” in Bullying and Harassment inthe Workplace: Developments in Theory, Research, and Practice,pp. 149–174, 2011.

[45] C. A. Sprigg and P. R. Jackson, “Call centers as lean serviceenvironments: job-related strain and themediating role of workdesign,” Journal of Occupational Health Psychology, vol. 11, no. 2,pp. 197–212, 2006.

[46] G. A. M. Ariens, P. M. Bongers, W. E. Hoogendoorn, I. L.D. Houtman, G. Van Der Wal, and W. Van Mechelen, “Highquantitative job demands and low coworker support as riskfactors for neck pain: results of a prospective cohort study,”Spine, vol. 26, no. 17, pp. 1896–1901, 2001.

[47] M. R. Tuckey, M. F. Dollard, P. J. Hosking, and A. H. Winefield,“Workplace bullying: the role of psychosocial work environ-ment factors,” International Journal of Stress Management, vol.16, no. 3, pp. 215–232, 2009.

[48] N. Djurkovic, D.McCormack, and G. Casimir, “Workplace bul-lying and intention to leave: the moderating effect of perceivedorganisational support,”Human Resource Management Journal,vol. 18, no. 4, pp. 405–422, 2008.

[49] R. Law, M. F. Dollard, M. R. Tuckey, and C. Dormann,“Psychosocial safety climate as a lead indicator of workplacebullying and harassment, job resources, psychological healthand employee engagement,” Accident Analysis and Prevention,vol. 43, no. 5, pp. 1782–1793, 2011.

[50] H. Cooper-Thomas, D. Gardner, M. O’Driscoll, B. Catley, T.Bentley, andL. Trenberth, “Neutralizingworkplace bullying: thebuffering effects of contextual factors,” Journal of ManagerialPsychology, vol. 28, no. 4, pp. 384–407, 2013.

[51] D. Salin and H. Hoel, “Organisational causes of workplacebullying,” in Bullying and Harassment in the Workplace: Devel-opments inTheory, Research, and Practice, S. Einarsen, H. Hoel,D. Zapf, and C. Cooper, Eds., pp. 227–243, Taylor & Francis,London, UK, 2nd edition, 2011.

Research ArticleEstimating the Impact of Workplace Bullying:Humanistic and Economic Burden among Workers withChronic Medical Conditions

A. Fattori,1 L. Neri,1 E. Aguglia,2 A. Bellomo,3 A. Bisogno,4 D. Camerino,1

B. Carpiniello,5 A. Cassin,6 G. Costa,1,7 P. De Fazio,8 G. Di Sciascio,9 G. Favaretto,10

C. Fraticelli,11 R. Giannelli,12 S. Leone,13 T. Maniscalco,14 C. Marchesi,15,16

M. Mauri,17 C. Mencacci,18 G. Polselli,19 R. Quartesan,20 F. Risso,21 A. Sciaretta,22

M. Vaggi,23 S. Vender,24 and U. Viora25

1 Department of Clinical Sciences and Community Health, University of Milan, 20124 Milan, Italy2 Department of Psychiatry, University of Catania, 95131 Catania, Italy3 Department of Clinical and Experimental Medicine, University of Foggia, 71121 Foggia, Italy4 Department of Mental Health, UO Cava de’Tirreni, 84013 Salerno, Italy5 Department of Public Health, Clinical and Molecular Medicine, University of Cagliari, 09124 Cagliari, Italy6 Department of Mental Health, ASS n.6, 33170 Pordenone, Italy7 IRCCS Maggiore Policlinico Hospital, Ca’Granda Foundation, 20124 Milan, Italy8 Department of Health Sciences, School of Specialization in Psychiatry, Magna Græcia University of Catanzaro,88100 Catanzaro, Italy

9 Psychiatric Neuroscience Group, Department of Neurological and Psychiatric Sciences, University of Bari,70125 Bari, Italy

10Department of Mental Health, ULSS 7, Conegliano, 31015 Treviso, Italy11 Unit of Psychiatry, Sant’Anna Hospital, 22100 Como, Italy12National Association of Rheumatic Patients (ANMAR), 00153 Rome, Italy13National Association for Inflammatory Bowel Disease (AMICI), 20131 Milan, Italy14Department of Mental Health, AULSS Legnago Hospital, Legnago, 37045 Verona, Italy15Department of Neuroscience, Psychiatry Unit, University of Parma, 43121 Parma, Italy16Mental Health Department, AUSL, 43126 Parma, Italy17Department of Experimental and Clinic Medicine, Section of Psychiatry, University of Pisa, 56100 Pisa, Italy18Depression Unit, Neuroscience Department, Fatebenefratelli Hospital, 20137 Milan, Italy19Department of Neurology and Psychiatry, Policlinico Umberto I, Sapienza University of Rome, 00185 Rome, Italy20Division of Psychiatry, Clinical Psychology and Psychiatric Rehabilitation, Department of Medicine, New Faculty of Medicine,University of Perugia, Sant’Andrea delle Fratte, 06156 Perugia, Italy

21Unit of Psychiatry, A.S.L. CN1, 12100 Cuneo, Italy22Servizio Psichiatrico di Diagnosi e Cura (SPDC), Tivoli Hospital, 00019 Rome, Italy23Department of Mental Health, ASL3, 16125 Genoa, Italy24Department of Clinical and Experimental Medicine, Psychiatry, Faculty of Medicine and Surgery, University of Insubria,21100 Varese, Italy

25National Association ANAP Onlus, Rivoli, 10098 Turin, Italy

Correspondence should be addressed to L. Neri; [email protected]

Received 16 January 2015; Revised 12 March 2015; Accepted 27 April 2015


Copyright © 2015 A. Fattori et al.This is an open access article distributed under theCreative CommonsAttribution License, whichpermits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.


http://dx.doi.org/10.1155/2015/708908


Background. Although the prevalence of work-limiting diseases is increasing, the interplay between occupational exposures andchronic medical conditions remains largely uncharacterized. Research has shown the detrimental effects of workplace bullyingbut very little is known about the humanistic and productivity cost in victims with chronic illnesses. We sought to assess workproductivity losses and health disutility associated with bullying among subjects with chronic medical conditions. Methods.Participants (𝑁 = 1717) with chronic diseases answered a self-administered survey including sociodemographic and clinical data,workplace bullying experience, the SF-12 questionnaire, and theWork Productivity Activity Impairment questionnaire. Results.Theprevalence of significant impairment was higher among victims of workplace bullying as compared to nonvictims (SF-12 PCS: 55.5%versus 67.9%, 𝑝 < 0.01; SF-12 MCS: 59.4% versus 74.3%, 𝑝 < 0.01). The adjusted marginal overall productivity cost of workplacebullying ranged from 13.9% to 17.4%, corresponding to Italian Purchase Power Parity (PPP) 2010 US$ 4182–5236 yearly. Associationestimates were independent and not moderated by concurrent medical conditions. Conclusions. Our findings demonstrate that theburden on workers’ quality of life and productivity associated with workplace bullying is substantial. This study provides key datato inform policy-making and prioritize occupational health interventions.

1. Introduction

All developed countries are facing a sustained shift in thedemographic composition of their population and are thusdevoting major effort to increasing the work participationrate of aging and disabled people [1]. Together with long-term health problems and chronic diseases, prevalence ofwork-limiting disabilities increases with age [2]: it has beenestimated that 72% of all-causes Disability-Adjusted LifeYears occur in subjects under 60 years old and more thanthree-quarters of old workers have at least one chronic healthcondition that requires management [3, 4]. In addition, themajority of workers with chronic illnesses continue to workand have to deal with several workplace risk factors [5, 6].However, the interplay between occupational exposures andchronic medical conditions remains largely uncharacterized,thus limiting the potential for effective preventive and thera-peutic actions.

Workplace bullying is a common and severe occupationalstressor and imbalance of power, harm, and systematic rep-etition over time represent its key elements [7]. The adverseeffects of workplace bullying on victims’ psychological healthspan from mild anxiety and depression to severe posttrau-matic stress symptoms [8–14]. Similarly, workplace bullyingalso has a detrimental impact on organizational outcomes,such as job satisfaction, organizational commitment, andintention to leave [15, 16].

Despite evidence showing that workplace bullying maybe associated with a significant financial burden for victimsand organizations, cost estimates are difficult to comparedue to different currencies, methodologies, time frames,and the selection of different cost drivers (e.g., health carecost, productivity and performance loss, sick leave, andreplacement costs) [17].

Research has shown that having a disability is a risk factorfor being bullied [18–21] and there is some evidence showinglongitudinal associations between mental health problemsand subsequent exposure to bullying at work [22–25]. Targetsof workplace bullying with preexisting chronic diseases couldexperience worse consequences and perceive themselves asbeing bullied more frequently compared to colleagues withno other medical conditions [12, 26]. Furthermore, exposureto psychosocial stressors at work may play an important rolein retirement behavior and labor supply decisions amongworkers with chronic diseases [27, 28].

Although psychosocial factors and chronic conditions areboth emerging issues in occupational medicine, very littleis known about the humanistic and productivity cost ofbullying at work in workers with chronic illness. Empiricalresearch would help health authorities and employers pri-oritize the allocation of limited resources for occupationalhealth interventions [29]. As a consequence, the assessmentof health-related quality of life and cost-effectiveness analysesare gaining importance in occupational medicine becausedecision-makers need comparable and accurate informationin order to achieve the greatest health improvement fortheir workforce. In the present study, we sought to evaluatework productivity losses and health disutility associated withworkplace bullying among patients with different chronicmedical conditions.


2.1. Participants and Procedures. The present study is ajoint analysis of the Liberamente and MOSAICO researchdatasets. Both studies aimed at evaluating the quality of life,treatment satisfaction, social participation, and health careutilization of patients with commonmedical conditions suchas major depression disorder (Liberamente study), inflam-matory bowel disease, psoriasis and autoimmune arthritis(MOSAICO study).

The Liberamente study was carried out between June andJuly 2013 in 18 outpatient referral centers for diagnosis andtreatment of psychiatric disorders across all Italian regions.Patients referred to the centers for psychiatric conditionswere screened for eligibility by a psychiatrist during a regularfollow-up visit at the clinic. We included adult patients with aclinical diagnosis of depression (i.e., recurrent depressive dis-order, major depressive episode, adjustment disorder, mixedaffective disorder, dysthymia, and other persistent depressivedisorders) with the exclusion of bipolar disorders. Seven hun-dred patients agreed to participate in the research completingan anonymous self-administered paper-and-pencil question-naire. Concurrently, the same psychiatrist recorded relevantclinical characteristics in a standardized data collection form.To preserve anonymity of data collection while matchingclinical and patient-reported information, the psychiatristhanded the data collection form to the patient at the end of thevisit.Thepatients sealed both the data collection form and the


self-administered questionnaire in an anonymous envelopeto return to the research team.

The MOSAICO study was carried out between April andOctober 2014. We invited the members of Patients’ Asso-ciations for people with Crohn’s Disease, ulcerative colitis,psoriasis, ankylosing spondylitis, rheumatoid arthritis, andpsoriatic arthritis to take part in the study. The surveyadopted a dual methodology. A web survey was posted onthe associations’ web sites. Respondents were 143 registeredpatients with autoimmune arthritis, 397 patients with inflam-matory bowel disease, and 236 patients with psoriasis. Allparticipants accessed the survey from their personal account.Additionally, 152, 52, and 37 patientswith autoimmune arthri-tis, inflammatory bowel disease, and psoriasis, respectively,elected to complete a self-administered paper-and-pencilversion of the survey.

2.2. Measures. Surveys included sociodemographic andoccupational information, the Work Productivity and Activ-ity Impairment scale, and the SF-12v1 (RAND).

2.2.1. Workplace Bullying. Research has essentially used twomethodologies to assess workplace bullying: (i) investigatingsubjects’ perceptions of victimization (self-labelling method)and/or (ii) their exposure to specific bullying behaviors(behavioral experience method) [30]. In this study, work-place bullying was measured using the self-labelling methodby providing respondents with a theoretical definition ofworkplace bullying (“bullying takes place when one or morepersons systematically and over time feel that they havebeen subjected to negative treatment on the part of one ormore persons, in a situation in which the person(s) exposedto the treatment have difficulty in defending themselvesagainst them. It is not bullying when two equally strongopponents are in conflict with each other” [31, pages 190-191]).Participants responded to a single-item question (“accordingto this definition have you been subjected to bullying at theworkplace during the last six months?”) using a five-pointscale from 1 (never) to 5 (yes, many times a week). Subjectsreporting a frequency of bullying of 3 or above on the five-point scale were considered victims of workplace bullyingfor the purpose of this analysis. In addition, the length oftime for which workplace bullying had been experienced wasassessed.

2.2.2. Outcomes

Health-Related Quality of Life. The SF-12 questionnaire(RAND, [32]) is a 12-item generic health profile measureassessing patients’ perception of their own mental andphysical health. Ratings use a 0–100-point scale. A 𝑡-scorecalculated on the normative values of the Italian generalpopulation is obtained from raw scores. Patients reportingscores lower than 42.0 and 43.85 on the SF-12 mental andphysical composites, respectively, were classified as signifi-cantly impaired [33, 34].Health Utility. The SD-6D utility index represents thevalue assigned to a specific health status characterized by

the impairments, functional states, perceptions, and socialopportunities that are influenced by disease, injury, treat-ment, or policy. The scoring algorithm of the Short Form 6Dimension (SF-6D) is a two-step process: in the first step,responses to SF-12 questions are used to define a responsevector representing the patient’s health state (classificationsystem); then, in the second step, the vector is converted intoa utility value using a utility function obtained from a sampleof the general population. The SF-6D classification systemincludes six multilevel dimensions (physical functioning,role of participation, social functioning, bodily pain, mentalhealth, and vitality) and describes 18,000 health states [35].The SF-6D utility index was calculated according to theutility function observed by Brazier et al. [35] using standardgamble experiments carried out in a sample of the generalpopulation. Scale ratings range from 0 (death) to 1 (perfecthealth).Productivity Loss. Economic evaluations conducted from anemployer’s perspective express the benefit of occupationalhealth interventions in terms of health-related productivity,which is translated into a monetary value and may also bereferred to as an indirect cost [29]. The Work ProductivityActivity Impairment questionnaire [36] consists of 4 items:(Q1) hours lost due to health problems; (Q2) hours lost dueto any other reason; (Q3) hours actually worked; and (Q4)the degree of which health problems affected productivitywhile at work. Responses ranged from 0 (“My health problemhad no effect on my work”) to 10 (“My health problemcompletely preventedme fromworking”). Estimationmetricswere calculated as percentage productivity losses, with highervalues indicating a greater proportion of time lost at work(less productivity). The following equations were calculated.

Equation for sick leave is as follows:

[Q1(Q1 +Q3)

] ∗ 100. (1)

Equation for work impairment while at work or presen-teeism is as follows:

Q4 ∗ 10. (2)

Equation for overall work productivity loss is as follows:

{SickLeave + [(1 − SickLeave) ∗ Presenteeism]}

∗ 100.

(3)

Percentage productivity losses were converted to the cor-responding share of the Italian Purchase Power Parity (PPP)per capita Gross Domestic Product (2010, US$ 31,090) [37]which allows cross-national comparisons. PPP represents thereal exchange rate (nominal exchange rate adjusted for theprice index), that is, how much money would be neededto purchase the same goods and services in two differentcountries.

2.2.3. Demographic Information. Surveys included a sectionon sociodemographic characteristics. We recorded patients’age, gender, education level, marital status, employment


Table 1: Characteristics of study sample across different chronic diseases. 𝑝 values represent significant levels of 𝜒2 for categorical variables,one-way ANOVA for continuous variables.

Total(𝑁 = 1717)

AA(𝑁 = 295)

IBD(𝑁 = 449)

MDD(𝑁 = 700)

PSO(𝑁 = 273) 𝑝

𝑛 (%) 𝑛 (%) 𝑛 (%) 𝑛 (%) 𝑛 (%)Women (𝑁 = 1586)∗ 978 (61.7) 172 (72.9) 243 (57.5) 436 (65.2) 127 (49.2) <0.0001Family (𝑁 = 1592)∗ 969 (60.9) 170 (72.0) 240 (56.7) 376 (55.7) 183 (71.0) <0.0001University education (𝑁 = 1589)∗ 315 (19.8) 45 (19.1) 78 (18.4) 116 (17.3) 76 (29.5) 0.0004Employment (𝑁 = 1603)∗ <0.0001

Employed 897 (56.0) 129 (54.7) 268 (63.4) 334 (48.7) 166 (64.3)Inactive 293 (18.3) 20 (8.5) 62 (14.7) 177 (25.8) 34 (13.2)Retired 183 (11.4) 67 (28.4) 43 (10.2) 60 (8.75) 13 (5.04)Unemployed 230 (14.3) 20 (8.47) 50 (11.8) 115 (16.8) 45 (17.4)

Contract (𝑁 = 1591)∗ <0.0001Temporary 128 (8.05) 12 (5.17) 35 (8.29) 62 (9.13) 19 (7.36)Permanent 606 (38.1) 90 (38.8) 178 (42.2) 219 (32.3) 119 (46.1)Self-employed/employer 151 (9.49) 23 (9.91) 54 (12.8) 46 (6.77) 28 (10.8)

Job demand (WAI) (𝑁 = 882)∗ <0.0001Physical demand 80 (9.07) 4 (3.17) 17 (6.37) 47 (14.6) 12 (7.23)Mental demand 416 (47.2) 56 (44.4) 121 (45.3) 139 (43.0) 100 (60.2)Mixed demand 386 (43.8) 66 (52.4) 129 (48.3) 137 (42.4) 54 (32.5)

Mean (sd) Mean (sd) Mean (sd) Mean (sd) Mean (sd) pAge (𝑁 = 1603)∗ 46.8 (13.1) 48.8 (10.4) 42.0 (12.2) 46.1 (10.9) 44.1 (8.79) <0.0001Workforce (𝑁 = 1717)∗ 1603 (93.4) 236 (80.0) 423 (94.2) 686 (98.0) 258 (94.5) <0.0001Time since diagnosis (years) (𝑁 = 1593)∗ 10.4 (10.2) 13.3 (10.1) 12.4 (9.45) 5.89 (7.38) 12.5 (11.5) <0.0001Hospitalization (days) (𝑁 = 1593)∗ 2.37 (8.38) 1.90 (7.00) 3.50 (10.9) 2.55 (8.51) 0.72 (2.79) 0.0004∗Number of valued cases for each variable.

status, employment contract, and preeminent job demand(either physical, mental, or mixed) with items from theWorkAbility Index (WAI, [38]). Workforce status was definedbased on patients’ age (between 18 and 65 years). We classi-fied employment, inactivity, retirement, and unemploymentstatus using the International Labour Office definition [39].Common medical information in both datasets included thenumber of days of hospitalization in the past 12 months, timesince chronic disease onset (years), and chronic conditiontype (major depression disorder (MDD), autoimmune arthri-tis (AA), psoriasis (PSO), and inflammatory bowel disease(IBD)). We used number of days of hospitalization ratherthan overall health care utilization rates (i.e., outpatientsvisits, mental health services) as a proxy of chronic diseaseseverity to minimize the information bias due to the inabilityto discern between bullying-related medical encounters andthose caused by the cooccurring chronic medical condition.

2.3. Analysis. We computedmeans and standard deviation orabsolute and relative frequency of continuous and categoricalvariables, respectively. We evaluated differences in sociode-mographic and clinical characteristics across diagnosis statuswith 𝜒2 or one-way ANOVA as appropriate. Unadjustedand adjusted quality-of-life penalty, health disutility, andproductivity losses associated with workplace bullying were

estimated with generalized linear models. To account for theskewed distribution of outcomes, we fitted OLS regressionswith log link function in the analysis of SF-12 and SF-6D scores. Additionally, we fitted gamma regressions forthe analysis of lost productivity time (WPAI metrics). Allmodels were adjusted for patients’ age, gender, education,marital status, job demand, contract, hospitalization days,diagnosis, time since disease onset, and time since the onsetof workplace bullying.We also tested the interaction betweenchronic disease and self-reported bullying experience in allmodels. 𝑝 < 0.05 was considered statistically significant.Analyses were conducted with SAS 9.2.

3. Results

3.1. Sample Characteristics. Demographic and clinical char-acteristics of the sample are summarized in Table 1. Althoughthe majority of subjects were of working age (46.8 ± 13.1),only 56% of the sample were actually employed. Participantswith a paid job were more likely to be men (63.7% versus51.3%, 𝑝 < 0.01), were slightly younger (44.0 ± 9.8 versus46.3 ± 12.5, 𝑝 < 0.01), reported less hospitalization days(1.8 ± 6.5 versus 3.2 ± 10.4, 𝑝 < 0.01), and were morelikely to have tertiary qualifications (24.7% versus 13.6%,𝑝 < 0.01). Among subgroups, significant differences were


Table 2: Sample characteristics across workplace bullying status. 𝑝 values represent significant levels of 𝜒2 for categorical variables, one-wayANOVA for continuous variables.

No workplace bullying𝑁 = 632 (83.7%)𝑛 (%)

Workplace bullying𝑁 = 123 (16.3%)𝑛 (%)

𝑝

Women 367 (58.6) 68 (56.7) nsMarried or de facto 392 (62.2) 67 (54.9) nsUniversity education 154 (24.5) 22 (18.0) nsJob security 520 (84.1) 98 (15.9) nsJob demand ns

Physical dem. 59 (86.8) 9 (13.2)Mental dem. 303 (85.1) 53 (14.9)Mixed dem. 261 (81.6) 59 (18.4)

Mean (sd) Mean (sd) pAge 44.6 (10.9) 47.2 (12.3) 0.02Time since diagnosis (years) 9.89 (9.71) 10.0 (9.97) nsHospitalization (days) 1.61 (5.83) 1.88 (4.46) ns

seen in employment status: subjects with MDD and PSOreported, respectively, the highest percentage of inactivity(25.8%) and unemployment (17.4%), while 28.4% of patientswith AA were retired. In most subgroups, there was greaterrepresentation of women and the overall prevalence of jobswith considerable physical demands was quite low. Apartfrom MDD, with most subjects reporting a more recentdiagnosis (5.89 ± 7.38), most other chronic diseases had beenaffecting participants for a long time (12.4 to 13.3 years).

3.2. Prevalence of Workplace Bullying. Table 2 shows theprevalence of workplace bullying in the whole sample. Onehundred and twenty-three subjects (16.3%) labelled them-selves as victims of bullying at work. Bullied subjects wereslightly older (44.6 ± 10.9 versus 47.2 ± 12.3; 𝑝 = 0.02).No statistically significant differences were found in bullyingprevalence across the different chronic diseases (AA 16.2%,IBD 15.4%, MDD 17.6%, and PSO 15.1%, 𝑝 = 0.89).

Eighty-one percent of bullied subjects had a preexistentmedical condition before bullying onset. However, in thesubgroup of patientswithMDD, 30% reported thatworkplacebullying had occurred before the onset of depression.

3.3. Workplace Bullying and Productivity Losses. The meanaverage weekly full-time equivalent sick hours were 6.58± 11.92, and corresponding average sick-leave rate was16.4% ± 29.8. Work impairment was 41.9% ± 31.6, whereasthe overall productivity loss (absenteeism + presenteeism)was 46.5% ± 33.2. Unadjusted productivity losses due tosick leave and presenteeism were both associated with work-place bullying (Figure 1(a), 𝑝 < 0.001). These associationswere both robust to adjustment for possible confounders(Figure 1(b), 𝑝 < 0.001) and were not moderated bydisease status (𝑝 for interactions with diagnosis >0.05). Therelative risk of sick leave associated with workplace bullyingwas 1.86 (95% CI: 1.30–2.82). This estimate was robust toadjustment for age, gender, education, chronic disease status,

and contract type (temporary/long-term contract). Amonghypothesized confounders, only days of hospitalizations wereassociated with productivity losses (Table 3). The adjustedmarginal overall productivity cost of workplace bullyingranged from 13.9% (IBD) to 17.4% (PSO), corresponding toPPP 2010US$ 4182–5236 yearly.

3.4. Workplace Bullying and Health-Related Quality of Life.The average scores of quality of life were 41.9 ± 10.6, 39.8 ±11.3, and 0.664 ± 0.102 for the SF-12 PCS, SF-12 MCS, and SF-6D indexes, respectively. Among employed patients, 56.7%and 57.0% reported significant impairment as defined by theSF-12 PCS and SF-12 MCS scales, respectively. Workers whoself-reported bullying at work were more likely classified assignificantly impaired on both scales compared to nonvictims(SF-12 PCS: 55.5% versus 67.9%, 𝑝 < 0.01; SF-12 MCS:59.4% versus 74.3%, 𝑝 < 0.01). Unadjusted health-relatedquality-of-life scores were associated with workplace bullying(Figure 2(a), 𝑝 < 0.001). These associations were bothrobust to adjustment for possible confounders (Figure 2(b),𝑝 < 0.001) and were not moderated by disease status (𝑝for interactions with diagnosis >0.05).The adjusted marginaldisutility associated with workplace bullying ranged from0.048 (AA) to 0.052 (PSO). Figure 3 illustrates unadjusted SF-6D scores of workers who self-reported workplace bullyingas compared to those who had not experienced workplacebullying.

Additionally, health-related quality-of-life scores wereassociated with days of hospitalization, gender, marital status,education, job security, and diagnosis (Table 3).

4. Discussion

4.1. Prevalence of Workplace Bullying. In this large multi-center cross-sectional study among workers with commonchronic conditions [40–43], the prevalence of workplacebullying was 16% and most workplace bullying started after


0102030405060708090

100

Prod

uctiv

ity lo

ss (%

)

No workplace bullyingWorkplace bullying

38.0

10.5

42.733.5

12.8

39.9 40.5

16.1

44.636.2

6.9

40.2

65.6

30.8

68.5

44.0

10.3

53.0 53.6

29.3

58.268.3

28.4

79.8

Impairment Sick leave Totalproductivity

loss


loss


loss


lossAutoimmune arthritis Inflammatory bowel disease Major depression disorder Psoriasis

(a)

36.5

11.0

43.8

28.7

9.0

36.5 37.3

15.0

42.0 38.9

10.1

46.051.9

22.2

60.5

40.8

18.2

50.4 53.0

30.2

58.0 55.3

20.3

63.4

0102030405060708090

100

Prod

uctiv

ity lo

ss (%

)



loss


loss


loss


lossAutoimmune arthritis Inflammatory bowel disease Major depression disorder Psoriasis

(b)

Figure 1: Work productivity losses across workplace bullying status among workers with different chronic conditions. (a) Unadjusted scores.(b) Adjusted scores. Models included age, gender, time since diagnosis, time since workplace bullying onset, education, marital status, jobdemand, contract, diagnosis, and hospitalization days.

the onset of chronic disease. There is wide variation inprevalence estimates of workplace bullying across studies.Italian rates in the general working population range from4.8% in a public service organization to 31.4% among airportemployees [44]. Other prevalence studies have found rates of3.5% in Sweden up to 27% in North America [45, 46]. Thesediscrepancies are partially explained by different methods ofmeasurement and criteria used to define workplace bullying[30, 47, 48].

4.2. Workplace Bullying and Productivity Loss. We observeda significant association between workplace bullying and allcomponents of productivity. Workers who were not self-labelled as victims of workplace bullying showed WPAIscores similar to previous findings among patients with the

samemedical conditions [49–52]. However, participants whoself-reported workplace bullying showedmuch higherWPAIscores. Our estimates suggest that the potential economicimpact of preventive or therapeutic interventions addressingworkplace bullying on yearly overall productivity loss mightrange from about PPP 2010US$ 4200 to 5200 for each caseprevented. Although cost-of-illness studies provide valuableinformation on the overall burden of disease, they generallylead to unrealistic expectations about savings from therapy ascurrent treatments may reduce symptoms but are unable toeradicate the disease. Conversely, several effective interven-tions can be implemented at different levels to prevent andmanage workplace bullying (e.g., antibullying policy, codeof conduct, psychosocial risk analysis, and training) [53].Coupled with the huge impact on overall productivity loss,


Table 3: Significant association estimates (𝛼 < 5%) between workers sociodemographic and clinical characteristics and study outcomes.Estimates represent the change score in outcome for each unit change in the independent variables. Models included age, gender, time sincediagnosis, time since workplace bullying onset, education, marital status, job demand, contract, diagnosis, and hospitalization days.

Impairment Sick leave Total productivity loss SF-12 PCS SF-12 MCS SF-6D indexHospitalization (days) 0.027∗ 0.055∗∗∗ 0.025∗ −0.012∗∗∗ −0.007∗∗ −0.005∗∗∗

Men ns ns ns 0.069∗∗∗ ns 0.040∗∗∗

Living alone ns ns ns ns −0.060∗∗ nsHigh school or lower ns ns ns ns ns −0.028∗

Temporary work ns ns ns ns −0.057∗ nsDiagnosis

AA ns ns ns −0.208∗∗∗ ns −0.073∗∗∗

IBD ns ns ns −0.088∗∗∗ ns −0.052∗∗

MDD ns ns ns −0.066∗∗ −0.139∗∗∗ −0.043∗∗

PSO — — — — — —∗𝑝 < 0.05; ∗∗𝑝 < 0.01; ∗∗∗𝑝 < 0.001.

workplace bullying should be considered an overriding issuefor public health authorities and employers alike. Althoughour study cannot demonstrate causality of association, ourfindings help compare competing hypothetical scenarios toprioritize research investments. Our results demonstrate astrong association between sick-leave rates and exposure toworkplace bullying. Nonetheless, previous studies have foundrelatively weak relationships between workplace bullyingand absenteeism [15, 16]. This might be explained by theobservation that victims of bullying may enhance their effortand commitment when their work performance and self-esteem are impaired [54]. Such compensative mechanismmight not offset detrimental effects of bullying among victimswith concurrent chronic health conditions due to the greaterseverity of their psychological and psychosomatic complaints[12].

Additionally, contrary to the short reference time adoptedin this study (e.g., self-reported hours lost in the past week),most of the previous studies based their estimates on sick-leave events registered in administrative databases or adoptedcoarse self-reported measures (i.e., ever taken any sick leavedue to workplace bullying) which may lead to informationbias [26, 55, 56]. For example, as data repositories serve work-compensation procedures, their capture rate may be limitedto eventswhose duration is eligible for compensation. For thisreason, estimates from previous studies may underestimatethe real productivity burden of workplace bullying. Consis-tent with labor supply models, there is empirical evidencesuggesting that long-term sick leave is not an expression ofwithdrawal behaviors such as lateness, shorter sick leave, orreduced performance at work; on the contrary, longer spellsaremore frequently associatedwith serious illness rather thanreduced commitment and motivation [57, 58]. Consistentwith previous studies demonstrating the relationship betweenincivility at work and withdrawal behavior [59], we showedthat workplace bullying is associatedwith reduced attendance(i.e., either lateness or sick days) beyond the effect of concur-rent disabling medical conditions.

4.3. Workplace Bullying and Health-Related Quality of Life. Afurther important finding of our study was that workplace

bullying was associated with worse health-related quality-of-life scores above and beyond the detrimental effect of otherconcurrentmedical conditions.There is sparse evidence fromprevious studies that exposure to occupational psychosocialstrain is associated with reduced health-related quality of life[60–63]. To our knowledge, this is the first study assessing theassociation between workplace bullying and health-relatedquality of life. The workplace bullying penalty observed inour study was clinically significant for the SF-12 PCS, SF-12MCS, and SF-6D index according to the proposed thresholdsfor the minimal clinically important difference for HRQOL[64]. Additionally, the overall effect size observed in ourstudy was similar to the SF-12 physical composite (Cohen’s𝑑 = 0.42) compared to the SF-12 mental composite (Cohen’s𝑑 = 0.47). Of note, victims of workplace bullying were morelikely classified as significantly impaired on both SF-12 scales.The cut-off chosen represents the lowest octile of the scoredistribution in working populations and indicates a severelycompromised function.

The overall effect size observed for the SF-6D index wasmoderate (𝑑 = 0.57): the adjusted disutility associated withworkplace bullying corresponded to 18-19 days of healthylife lost for each year spent with the condition. The SF-6Dscores reported by patients who were not self-labelled asvictims of workplace bullying were comparable to figuresreported in previous studies among subjects with the samemedical condition (Figure 3) [65–69]. Exposure to workplacebullying is associated with posttraumatic stress reactions,anxiety, depression, and insomnia as well as chronic fatigue,psychosomatic symptoms,musculoskeletal and gastrointesti-nal disorders, headaches, and hypertension [8, 9, 12–14].

4.4. Strengths and Limitation. This study has severalstrengths. First, we evaluated the burden of workplacebullying on important outcomes among underresearchedgroups with different diseases. Second, we complied withrecommendations for reporting economic evaluations inoccupationalmedicine [29]. For example, in order to improvecomparability and interpretability of our findings and tominimize the likelihood of underestimation, we adopteda widely used questionnaire, and we identified the source


65.4

36.943.6

66.2

42.0 40.9

67.6

43.937.5

71.6

47.643.8

58.3

32.9 36.1

62.5

37.9 36.3

64.2

41.433.2

59.5

40.136.1

0

10

20

30

40

50

60

70

80

90

100

Qua

lity

of li

fe


SF-6D SF-12 PCS SF-12 MCS SF-6D SF-12 PCS SF-12 MCS SF-6D SF-12 PCS SF-12 MCS SF-6D SF-12 PCS SF-12 MCSAutoimmune arthritis Inflammatory bowel disease Major depression disorder Psoriasis

(a)

66.1

38.642.7

67.5

43.6 40.6

68.2

44.536.7

71.1

47.642.1

61.3

34.638.1

62.6

39.1 36.2

63.2

39.932.7

65.9

42.737.6


SF-6D SF-12 PCS SF-12 MCS SF-6D SF-12 PCS SF-12 MCS SF-6D SF-12 PCS SF-12 MCS SF-6D SF-12 PCS SF-12 MCSAutoimmune arthritis Inflammatory bowel disease Major depression disorder Psoriasis

0

10

20

30

40

50

60

70

80

90

100

Qua

lity

of li

fe

(b)

Figure 2: Health-related quality of life across workplace bullying status among workers with different chronic conditions. (a) Unadjustedscores. (b) Adjusted scores. Models included age, gender, time since diagnosis, time since workplace bullying onset, education, marital status,job demand, contract, diagnosis, and hospitalization days.

of price weights used and reported percent productivityloss for all components of indirect costs from the employerperspective [70]. By converting such findings into a financialmetric, we sought to help organizational and public healthstakeholders to better translate the impact of workplacebullying for people with chronic medical conditions. Third,our large sample size permitted adjustment for potentiallyimportant confounders thus reducing the likelihood of bias.Finally, community-based data on work productivity from aclinical population may present lesser degree of desirabilitybias compared to surveys conducted in occupational settings.

However, we must acknowledge some limitations. Werelied on a self-labelling measure of workplace bullying, the

most commonly adopted in epidemiological studies [48],whichmight have introduced information bias.Howdifferentestimation methods and measurements affect findings isstill underinvestigated [30]. We primed participants with awidely accepted theoretic definition of workplace bullying toimprove the accuracy of their subjective evaluation of vic-timization and power imbalance given the complexity of thephenomenon and potential for misinterpretation. Typicallyprevalence estimates yielded with the self-labelling approachare lower than those based on behavioral experiencemethods[47], so we used a broad cut-off for frequency of bullyingexperience (“now and then” to “many times a week”). Addi-tionally, cross-sectional studies cannot prove causality since


0.760.67

0.680.630.63

0.650.58

0.730.74

0.600.68

0.660.63

0.26 0.36 0.46 0.56 0.66 0.76 0.86General population

MDDMDD-no workplace bullying

MDD-workplace bullyingRheumatoid arthritis

AA-no workplace bullyingAA-workplace bullying

PSOPSO-no workplace bullying

PSO-workplace bullyingIBD

IBD-no workplace bullyingIBD-workplace bullying

Figure 3: SF-6D scores in workers self-labelled as victims compared to those not reporting workplace bullying across different chronicdiseases. Results from previous studies are reported for comparison [65–69].

a necessary criterion of causation is the appropriate temporalrelationship between the hypothesized risk factors and out-comes. Finally, we do not have information concerning theattrition rate of both studies. As a consequence, we cannotexclude the notion that selection bias may have occurred.However, the consistency of productivity loss and quality-of-life estimates found in our study with those published in theliterature [49–52, 65–69] supports the validity of our results.

Future studies could collect data to assess psychosocialrisk factors which may influence the associations observed.Although we did not observe any interaction between diseasestatus and workplace bullying (i.e., the burden of bullying isconsistent across different disease populations), our resultsmay not be generalized to all workers with chronic con-ditions. Further studies could evaluate the humanistic andindirect burden of victimization at work among patientswith an expanded range of chronic medical conditions (e.g.,cardiovascular disease, diabetes, and chronic obstructivepulmonary disease).

5. Conclusions

Our findings demonstrate that the burden of workplace bul-lying on quality of life and productivity is substantial amongworkers with common and severe chronic diseases. Theseassociations were independent of the underlying medicalconditions (psoriasis, autoimmune arthritis, inflammatorybowel syndrome, and depression). This study provides keydata to inform policy-making and prioritization of occupa-tional health interventions.




Alice Fattori and Luca Neri contributed equally to this work.Alice Fattori prepared the first draft of the paper, contributedto the interpretation of the results, and approved the finalversion of the paper. Luca Neri participated in the design

of Liberamente and MOSAICO studies, developed the studyconcept for this paper, performed data analysis, contributedto the interpretation of the results, approved the final versionof the paper, and takes full responsibility for data integrity.Donatella Camerino and Giovanni Costa contributed to theinterpretation of the results and approved the final versionof the paper. All the other authors contributed to eitherthe design of Liberamente or MOSAICO studies, performedresearch fielding, contributed to the interpretation of theresults, and approved the final version of the paper.

Acknowledgment

The authors thank Doxa Pharma for assisting directors ateach study and providing studies coordination for researchfielding.

References

[1] J. E. Ilmarinen, Promoting Active Ageing in theWorkplace, Euro-pean Agency for Safety andHealth atWork, Bilbao, Spain, 2012.

[2] J. E. Ilmarinen, “Aging workers,”Occupational and Environmen-tal Medicine, vol. 58, no. 8, pp. 546–552, 2001.

[3] C. D. Mathers, “Global burden of disease in 2002: data sources,methods and results,” Global Programme on Evidence forHealth Policy Discussion Paper 54, WHO, 2003.

[4] Centers for Disease Control and Prevention (CDC),Healthy Aging at Work, 2014, http://www.cdc.gov/niosh/topics/healthyagingatwork/.

[5] M. Silverstein, “Meeting the challenges of an aging workforce,”American Journal of Industrial Medicine, vol. 51, no. 4, pp. 269–280, 2008.

[6] J. Seitsamo and M. Klockars, “Aging and changes in health,”Scandinavian Journal of Work, Environment & Health, vol. 23,no. 1, pp. 27–35, 1997.

[7] S. Einarsen, H. Hoel, D. Zapf, and C. L. Cooper, “The conceptof bullying and harassment at work: the European tradition,”in Bullying and Harassment in the Workplace: Developments inTheory, Research, and Practice, S. Einarsen, H. Hoel, D. Zapf,and C. L. Cooper, Eds., Taylor & Francis Group, 2nd edition,2011.

[8] A. Hogh, E. G. Mikkelsen, and A. M. Hansen, “Individualconsequences of workplace bullying/mobbing,” in Bullying and


Harassment in theWorkplace: Developments inTheory, Research,and Practice, S. Einarsen, H. Hoel, D. Zapf, and C. L. Cooper,Eds., Taylor and Francis Group, 2nd edition, 2011.

[9] E. G. Mikkelsen and S. Einarsen, “Bullying in Danish work-life:prevalence and health correlates,” European Journal ofWork andOrganizational Psychology, vol. 10, no. 4, pp. 393–413, 2010.

[10] S. B. Matthiesen and S. Einarsen, “Psychiatric distress andsymptoms of PTSD among victims of bullying at work,” BritishJournal of Guidance & Counselling, vol. 32, no. 3, pp. 335–356,2004.

[11] S. A. Bond,M.R. Tuckey, andM. F.Dollard, “Psychosocial safetyclimate, workplace bullying, and symptoms of posttraumaticstress,”Organization Development Journal, vol. 28, no. 1, pp. 37–56, 2010.

[12] L. R.-M.Hallberg andM.K. Strandmark, “Health consequencesof workplace bullying: experiences from the perspective ofemployees in the public service sector,” International Journal ofQualitative Studies on Health and Well-being, vol. 1, no. 2, pp.109–119, 2006.

[13] M. B. Nielsen, N. Magerøy, J. Gjerstad, and S. Einarsen, “Work-place bullying and subsequent health problems,” Tidsskrift forDen norske laegeforening, vol. 134, no. 12-13, pp. 1233–1238, 2014.

[14] S. Einarsen and E. G.Mikkelsen, “Individual effects of exposureto bullying at work,” in Bullying and Emotional Abuse in theWorkplace: International Perspectives in Research and Practice,S. Einarsen, H. Hoel, D. Zapf, and C. L. Cooper, Eds., pp. 127–144, Taylor & Francis, London, UK, 2003.

[15] M. B. Nielsen and S. Einarsen, “Outcomes of exposure toworkplace bullying: a meta-analytic review,” Work and Stress,vol. 26, no. 4, pp. 309–332, 2012.

[16] H. Hoel, M. J. Sheehan, C. L. Cooper, and S. Einarsenin,“Organisaional effects of workplace bullying,” in Bullying andHarassment in theWorkplace: Developments inTheory, Research,and Practice, S. Einarsen, H. Hoel, D. Zapf, and C. L. Cooper,Eds., Taylor & Francis, 2nd edition, 2011.

[17] European Agency for Safety and Health at Work (EU-OSHA),Calculating the Cost of Work-Related Stress and PsychosocialRisks, European Risk Observatory Literature Review, Pub-lications Office of the European Union, Luxembourg City,Luxembourgh, 2014.

[18] S. L. Gortmaker, D. K. Walker, M. Weitzman, and A. M. Sobol,“Chronic conditions, socioeconomic risks, and behavioral-problems in children and adolescents,” Pediatrics, vol. 85, pp.267–276, 1990.

[19] J. VanCleave andM.M.Davis, “Bullying and peer victimizationamong children with special health care needs,” Pediatrics, vol.118, no. 4, pp. e1212–e1219, 2006.

[20] R. Nordhagen, A. Nielsen, H. Stigum, and L. Kohler, “Parentalreported bullying among Nordic children: a population-basedstudy,” Child: Care, Health and Development, vol. 31, no. 6, pp.693–701, 2005.

[21] I. Pittet, A. Berchtold, C. Akre, P.-A. Michaud, and J.-C. Surıs,“Are adolescents with chronic conditions particularly at risk forbullying?” Archives of Disease in Childhood, vol. 95, no. 9, pp.711–716, 2010.

[22] L. B. Finne, S. Knardahl, and B. Lau, “Workplace bullying andmental distress—a prospective study of Norwegian employees,”Scandinavian Journal of Work, Environment & Health, vol. 37,no. 4, pp. 276–287, 2011.

[23] M. B. Nielsen, J. Hetland, S. B. Matthiesen, and S. Einarsen,“Longitudinal relationships between workplace bullying and

psychological distress,” Scandinavian Journal of Work, Environ-ment & Health, vol. 38, no. 1, pp. 38–46, 2012.

[24] I. Reknes, S. Pallesen, N. Magerøy, B. E. Moen, B. Bjorvatn, andS. Einarsen, “Exposure to bullying behaviors as a predictor ofmental health problems among Norwegian nurses: results fromthe prospective SUSSH-survey,” International Journal of NursingStudies, vol. 51, no. 3, pp. 479–487, 2014.

[25] M. Kivimaki, M. Virtanen, M. Vartia, M. Elovainio, J. Vahtera,and L. Keltikangas-Jarvinen, “Workplace bullying and the riskof cardiovascular disease and depression,” Occupational andEnvironmental Medicine, vol. 60, no. 10, pp. 779–783, 2003.

[26] M. Kivimaki, M. Elovainio, and J. Vahtera, “Workplace bullyingand sickness absence in hospital staff,” Occupational and Envi-ronmental Medicine, vol. 57, no. 10, pp. 656–660, 2000.

[27] G. A. Geuskens, K. M. O. Hengel, L. L. J. Koppes, and J. F.Ybema, “Predictors of thewillingness and the ability to continueworking until the age of 65 years,” Journal of Occupational andEnvironmental Medicine, vol. 54, no. 5, pp. 572–578, 2012.

[28] S. Gabrielle, D. Jackson, and J. Mannix, “Adjusting to personaland organisational change: views and experiences of femalenurses aged 40–60 years,” Collegian, vol. 15, no. 3, pp. 85–91,2008.

[29] K. Uegaki, M. C. de Bruijne, A. J. van der Beek, W. vanMechelen, and M. W. van Tulder, “Economic evaluations ofoccupational health interventions from a company’s perspec-tive: a systematic review of methods to estimate the costof health-related productivity loss,” Journal of OccupationalRehabilitation, vol. 21, no. 1, pp. 90–99, 2011.

[30] M. B. Nielsen, G. Notelaers, and S. Einarsen, “Measuringexposure to workplace bullying,” in Bullying and Harassment inthe Workplace: Developments in Theory, Research, and Practice,S. Einarse, H. Hoel, D. Zapf, and C. L. Cooper, Eds., Taylor &Francis Group, 2nd edition, 2011.

[31] S. Einarsen and A. Skogstad, “Bullying at work: epidemiologicalfindings in public and private organizations,” European JournalofWork andOrganizational Psychology, vol. 5, no. 2, pp. 190–191,1996.

[32] G. Apolone, P. Mosconi, and L. Quattrocchi, Questionario SulloStato di Salute SF-12, Versione Italiana, 2005, edited by: Guerinie Associati.

[33] J. E. Ware Jr., M. Kosinski, and S. D. Keller, “A 12-item short-form health survey: construction of scales and preliminary testsof reliability and validity,”Medical Care, vol. 34, no. 3, pp. 220–233, 1996.

[34] R. M. D’Souza, L. Strazdins, L. L.-Y. Lim, D. H. Broom, andB. Rodgers, “Work and health in a contemporary society:demands, control, and insecurity,” Journal of Epidemiology andCommunity Health, vol. 57, no. 11, pp. 849–854, 2003.

[35] J. E. Brazier, R. Harper, N. M. B. Jones et al., “Validating theSF-36 health survey questionnaire: new outcome measure forprimary care,” British Medical Journal, vol. 305, no. 6846, pp.160–164, 1992.

[36] M. C. Reilly, A. S. Zbrozek, and E. M. Dukes, “The validity andreproducibility of a work productivity and activity impairmentinstrument,” PharmacoEconomics, vol. 4, no. 5, pp. 353–365,1993.

[37] World Bank, World Development Indicators Database, 2011,http://siteresources.worldbank.org/DATASTATISTICS/Resour-ces/GNIPC.pdf.

[38] K. Tuomi, J. Ilmarinen, A. Jahkola, L. Katajarinne, and A.Tulkki, Work Ability Index, Finish Institute of OccupationalHealth, Helsinki, Finland, 2nd edition, 1998.


[39] International Labour Organization (ILO), “Resolution con-cerning the International Classification of Status in Employ-ment (ICSE),” in Proceedings of the 15th Interna-tional Conference of Labour Statisticians, 1993, http://www.ilo.org/global/statistics-and-databases/statistics-overview-and-topics/status-in-employment/current-guidelines/lang–en/index.htm.

[40] R. C. Kessler and E. J. Bromet, “The epidemiology of depressionacross cultures,”Annual Review of PublicHealth, vol. 34, pp. 119–138, 2013.

[41] R. Di Domenicantonio, G. Cappai, M. Arca et al., “Occurrenceof inflammatory bowel disease in central Italy: a study based onhealth information systems,”Digestive and LiverDisease, vol. 46,no. 9, pp. 777–782, 2014.

[42] R. Saraceno, R.Mannheimer, and S. Chimenti, “Regional distri-bution of psoriasis in Italy,” Journal of the European Academy ofDermatology and Venereology, vol. 22, no. 3, pp. 324–329, 2008.

[43] F. Salaffi, R. De Angelis, W. Grassi, MArche Pain Prevalence,and INvestigation Group (MAPPING) study, “Prevalence ofmusculoskeletal conditions in an Italian population sample:results of a regional community-based study. I. The MAPPINGstudy,” Clinical and Experimental Rheumatology, vol. 23, no. 6,pp. 819–828, 2005.

[44] G. Giorgi, A. Arenas, and J. M. Leon-Perez, “An operativemeasure of workplace bullying: the negative acts questionnaireacross Italian companies,” Industrial Health, vol. 49, no. 6, pp.686–695, 2011.

[45] H. Leymann, “The content and development of mobbing atwork,” European Journal of Work and Organizational Psychol-ogy, vol. 5, no. 2, pp. 165–184, 1996.

[46] L. Keashly and K. Jagatic, “The nature, extent and impactof emotional abuse in the workplace: results of a state-widesurvey,” in Proceedings of the Meeting of the Academy ofManagement, Toronto, Canada, 2002.

[47] M. B. Nielsen, S. B. Matthiesen, and S. Einarsen, “The impactof methodological moderators on prevalence rates of workplacebullying. Ameta-analysis,” Journal of Occupational and Organi-zational Psychology, vol. 83, no. 4, pp. 955–979, 2010.

[48] M. B. Nielsen, Metodological issues in research on workplacebullying: operationalizations, measurements and samples [Ph.D.thesis], University of Bergen, Bergen, Norway, 2009.

[49] K. Kruntoradova, J. Klimes, L. Sedova, J. Stolfa, T. Dolezal,and A. Petrıkova, “Work productivity and costs related topatients with ankylosing spondylitis, rheumatoid arthritis, andpsoriasis,” Value in Health Regional Issues C, vol. 4, pp. 100–106,2014.

[50] A. Beck, A. L. Crain, L. I. Solberg et al., “Severity of depressionandmagnitude of productivity loss,”Annals of Family Medicine,vol. 9, no. 4, pp. 305–311, 2011.

[51] M. Vergara, A. Montserrat, F. Casellas et al., “A new validationof the spanish work productivity and activity impairmentquestionnaire-Crohn’s disease version,” Value in Health, vol. 14,no. 6, pp. 859–861, 2011.

[52] A. B. Kimball, J. Signorovitch, J. Xie et al., “The effects ofadalimumab treatment and psoriasis severity on self-reportedwork productivity and activity impairment for patients withmoderate to severe psoriasis,” Journal of the American Academyof Dermatology, vol. 66, no. 2, pp. e67–e76, 2012.

[53] S. Leka, M. Vartia, J. Hassard et al., “Best practice in interven-tions for the prevention andmanagement of work-related stressand qorkplace violence and bullying,” in The European Frame-work for Psychosocial RiskManagement: PRIMA-EF, S. Leka and

T. Cox, Eds., Institute of Work, Health and Organizatons (I-WHO), Nottingham, UK, 2008.

[54] H. Hoel, C. Rayner, and C. L. Cooper, “Workplace bullying,” inInternational Review of Industrial and Organizational Psychol-ogy, C. L. Cooper and I. T. Robertson, Eds., pp. 195–230, JohnWiley & Sons, Chichester, UK, 1999.

[55] A. Ortega, K. B. Christensen, A. Hogh, R. Rugulies, and V.Borg, “One-year prospective study on the effect of workplacebullying on long-term sickness absence,” Journal of NursingManagement, vol. 19, no. 6, pp. 752–759, 2011.

[56] P. J. O’Connell, E. Calvert, and D. Watson, Bullying in theWorkplace: Survey Reports, Department of Enterprise Trade andEmployment, Economic and Social Research Institute, Dublin,Ireland, 2007.

[57] C. M. Berry, A. M. Lelchook, andM. A. Clark, “Ameta-analysisof the interrelationships between employee lateness, absen-teeism, and turnover: implications for models of withdrawalbehavior,” Journal of Organizational Behavior, vol. 33, no. 5, pp.678–699, 2012.

[58] N. R. Ziebarth, “Long-term absenteeism and moral hazard—evidence from a natural experiment,” Labour Economics, vol. 24,pp. 277–292, 2013.

[59] M. Sliter, K. Sliter, and S. Jex, “The employee as a punchingbag: the effect of multiple sources of incivility on employeewithdrawal behavior and sales performance,” Journal of Orga-nizational Behavior, vol. 33, no. 1, pp. 121–139, 2012.

[60] D. J. Lerner, S. Levine, S. Malspeis, and R. B. D’Agostino, “Jobstrain and health-related quality of life in a national sample,”American Journal of Public Health, vol. 84, no. 10, pp. 1580–1585,1994.

[61] S. A. Stansfeld, H. Bosma, H. Hemingway, and M. G. Mar-mot, “Psychosocial work characteristics and social support aspredictors of SF-36 health functioning: the Whitehall II study,”Psychosomatic Medicine, vol. 60, no. 3, pp. 247–257, 1998.

[62] L. Neri, L. A. R. Rey, M. Gallieni et al., “Occupational stress isassociated with impaired work ability and reduced quality oflife in patients with chronic kidney failure,” The InternationalJournal of Artificial Organs, vol. 32, no. 5, pp. 291–298, 2009.

[63] C. Fekete, M. Wahrendorf, J. D. Reinhardt, M. W. M. Post,and J. Siegrist, “Work stress and quality of life in persons withdisabilities from four European countries: the case of spinalcord injury,”Quality of Life Research, vol. 23, no. 5, pp. 1661–1671,2014.

[64] S. J. Walters and J. E. Brazier, “What is the relationship betweenthe minimally important difference and health state utilityvalues? The case of the SF-6D,” Health and Quality of LifeOutcomes, vol. 1, article 4, 2003.

[65] J. Brazier, J. Roberts, A. Tsuchiya, and J. Busschbach, “Acomparison of the EQ-5D and SF-6D across seven patientgroups,” Health Economics, vol. 13, no. 9, pp. 873–884, 2004.

[66] S. A. H. Gerhards, L. E. De Graaf, L. E. Jacobs et al., “Economicevaluation of online computerised cognitive-behavioural ther-apywithout support for depression in primary care: randomisedtrial,” The British Journal of Psychiatry, vol. 196, no. 4, pp. 310–318, 2010.

[67] C. A. Marra, J. C. Woolcott, J. A. Kopec et al., “A comparison ofgeneric, indirect utility measures (the HUI2, HUI3, SF-6D, andthe EQ-5D) and disease-specific instruments (the RAQoL andthe HAQ) in rheumatoid arthritis,” Social Science & Medicine,vol. 60, no. 7, pp. 1571–1582, 2005.

[68] V. Strand, V. Sharp, A. S. Koenig et al., “Comparison of health-related quality of life in rheumatoid arthritis, psoriatic arthritis


and psoriasis and effects of etanercept treatment,” Annals of theRheumatic Diseases, vol. 71, no. 7, pp. 1143–1150, 2012.

[69] M. J. Buxton, L. A. Lacey, B. G. Feagan, T. Niecko, D. W.Miller, and R. J. Townsend, “Mapping from disease-specificmeasures to utility: an analysis of the relationships between theinflammatory bowel disease questionnaire and Crohn’s diseaseactivity index in Crohn’s disease and measures of utility,” Valuein Health, vol. 10, no. 3, pp. 214–220, 2007.

[70] W. B. F. Brouwer, N. J. A. van Exel, M. A. Koopmanschap, and F.F. H. Rutten, “Productivity costs before and after absence fromwork: as important as common?” Health Policy, vol. 61, no. 2,pp. 173–187, 2002.

Research ArticlePrognostic Factors of Returning to Work after SickLeave due to Work-Related Common Mental Disorders:A One- and Three-Year Follow-Up Study

Bo Netterstrøm, Nanna Hurwitz Eller, and Marianne Borritz

Department of Occupational and Environmental Medicine, Bispebjerg University Hospital, Bispebjerg Bakke 23,2400 Copenhagen NV, Denmark

Correspondence should be addressed to Bo Netterstrøm; [email protected]

Received 15 January 2015; Revised 14 March 2015; Accepted 20 March 2015


Copyright © 2015 Bo Netterstrøm et al.This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The aim of this paper was to assess the prognostic factors of return to work (RTW) after one and three years among people onsick leave due to occupational stress.Methods. The study population comprised 223 completers on sick leave, who participated in astress treatment program. Self-reported psychosocial work environment, life events during the past year, severity of the condition,occupational position, employment sector, marital status, and medication were assessed at baseline. RTW was assessed with datafrom a national compensation database (DREAM). Results. Self-reported high demands, low decision authority, low reward, lowsupport from leaders and colleagues, bullying, high global symptom index, length of sick leave at baseline, and stressful negativelife events during the year before baseline were associated with no RTW after one year. Low work ability and full-time sick leave atinclusion were predictors after three years too. Being single was associated with no RTW after three years. The type of treatment,occupational position, gender, age, and degree of depression were not associated with RTWafter one or three years.Conclusion.Theimpact of the psychosocial work environment as predictor for RTW disappeared over time and only the severity of the conditionwas a predictor for RTW in the long run.

1. Introduction

Work-related common mental disorders such as stressaccount for a significant portion of sick leave in modernsociety. Stress conditions are associated with great personalsuffering as well as economic problems due to sick leave[1]. Additionally, sick leave is a major risk factor for earlywithdrawal from the labor market [2] with reports of only50% of people on sick leave for more than six months formental health disorders return to work (RTW) [3]. Thesefindings have led to growing interest in the evaluation of stressmanagement interventions and their effect on RTW [4, 5].

A number of reviews and meta-analyses including aCochrane review have reviewed randomized controlled trialsof stress treatment programs and concluded that they aremore effective at symptom reduction than no treatment. Itwas also determined that cognitive behavioral therapy, CBT,is particularly more effective than other therapies in reducing

symptoms [6–9]. However, findings for the impact of CBT onRTWare inconsistent and do not support a significant impactof CBT on RTW [6].

The inconsistent findings for RTW as an outcomemay bedue to considerable heterogeneity in jurisdictional contextssuch as national differences in labor market regulations andofficial sick leave policies, which hamper the ability to com-pare study findings from different countries [10, 11]. Therecan also be considerable heterogeneity in the individualsincluded in studies of RTWwith regard to the course of stressdevelopment and reasons for being stressed including bothprivate and work-related stressors and coping with stress.Many studies on the effect of stress treatment programs haveincluded volunteers from a certain workplace or organizationbut have not used sick leave as inclusion criteria. This toomay lead to inconclusive findings for RTW outcomes asparticipants may not be sufficiently impaired at inclusion toshow improvement [5, 12–14].


http://dx.doi.org/10.1155/2015/596572


Between 2010 and 2012, a stress treatment study wasconducted at the Department of Occupational and Envi-ronmental Medicine at the Bispebjerg Hospital in Denmark.Individuals on sick leave due to stress were randomized intoone of four treatment groups: (I) group-based psychody-namic therapy and body awareness; (II) individual problemsolution therapy (PST) [4] with an eight-week mindfulnesscourse; (III) control treatment of individual therapywith psy-chologists outside the study team (treatment as usual (TAU));and (IV) waitlist control group (WLCG), who received PSTafter three months. Initially, the data were analyzed to revealany differences in the effects of the various interventions onsymptom reduction and RTW.The two interventions ((I) and(II)) led to significant improvements in symptoms comparedto the waitlist group [15, 16]. In addition, the RTW rates forthe intervention groups were significantly higher after threemonths compared to both control groups [15, 16]. We havenow followed the participants for three years from inclusionto the study in order to evaluate the long-term effect oftreatment and other prognostic factors measured at baselinefor RTW.


From June 2010 to December 2010, all general practitionersin the Capital Region of Denmark (1.6 million inhabitants)were invited to refer patients with work-related commonmental disorder to our project. The purpose of the studyand criteria for participation were described in the invitation.The inclusion criteria were as follows: the participant had to(1) be on full-time or part-time sick leave, (2) be employedor self-employed, (3) have significant symptoms of work-related common mental disorder for at least 2 months,and (4) be motivated to participate. The exclusion criteriawere (1) current abuse of alcohol or psychoactive stimulants,(2) major psychiatric disorder, and (3) significant somaticdisorder assumed to be the primary cause of the stresscondition. Details regarding treatment and methods havebeen previously described [15, 16].

All procedures followed were in accordance with theHelsinki Declaration of 1975, as revised in October 2013.

2.1. Dependent Variable

2.1.1. RTW. Data on employment status one and three yearsafter inclusion in the study was obtained from the DREAMdatabase (Danish Register for Evaluation of Marginalization[10, 22]). DREAM is a registry of the Labor Market Authorityof all public transfer payments. It contains data on all Danes,including those who receive economic compensation due tosick leave, unemployment, retirement, and so forth. The dataon sick leave are reported as soon as the employer reports acase of sick leave of duration of two weeks or more amonghis or her employees. Sick leave compensation normallyterminates after one year according to the regulation. Manypeople on sick leave thereafter are transferred to othercompensation systems. Therefore, we only considered a caseto have returned to the labor market if there was actually no

compensation of any type at the time of census. This meantthat there were two possible assessments: (1) work, that is,full-time or part-time before sick leave, or (2) case, that is, sickleave (part-time incl.), unemployment, education, maternityleave, retirement, or death. The data in the database did notallow us to distinguish between full-time and part-time sickleave. After one year, 67 pct. were at work, 17 pct. were onsick leave (full- and part-time incl.), 11 pct. were unemployed,3 pct. were under education, and 2 pct. had retired. Thesefigures were almost the same after three years.

2.2. Independent Variables

2.2.1. PsychosocialWork Environment Risk Characteristics. Asthe work environment was believed to be the reason for sickleave, the variables describing the work environment werethose believed to be of greatest significance to RTW. Theexplaining variables measured at baseline were the following.

We used the full scales on demands, decision authority,skill discretion, meaningfulness, predictability, reward, roleclarity, justice, and social support from leaders and colleaguesfrom the Copenhagen Psychosocial Questionnaire (COP-SOQ) [17]. As part of the sessions during the intervention,the stressors were evaluated, and the participant rated thestressors in collaboration with the therapist. Ratings rangedfrom 1 = no or low influence to 4 = very high influence.The possible work-related stressors were bad management,bad work environment, reorganization, and work pressure.The scores from the work-related stressors were summed anddivided by four to calculate the work environment factorindex (maximum score 4). The participants were also askedwhether bullyingwas a stressor.This issuewas not included inthe averagedmeasure because we find this stressor verymuchdifferent from reorganization and work pressure.

2.2.2. Life Events during the Last Year. However, it is wellknown that people suffering from stress experience stressorsin both their work and their private lives [18, 19]. Thebaseline questionnaire also included information about lifeevents during the last year, that is, problems with colleagues,getting fired, death in family, divorce, and economic troubles(yes/no). Life events were summed and divided by five tocalculate the life event one-year score (maximum score 5).

2.2.3. Demographics. Gender, age, occupational position(blue, white collar, or academic worker), employment in thepublic or private sector, and marital status were recorded.

The type of treatment in the program was also used as anindependent variable.

2.2.4. Seriousness of Stress Conditions. The seriousness ofstress conditions was estimated in several ways.

(a) The form of sick leave, that is, being on full-time sickleave at time of inclusion or not and number of dayson sick leave before attending the project.

(b) Work ability, measured on a scale ranging from 0 to10 using the following question: “Assess your work


Table 1: Analyses of excluded participants and drop-outs compared to those who fulfilled treatment. Excluded comprises those excluded dueto major psychiatric disease (𝑁 = 17) and absence from treatment (𝑁 = 16). Drop-outs are those who did not attend the treatment at alleven though they fulfilled the inclusion criteria.

Intervention Excluded 𝑃 Drop-outs 𝑃

𝑁 223 33 12Age, mean (SD) 44.2 (8.8) 43.2 (10.8) 0.60 40.9 (7.7) 0.19Women % 80.3 65.2 0.08 66.7 0.21Blue collar workers % 34.3 30.8 0.95 (NA) (NA)Married % 62.2 53.8 0.37 (NA) (NA)Global symptom index (GSI) (SCL92), mean (SD) 1.23 (0.54) 1.66 (0.75) 0.005 (NA) (NA)Depression score (SCL92), mean 1.74 (0.76) 2.42 (1.02) 0.002 (NA) (NA)At work after 1 year % 67.3 51.8 0.20 41.7 0.07At work after 3 years % 68.9 54.5 0.31 34.8 0.04

ability as ten points when you are at your best. Howdo you rate your work ability currently?”

(c) The degree of stress, measured using the followingquestion: “‘Stress’ is a condition characterized byunrest, agitation, or anxiety and/or sleeping prob-lems. Do you experience stress at themoment?”Therewere five options for answers ranging from “not at all”to “always” [20].

(d) The SCL92, a validated scale on 92 questions onpsychological symptoms, used to calculate the globalsymptom index (GSI) and its nine subscales accord-ing to [21].

(e) Medication with antidepressants at baseline (yes/no).(f) Alcohol consumption.

2.3. Statistical Analyses. First, an analysis of the differencesbetween the participants of the study, the excluded and thedrop-outs, was conducted as well as the mean and standarddeviation (SD) of the demographic variables to describe theparticipants. Next two sets of analysis were conducted. First,the participants at work (𝑁 = 150) were compared to cases(𝑁 = 73) after one year. Next, those at work at both theone- and the three-year census (𝑁 = 111) were comparedto those who were cases at both times (𝑁 = 32). A Student’s𝑡-test was used to evaluate the differences in continuousvariables, and a chi-square test was used in connectionwith the categorical variables. A series of bivariate logisticregression analyses were conducted to reveal if the variableswere significantly associated with the outcome. Correlationsof the explanatory variables, covariates, and outcome wereanalyzed to reveal any multicollinearity between the vari-ables. Next, several multivariate logistic regression analyseswith RTW at one- and three-year follow-up as dependentvariable were conducted.The included independent variableswere chosen so that multicollinearity was not present. Thechosen variables were decision authority, bullying, workability index, and full-time/part-time sick leave at baseline.In model 1, the adjustment factors were age, gender, maritalstatus, and occupational position. Model 2 included GSIand model 3 in addition life events. These analyses wererepeated in general linear models (GLM) in order to evaluate

any interactions between the independent variables. Finally,in model 4 multivariate logistic regression analyses wereconducted with the four chosen independent variables forcedinto the model at the same time.

3. Results

From August 4, 2010, to April 8, 2011, 320 potential partic-ipants were referred to the study, of which 268 fulfilled theinclusion criteria. The procedure of randomization of treat-ment has been previously described [15]. However, only 223individuals completed the treatment (87.8%). Twelve peopledid not show up or decided that they did not want to par-ticipate (drop-outs). Of the remaining 33, 17 were excludeddue to major psychiatric disorder during the first weeks oftreatment, and 16 participants were excluded because theydid not complete the treatment or were absent more thantwo times during the duration of treatment. Table 1 shows thecharacteristics of the participants in the intervention groupcompared to the individuals who were excluded or did dropout. The excluded persons were predominately women, wereless educated, and had higher symptom level scores but wereon the other variables comparable to the intervention group.

At one-year follow-up, 150 participantswereworking full-time, whereas 73 were not. After three years, 111 of the 150participants were still working, whereas only 32 were beinga case both years. The results of the one-year follow-up ofthe bivariate 𝑡-tests including one explanatory variable at atime with RTW as a dependent variable showed that severalwork environment risk factors were significantly associatedwith not being at work after one year. High demands, lowdecision authority, low reward, low support from leadersand colleagues, and being bullied were all self-reportedbaseline risk factors among those not being at work. Thework environment index and life events one-year index wereassociated with no RTW. The seriousness of the condition inthe form of GSI, work ability index, full-time sick leave atinclusion, and number of days on sick leave before inclusionwas significantly greater in the group that had not returned atwork after one year (Table 2).

In contrast, the only factors significantly associated withbeing a case after both one and three years were being on


Table2:Bivaria

teassociations

betweenindepend

entfactorsatbaselin

eand

RTW

full-tim

eorn

oton

eand

threey

earsaft

erbaselin

e.

One

year

Both

onea

ndthreey

ears

Full-tim

ejob

Case

𝑃Fu

ll-tim

ejob

Case

𝑃

𝑁150

73111

32Dem

ograph

ics

Age,m

ean(SD)

44.8(8.6)

42.8(9.7)

0.14

44.7(7.8)

43.6(11.1)

0.52

Wom

en%

79.3

82.2

0.38

74.6

70.7

0.68

Blue

collar/whitecollar/academ

icworker%

34.9/49.7

/15.6

37.3/53.3/9.3

0.41

30.6/54.8/14.5

39.5/52.6/7.9

0.42

Publicsector

%68.4

71.2

0.71

68.2

74.5

0.46

Marrie

d%

64.6

57.1

0.30

62.4

35.9

0.005

Alcoh

ol(U

/week),m

ean(SD)

4.3(5.1)

3.6(4.6)

0.29

4.3(4.5)

3.9(5.0)

0.66

Treatm

entg

roup

%I

70.0

30.0

0.87

80.4

19.6

0.68

II63.8

36.2

71.2

28.6

TAU(III)

64.8

35.2

71.8

28.2

WLC

G(IV)

65.5

34.5

70.6

29.4

Psycho

socialworkenvironm

ent

Highdemands,m

ean(SD)

61.9(12.4)

66.6(15.5)

0.02

61.4(12.2)

65.6(15.6)

0.11

Lowdecisio

nandauthority,m

ean(SD)

56.1(21.7

)64

.7(22.0)

0.007

56.6(21.5

)62.8(22.2)

0.16

Lowskill

discretio

n,mean(SD)

32.5(16.0)

35.3(16.5)

0.24

32.3(15.5)

35.9(16.2)

0.24

Lowmeaning

fulness,mean(SD)

32.1(19

.6)

32.0(21.2

)0.98

32.2(20.5)

32.2(20.9)

0.99

Lowpredictability,mean(SD)

55.1(23.5)

61.6(22.9)

0.06

56.6(23.0)

56.6(21.1)

0.91

Lowreward,mean(SD)

49.7(24.2)

57.4(24.7)

0.03

51.9(24.1)

57.2(26.9)

0.29

Lowrolecla

rity,mean(SD)

43.5(22.3)

42.0(22.4)

0.67

44.1(22.3)

37.2(18.2)

0.11

Lowsupp

ortfrom

leader,m

ean(SD)

52.3(24.9)

60.2(26.4)

0.03

53.6(25.0)

56.4(27.7

)0.57

Lowdegree

ofjustice,mean(SD)

55.3(20.4)

59.0(19.8

)0.21

57.3(20.8)

60.2(20.1)

0.50

Bullying,mean(SD)

1.4(0.9)

1.7(1.1)

0.05

1.4(0.9)

1.6(1.1)

0.41

Workenvironm

entind

ex(4

items),m

ean(SD)

2.3(0.4)

2.4(0.6)

0.05

2.2(0.4)

2.3(0.5)

0.50

Lifeeventslastyear

score(5items),m

ean(SD)

1.3(1.9)

1.4(2.3)

0.02

1.5(1.8)

1.5(1.8)

0.89

Severityof

cond

ition

Globalsym

ptom

index(G

SI)(SC

L92),m

ean(SD)

1.2(0.5)

1.3(0.5)

0.05

1.23(0.5)

1.3(0.6)

0.35

Somaticsymptom

score(SC

L92),m

ean(SD)

1.3(0.8)

1.5(0.8)

0.17

1.4(0.8)

1.50.15

Depressionscore(SC

L92),m

ean(SD)

1.7(0.8)

1.8(0.8)

0.56

1.7(0.8)

1.8(0.8)

0.57

Anx

ietyscore(SC

L92),m

ean(SD)

1.3(0.7)

1.4(0.7)

0.19

1.4(0.7)

1.4(0.7)

0.98

Antidepressantm

edication%

15.9

18.8

0.69

15.2

17.2

0.78

Workabilityindex,mean(SD)

2.8(2.2)

1.8(2.0)

0.001

2.9(2.2)

1.7(2.1)

0.04

Stress,m

ean(SD)

3.7(0.9)

4.0(0.9)

0.11

3.8(0.9)

4.1(0.9)

0.19

Full-tim

esickleavea

tbaseline

%62.5

76.8

0.03

61.5

82.1

0.02

Daysw

ithfull-tim

esickleavea

tbaseline,m

ean(SD)

40.9(49.4

)57.0(80.9)

0.04

43.6(51.2

)46

.4(58.4)

0.68

TAU:treatmentasu

sual(III).

WLC

G:w

aitlist

controlgroup

(IV).


Table 3: Correlations between RTW after one year and independent variables at baseline.

1 2 3 4 5 6 7 8 9 10 11 12 13 141 RTW after one year 1 −0.160 −0.181 −0.130 −0.146 −0.147 −0.141 −0.171 −0.124 −0.133 0.208 −0.103 −0.145 −0.1232 High demands 1 0.194 0.134 0.156 0.150 0.078 0.186 −0.066 −0.028 −0.083 0.210 0.054 0.0453 Low decisionauthority 1 0.351 0.381 0.253 0.232 0.296 −0.006 0.259 −0.032 0.194 0.053 −0.084

4 Low predictability 1 0.603 0.609 0.397 0.211 0.117 0.187 −0.086 0.222 0.176 0.1475 Low rewards 1 0.737 0.486 0.265 0.246 0.276 −0.044 0.217 0.182 0.0676 Low support fromleader 1 0.447 0.189 0.130 0.195 −0.101 0.165 0.160 0.094

7 Low support fromcolleagues 1 0.182 0.215 0.232 −0.063 0.110 −0.130 0.096

8 Work environmentindex 1 0.205 0.389 −0.137 0.248 0.001 −0.015

9 Life event index 1 0.202 −0.008 0.197 0.045 −0.02810 Bullying 1 0.031 0.179 −0.069 −0.13011 High work abilityindex 1 −0.253 −0.304 −0.079

12 Global symptomindex 1 0.079 −0.075

13 Full-time sick leaveat baseline 1 0.182

14 Days of sick leaveat baseline 1

Bold: 𝑃 < 0.05. Bold and italic: 𝑃 < 0.01.

Table 4: Logistic regression analyses of prognostic variables for RTW after one year.

Prognostic variables Model 1 Model 2 Model 3 Model 4OR (95% CI) OR (95% CI) OR (95% CI) OR (95%)

Low decision authority 0.982 (0.970–0.999) 0.987 (0.971–1.004) 0.986 (0.971–1.002) 0.990 (0.975–1.007)Bullying 0.715 (0.513–0.994) 0.677 (0.474–0.966) 0.634 (0.426–0.926) 0.731 (0.513–1.040)Work ability index 1.220 (1.038–1.434) 1.285 (1.074–1.537) 1.356 (1.113–1.612) 1.235 (1.029–1.401)Full-time sick leave at baseline 0.431 (0.222–0.830) 0.430 (0.223–0.830) 0.454 (0.234–0.880) 0.511 (0.251–1.010)In models 1–3, the prognostic variable was analyzed separately with adjustment for the following.Model 1: age, gender, marital status, and occupational position.Model 2: model 1 and global symptom index.Model 3: model 2 and life events.Model 4: all four prognostic variables forced into themodel adjusted for age, gender, occupational position, global symptom index, marital status, and life eventlast year.

full-time sick leave at baseline, low work ability index score,and being single.

The correlation analysis of data shown in Table 3 demon-strated that the number of days of sick leave at time ofinclusion in the study was not correlated with any of thescales measuring degree of symptoms. The scales measuringdegree of symptoms were all significantly intercorrelated. Inaddition, the different measures of work environment werecorrelated, and the data on symptom degree were correlatedwith scales of work environment. Therefore, the data setshowed multiple collinearity problems.

The multivariate logistic regression analyses showed thathigh work ability index, bullying, and full-time sick leaveprior to inclusion in the study were significantly associatedwith RTW after one year after full adjustment (Table 4). Lowdecision authority remained significant after adjustment for

age, gender, and occupational position, but not after fur-ther adjustments. No interactions between the independentvariables were found in the GLM-analyses. However, onlywork ability remained significant with RTW after one year,when the chosen independent variables were forced into themodel with full adjustments (Table 4). Demands and socialsupport from leaders as well as colleagues were significantlyassociated with RTW after adjustment for age, gender, andoccupational position whereas the predictability and rewardswere only borderline significant (0.1 > 𝑃 > 0.05) (datanot shown). Life events were not significantly associated withRTW after adjustment for age, gender, and occupationalposition.

After three years only full-time sick leave at baseline andlow work ability were significantly negatively associated withRTW even after adjustments as above (data not shown).


4. Discussion

In this study of predictors of being at work one year afterinclusion in a stress treatment project, we determined thatself-reported psychosocial work environment, life events,part-time sick leave, and length of sick leave at baseline wereof significance to being at work, whereas type of treatment,employment grade, degree of depression, and life events werenot.The association between psychosocial work environmentfactors and RTW disappeared after three years.

The outcome was “being at work or not.” The data toachieve the outcome were collected fromDREAM, a nationalregister of public compensation. However, the validity ofthese data may be questioned [10, 11, 22]. The participantswho were in the “not being at work” group may be verydifferent as some are in fact still disabled from stress, whereasothers are on education, have retired, or are not employed.It was not possible to select only those still disabled due tostress due to the categories in the register. The results shouldtherefore be read as predictors for “being at work or beingsomething else” and not “being at work or being disabled bystress.” A bad work environment may indeed be a reason forwanting to pursue further education or retire, but nonethelessa positive working environment predicts RTW after one year.

Themost important result was that working environmenthas significance to an early RTW. Those who were notreturned to the workplace one year after inclusion in theproject had reported significantly more demands and lessdecision authority, reward, and support at baseline, thanthe group which was working full-time after one year. Thechance of getting back is larger if you felt comfortable atwork before getting stressed. An early return to work isimportant as this prevents withdrawal from the labor market[2, 3]. Also, a successful RTW-process is a success to theworkplace and may prevent other cases of long time sickleave.When the process has a positive result thiswill spread inthe organization. However, when the working environment isbad the chance of getting back is smaller, the RTW-process ishard, and you may be squeezed out of the workplace. Thiscould not be demonstrated by the analyses including only theparticipantswhowere full-timeworkers at both one and threeyears compared to those who were cases both years. The datafrom the DREAM database did not give us the opportunityto analyze part-time employees separately. However, thosewho were cases at both years had a nonsignificant tendencyof experiencing high demands at baseline.

We chose to include variables related to life events, asconditions in private life may also be of significance to RTWand may delay RTW. However, it was not possible to revealany significance of these variables as they relate to RTW.

The self-reported psychosocial working environment wasof some significance to the outcome, as high demands, lowdecision authority, bullying, low rewards, and low socialsupport from leaders and colleagues were all significantlyassociated with “not being at work.” However, after adjust-ment for severity of the condition, these associations wereinsignificant. The seriousness of the stress condition in theform of number of days on sick leave, being on full-time sickleave, GSI, and most pronounced work ability index was also

significantly associated with RTW. A recent Danish paper hasreported similar results on the association between depressivesymptoms and long-term sickness absence, but in that studypoor psychosocial work environment did not predict sicknessabsence [23]. However, the psychosocial work environmentwas assessed by the use of unit level aggregated measures onwork environment. Though structural conditions may be thesame for several individuals working in the same unit of anorganization, the work environment may be perceived verydifferently by single individuals.This difference in perceptionmay be the reason that our findings differ from the findingsby Hjarsbech et al.

Low social support from leaders, low social support fromcolleagues, and bullying were all associated with RTW afterone year. In line with this finding are the findings by Arendset al. that associate conflicts with a superior with recurrentsick leave [4]. If you expect problems at the workplace, it is ofcourse not easy to return.

The finding that the type of treatment was not associatedwith RTW after one and three years is in accordance withearlier findings [6]. This might reflect the fact that althoughtreatment accelerates the RTW-process, the severity of thecondition and other factors are more important in the longrun [4, 5, 11].

In this study, the severity of the disorder (full-time sickleave and poor self-rated work ability) was found to be themain predictor in the long run in addition to being single,which is a main finding in many studies on the relationbetween marital status and disease.

Our results may be questioned, as the study has severalweaknesses.The study size is rather small, including only 223participants who fully completed the study. This includes therisk of determining findings by chance, as a small numberof participants may completely change the results. However,the strength of the study is the well-validated outcome andextensive exposure measures.

Consent

Informed consentwas obtained fromall patients for inclusionin the study. Feedback to the referral GP was given for allparticipants.


N. H. Eller, M. Borritz, and B. Netterstrøm declare that theyhave no conflict of interests.

Acknowledgments

The study was funded by the TrygFonden and the DanishWork Environment Fund.

References

[1] R. Z. Goetzel, S. R. Long, R. J. Ozminkowski, K. Hawkins,S. Wang, and W. Lynch, “Health, absence, disability, and pre-senteeism cost estimates of certain physical and mental health


conditions affecting U.S. employers,” Journal of Occupationaland Environmental Medicine, vol. 46, no. 4, pp. 398–412, 2004.

[2] M. Henderson, N. Glozier, and K. H. Elliott, “Long termsickness absence,” BritishMedical Journal, vol. 330, no. 7495, pp.802–803, 2005.

[3] M. F. Andersen, K. M. Nielsen, and S. Brinkmann, “Meta-synthesis of qualitative research on return to work amongemployees with commonmental disorders,” Scandinavian Jour-nal of Work, Environment and Health, vol. 38, no. 2, pp. 93–104,2012.

[4] I. Arends, J. J. L. van der Klink, W. van Rhenen, M. R. de Boer,and U. Bultmann, “Predictors of recurrent sickness absenceamong workers having returned to work after sickness absencedue to common mental disorders,” Scandinavian Journal ofWork, Environment & Health, vol. 40, no. 2, pp. 195–202, 2014.

[5] L. Blank, J. Peters, S. Pickvance, J. Wilford, and E. MacDonald,“A systematic review of the factors which predict return to workfor people suffering episodes of poor mental health,” Journal ofOccupational Rehabilitation, vol. 18, no. 1, pp. 27–34, 2008.

[6] I. Arends, D. J. Bruinvels, D. S. Rebergen et al., “Interventionsto facilitate return towork in adults with adjustment disorders.,”TheCochrane Database of Systematic Reviews, vol. 12, Article IDCD006389, 2012.

[7] R. W. B. Blonk, V. Brenninkmeijer, S. E. Lagerveld, and I. L.D. Houtman, “Return to work: a comparison of two cognitivebehavioural interventions in cases of work-related psycholog-ical complaints among the self-employed,” Work & Stress, vol.20, no. 2, pp. 129–144, 2006.

[8] K. M. Richardson and H. R. Rothstein, “Effects of occupationalstress management intervention programs: a meta-analysis,”Journal of Occupational Health Psychology, vol. 13, no. 1, pp. 69–93, 2008.

[9] J. J. L. van der Klink, R. W. B. Blonk, A. H. Schene, and F. J. H.van Dijk, “The benefits of interventions for work-related stress,”American Journal of Public Health, vol. 91, no. 2, pp. 270–276,2001.

[10] T. Lund, M. Labriola, K. B. Christensen, U. Bultmann, andE. Villadsen, “Return to work among sickness-absent Danishemployees: prospective results from the Danish Work Envi-ronment Cohort Study/National Register on Social TransferPayments,” International Journal of Rehabilitation Research, vol.29, no. 3, pp. 229–235, 2006.

[11] C. D. G. Stoltenberg and P. G. Skov, “Determinants of return towork after long-term sickness absence in six Danish municipal-ities,” Scandinavian Journal of Public Health, vol. 38, no. 3, pp.299–308, 2010.

[12] E. P. M. Brouwers, B. Terluin, B. G. Tiemens, and P. F.M. Verhaak, “Predicting return to work in employees sick-listed due to minor mental disorders,” Journal of OccupationalRehabilitation, vol. 19, no. 4, pp. 323–332, 2009.

[13] U. Bultmann, R. Rugulies, T. Lund, K. B. Christensen, M.Labriola, and H. Burr, “Depressive symptoms and the riskof long-term sickness absence: a prospective study among4747 employees in Denmark,” Social Psychiatry and PsychiatricEpidemiology, vol. 41, no. 11, pp. 875–880, 2006.

[14] L.-G. Engstrom and S. Janson, “Stress-related sickness absenceand return to labour market in Sweden,”Disability and Rehabil-itation, vol. 29, no. 5, pp. 411–416, 2007.

[15] B. Netterstrøm, L. Friebel, and Y. Ladegaard, “Effects of a mul-tidisciplinary stress treatment programme on patient return towork rate and symptom reduction: results from a randomised,

wait-list controlled trial,” Psychotherapy and Psychosomatics,vol. 82, no. 3, pp. 177–186, 2013.

[16] B. Netterstrøm, L. Friebel, and Y. Ladegaard, “The effects of agroup based stress treatment program (the Kalmia concept) tar-geting stress reduction and return to work. A randomized, wait-list controlled trial,” Journal of Environmental and OccupationalScience, vol. 1, pp. 111–120, 2012.

[17] T. S. Kristensen, H. Hannerz, A. Høgh, and V. Borg, “TheCopenhagen psychosocial questionnaire—a tool for the assess-ment and improvement of the psychosocial work environment,”Scandinavian Journal of Work, Environment & Health, vol. 31,no. 6, pp. 438–449, 2005.

[18] U. Lundberg, “Stress hormones in health and illness: the rolesof work and gender,” Psychoneuroendocrinology, vol. 30, no. 10,pp. 1017–1021, 2005.

[19] L. F. Portela, L. Rotenberg, A. L. Pereira Almeida, P. Lands-bergis, and R. H. Griep, “The influence of domestic overloadon the association between job strain and ambulatory bloodpressure among female nursing workers,” International Journalof Environmental Research and Public Health, vol. 10, no. 12, pp.6397–6408, 2013.

[20] A.-L. Elo, A. Leppanen, and A. Jahkola, “Validity of a single-item measure of stress symptoms,” Scandinavian Journal ofWork, Environment & Health, vol. 29, no. 6, pp. 444–451, 2003.

[21] L. R.Olsen, E. L.Mortensen, andP. Bech, “TheSCL-90 and SCL-90R versions validated by item response models in a Danishcommunity sample,”Acta Psychiatrica Scandinavica, vol. 110, no.3, pp. 225–229, 2004.

[22] M. Labriola and T. Lund, “Self-reported sickness absence as arisk marker of future disability pension. Prospective findingsfrom the DWECS/DREAM study 1990–2004,” InternationalJournal of Medical Sciences, vol. 4, no. 3, pp. 153–158, 2007.

[23] P. U. Hjarsbech, K. B. Christensen, R. V. Andersen, V. Borg,B. Aust, and R. Rugulies, “Do psychosocial working conditionsmodify the effect of depressive symptoms on long-term sicknessabsence?” American Journal of Industrial Medicine, vol. 56, no.11, pp. 1329–1340, 2013.

Research ArticleDo Italian Companies Manage Work-Related Stress Effectively?A Process Evaluation in Implementing the INAIL Methodology

Cristina Di Tecco, Matteo Ronchetti, Monica Ghelli, Simone Russo,Benedetta Persechino, and Sergio Iavicoli

Department of Occupational and Environmental Medicine, Epidemiology and Hygiene,Italian Workers’ Compensation Authority (INAIL) Research Area, Via Fontana Candida 1,Monte Porzio Catone, 00040 Rome, Italy

Correspondence should be addressed to Cristina Di Tecco; [email protected]

Received 16 January 2015; Revised 4 August 2015; Accepted 6 September 2015

Academic Editor: Yvonne F. Heerkens

Copyright © 2015 Cristina Di Tecco et al. This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.

Studies on Intervention Process Evaluation are attracting growing attention in the literature on interventions linked to stress and thewellbeing of workers.There is evidence that some elements relating to the process and content of an interventionmay have a decisiverole in implementing it by facilitating or hindering the effectiveness of the results. This study aimed to provide a process evaluationon interventions to assess and manage risks related to work-related stress using a methodological path offered by INAIL. The finalsample is composed of 124 companies participating to an interview on aspects relating to each phase of the INAIL methodologicalpath put in place to implement the intervention. INAIL methodology has been defined as useful in the process of assessing andmanaging the risks related to work-related stress. Some factors related to the process (e.g., implementation of a preliminary phase,workers’ involvement, and use of external consultants) showed a role in significant differences that emerged in the levels of risk,particularly in relation to findings from the preliminary assessment.Main findings provide information on the key aspects of processand content that are useful in implementing an intervention for assessing and managing risks related to work-related stress.

1. Introduction

Psychosocial risks are widely recognised as emerging risksto the health and safety of workers and are linked toworkplace problems such as work-related stress, harassmentor bullying, and workplace violence [1]. They are one of themost challenging issues to be faced, not only because oftheir widespread increase in Europe, but also in considerationof the significant related socioeconomic costs not only forcompanies but for society as a whole [2].

The latest pan-European opinion poll on OccupationalSafety and Health (OSH) conducted by the European Agencyfor Safety andHealth atWork [3] reported that 51%ofworkersconsider work-related stress common in their workplace.Furthermore, four out of ten workers sustained that stresswas not managed adequately in their organisation. A recentsurvey on health and safety at work carried out by INAIL onover 8000 Italian workers indicated that workers generallyfeel more exposed to work-related stress than to any other

risk in the workplace [4]. Since the 1970s, studies have beendeveloped to investigate psychosocial risks and their impactsand to provide practical solutions at the organisational andpolicy levels to manage them [5].

On the European level, efforts have been made to guidecompanies in assessing and handling these risks, includingtechniques for developing strategies and tools for managingthem.The European Framework Agreement (2004) providesemployers and employees with a reference framework foridentifying, preventing, andmanaging work-related stress onthe organisational level. This has since been accompanied bya series of methodological proposals from different Europeancountries to offer companies solutions that are both effectiveand sustainable in managing psychosocial risks at work.

In Italy, the inclusion in the specific OSH legislation (Leg-islative Decree 81/08 and amendments) of the World HealthOrganization definition of health as a “state of completephysical, mental, and social wellbeing and not merely theabsence of disease or infirmity” [6] served as the basis for


http://dx.doi.org/10.1155/2015/197156


protection against psychosocial risks at work, particularlythose related to work-related stress [7]. This LegislativeDecree, implementing the European Agreement on work-related stress, established the employer’s responsibility forassessing and managing risks related to work-related stress,with the collaboration of company OSH professionals. Thishas led to the need for scientific strategies and effectivetools to enable the company to fulfill these requirementsusing sustainable risk management strategies, in compliancewith the guidelines provided by the Permanent ConsultativeCommission for Occupational Health and Safety [8].

Drawing on its national and international research onthe issue, in 2011, the INAIL Research Area developed amethodological proposal for assessing and managing risksrelated to work-related stress [9]. The proposal reflects theminimum requirements and the methodological criteriaidentified in regulatory terms [10] and it is based on the riskmanagement framework. It therefore starts by identifying andestimating/measuring the risk and identifies what should beconsidered the key resources, strategies, and measures forcorrecting, controlling, and preventing it, using a participa-tory process [10–12]. It offers a dynamic path, made up offour key phases and based on a continuous improvementcycle [13], which should involve company OSH professionalsand the active participation of workers right from the initialplanning stages.

Offering tools that are scientifically proven and easy touse [8, 14], the INAIL approach has now been employed bya large number of Italian companies (more than 6000) in thepublic and private sectors, in many different fields of business(health, services, education, construction, etc.).

2. The Four Phases of the Methodological Path

The INAIL methodology represents an intervention forassessing and managing the risks related to work-relatedstress in four phases, each with its own specific objectives,activities, and supporting tools (Table 1). The first, prelim-inary phase, outlines activities for planning and managingthe entire process of risk assessment, which is essential forensuring the accuracy and effectiveness of the subsequentphases. In addition to establishing a steering group respon-sible for planning the assessment process, during this phase,it is important to ensure workers are involved, using notonly communication strategies but also training, if necessary,for those concerned. Lastly, during this phase, homogeneousgroups of workers are identified (see [8] for a full definitionof homogeneous groups) onwhich to implement themethod-ological path.

The second phase is the actual preliminary assessment,analysing the outcome indicators, consisting of sentinelevents, work content factors, and work context factors linkedto the work-related stress [15]; these objective and/or veri-fiable indicators are gathered using a checklist compiled byeach homogeneous worker group. The third phase, in-depthassessment, comprises a detailed analysis of work content andcontext factors, from the workers’ point of view. The Italianversion of the HSE (Health and Safety Executive) IndicatorTool is available for this phase [16], but additional tools or

ones thatmay bemore suitable (focus groups, semistructuredinterviews, and meetings) can also be used depending on thecharacteristics of the company making the assessment (e.g.,small enterprises, specific economic sectors).

The fourth and final phase involves managing the condi-tions of risk that have come to light in the previous phases,developing corrective or preventive actions, and verifyingtheir effectiveness, based on the outcomes of risk assessment.The final phase also aims to develop a risk monitoring plan,which will allow for a new cycle of intervention two or threeyears from the conclusion of the previous one, as required bythe cyclical, dynamic nature of the path offered [11, 13].

3. This Study

Despite the fast-growing numbers of methodologicalapproaches to assess and manage this kind of riskand to confirm the effectiveness of their outcomes inmethodological terms [8, 13], only very few studies have fullyanalysed the process implemented by companies to assessand manage risks related to work-related stress, aimed atunderstanding how the intervention has been implementedand its main impacts on the effectiveness of the results [16].

Some UK studies in this direction have explored progressin implementing the Management Standards approach,developed and offered to companies by the Health andSafety Executive [17, 18]. As defined by Mellor and colleagues(2011) “the detailed processes through which a programmehas unfolded can explain its success or failure” (page 1041).It follows that investigating the ways of implementing anintegrated approach for assessing and managing risks relatedto work-related stress can be useful for verifying how appro-priate the actual process has been, also in view of the ultimategoal of adequately detecting the risk being investigated andavoiding false conclusions and inconsistent results [16, 19].So detailed analysis is needed to identify those elements ofthe process that are needed to ensure its effectiveness and, iflacking, put its validity at risk.

This analysis refers to studies on Intervention ProcessEvaluation, a topic that is attracting growing attention in theliterature on interventions linked to stress and the wellbeingof workers [16, 20–22]. Previous studies indicate some keyfactors related to the process and content of an interventionto manage stress in the workplace, making for its effectiveimplementation.These factors are significant for assessing theactual measures taken but also for assessing the processesemployed to put the measures into place [19, 22]. Somealso refer to guiding principles and typical aspects of therisk management approach that should be considered whenimplementing methods based on these strategies [13, 23]. Abrief overview of the main ones follows.

Any intervention stands or falls on the ability to planand conduct the process from the perspective of projectmanagement [18, 24]. As mentioned previously, the INAILmethodology requires the establishment of an assessmentmanagement group, or steering group, responsible for plan-ning and managing the assessment, and including a nom-inated manager as well as OSH professionals, including


Table 1: The four phases of the methodological path.

Phase Aim Activities and tools

Preliminary phase To prepare the organisation forsubsequent assessment processes

(i) Establishment of a steering group to manageassessment (employer, managers, OSH professionals,workers, and other health-related organizational figuressuch as the human resources manager, internaloccupational psychologists),(ii) development of a communications/employeeengagement strategy (meetings, training, etc.),(iii) drafting the risk assessment plan (scheduling eachphase, actions, and players involved),(iv) identification of homogeneous groups of workerson which the assessment is to be made.

Preliminary assessment

To assess objective and verifiableindicators associated with work-relatedstress under three main headings: (1)sentinel events (e.g., injury rates, absencedue to sickness, and turnover), (2) workcontent factors (e.g., work load, workinghours, and working environment), (3)work context factors (e.g., interpersonalrelationships work/home interface)

A checklist is compiled for each homogeneous group ofworkers, with their participation.

In-depth assessment To assess employees’ perceptions aboutwork content/context factors

Italian version of the Management Standards IndicatorToolmade up of 35 items corresponding to the sevenManagement Standards: Demands, Control, ManagerialSupport, Peer Support, Relationships, Role, andChange.

Interventions and monitoring

To manage work-related stress byidentifying corrective measures andinterventions based initially on thefindings from the preliminaryassessment. To outline a monitoring plan

A focus group guide to help organisations set up focusgroups to collect detailed information for interpretingthe results of the previous steps and identifying the bestsolutions.

workers’ representatives for safety. This group has key func-tions in correctly implementing the process (Table 1),many ofwhich are linked to the success of the intervention and havealso been defined as crucial in the guidelines drawn up bynational supervisory bodies [25].

The success of an intervention depends not only on howthe process is managed upstream, but also on managementsupport during the intervention (e.g., information and clearcommunication with staff) and the active participation ofmanagers [18, 24]. Several studies have also indicated thatthe active involvement of workers [21, 26] plays a key rolein the success of interventions. In any methodological pathfor assessing and managing work-related stress, workers’participation contributes to the correct estimation of risk, asthey are an essential source of information about their ownworking conditions.Their involvement also aims to boost thelevels of knowledge and internal skills with a view of creatinga cycle of continuous improvement [10].

Another key factor is the level of specific skills in assessingthe risk and stress of the people managing the process [27].If these are inadequate, specific training must be laid on forthose involved.

In some cases, companies may use external consultantswhen implementing an intervention [18]. However, one basic

criterion for developing the INAIL method is that companiesshould be able to implement the assessment andmanagementprocess autonomously. Naturally, however, companies candecide to use an external consultant if they deem it essential.

Another enabling factor in the approach to assess andmanage the risks related to work-related stress is the pos-sibility of combining tailored and contextualized tools withstandard ones to assess the needs linked to the specificsof each organisation [18]. The INAIL methodology wasdeveloped with modularity and flexibility in mind, allowingfor the use of supporting tools to achieve comprehensive riskassessment in compliance with the each company’s specificfeatures (e.g., size, business sector).

In line with evidence from previous studies mentionedabove, the aim of this study was to assess the implementationof the INAIL methodology in a large sample of Italian com-panies. Thus, first of all, we explored the presence of factorsreported in the literature as related to the process evaluationof an intervention tomanage stress in the workplace (namely,the company’s ability to plan and manage the process, train-ing, OSH professionals and workers’ involvement, possibilityof combining tailored and contextualized tools, etc.). Then,we analysed the relationship between these factors and thefindings from the two assessment phases to understand their


Table 2: Main aspects investigated in implementing the process for assessing and managing risks related to work-related stress.

Preliminary phaseInformation and communication strategyParticipation/involvement of workersCompany and external figures involved

Specific training

Preliminary assessmentCompany and external figures involved and ways of involvement

Specific trainingParticipation/involvement of workers

In-depth assessment Reasons that led to this phase being conducted Use of additional tools

Interventions and monitoringAdoption of interventions/corrective measures and typesAssessment of the effectiveness of management measures

Monitoring plan

impact on changes in risk levels. Finally, we explored theperceptions of the usefulness of the whole methodologicalpath and its assessment phases.

Studies that verified the validity and soundness of thesupporting tools [8, 14, 28] led to this follow-up analysis ofthe processes put into place by companies for implementingthe methodological path proposed by INAIL.


The sample of companies involved in this study was extractedfrom the INAIL web platform database [8]. Two maincriteria were followed in selecting them: (1) companies thathad already completed the INAIL methodological path andhad therefore used both the checklist and the IndicatorTool for the assessment phases; (2) companies where thehomogeneous groups comprised more than six workers, formethodological reasons relating to the use of the IndicatorTool.

The resulting sample consisted of 339 companies that hademployed a work-related stress assessment and managementintervention using INAIL methodology. We sent these com-panies a letter describing the investigation and its purpose,contents, how it would be conducted, and how data wouldbe handled. The letter asked them if they would complete aquestionnaire presented during a telephone interview withan occupational psychologist. Of the 339 companies, 124agreed to the interviews (37% response rate); this gave 330homogeneous groups of workers, meaning 330 checklists and4500 questionnaires.

Most of the companies had up to 50 employees (22%from 1–9 and 40% from 10–50), 22% from 51–250 and 16%more than 251. The five most frequent business sectors wereservices (21.8%), manufacturing (17.7%), professional sector,healthcare and social welfare (16.9%), scientific and technicalones (12.1%), and construction (5.6%).

INAIL occupational psychologists conducted the phoneinterviews with an internal representative of the companiesin the work-related stress assessment steering group. Thequestionnaire comprised 22 items to analyse aspects relatingto each phase of the method. Some of the questions inves-tigated qualitative aspects of the assessment and related tothe perceived level of usefulness of these aspects (using aLikert type scale from 1 = completely useless to 5 = completely

useful) and of the method as a whole and any difficulties thathad been met. Other questions were designed to investigatehow the single phases of the assessment process were carriedout and are in fact the most significant items for achievingthe objectives of the follow-up analysis. Table 2 shows someexamples relating to the various aspects.

In addition to the data collected during the interviews,the results of the two assessment phases were extractedfrom the web platform; these were the checklist findingsand the Indicator Tool, for each homogeneous group in thecompanies involved. They are described below.

4.1. PreliminaryAssessment Results. Achecklist was compiledfor each homogeneous group assessed [8] to gather a range ofindicators (sentinel events, work content, and context factors)on a dichotomous scale. Each statement in the checklistcontributes to an overall score. The sum of the scores forthe three areas establishes the position of the homogeneousgroup based on a table of levels of risk: low, medium, andhigh.

4.2. In-Depth Assessment Results. The self-report question-naire used to obtain details is the Italian version of theUK HSE Management Standards Indicator Tool [14, 28]. Itcomprises 35 items measuring the seven dimensions (Man-agement Standards) used for describing the indicators ofwork context and content, corresponding to seven ideal con-ditions/states to be achieved for the prevention and reductionof the risks related to work-related stress in companies. Theoutput is a profile of the levels of risk of each homogeneousgroup for each of the seven dimensions of the questionnaire.

5. Analysis

Data were analysed using the IBM SPSS Statistics version21. Percentage frequencies were calculated for the multiple-choice questions based on the total number of answers, andcomparisons with other questions in the follow-up ques-tionnaire were made by processing the double-entry tables.For questions without multiple-choice responses, parametrictests such as the 𝑡-test and ANOVA were used to verifyrelations between the variables compared. Nonparametrictests were also used such as Chi-square (𝜒2) and the Kruskal-Wallis test, a one-way analysis of variance by ranks which is


Table 3: Significant differences in scores from the preliminary assessment for responders that had implemented the preliminary phase.

M SD 𝑝 Shapiro 𝑡-test Kruskal-Wallis 𝜒2

Yes 1.32 0.506 0.000𝐹 = 34.370 (0.000) 0.049 3.92 (0.140)

No 1.00 0.000 0.380

considered the nonparametric equivalent to the ANOVA andestablishes whether the difference(s) between the mediansfor one or more subsamples are due to chance or arestatistically significant.The samples and the subsamples wereverified for normality of distribution using the 𝑝-Shapiro-Wilk test, so as to establish how to consider the parametricand nonparametric statistics each time. We took 𝑝 < 0.05 assignificant.

6. Results

In keeping with the aims of the study, the results for somekey factors that have emerged from the literature as relatedto successfully assessing and managing risks related to work-related stress are presented below, including the relationshipbetween such factors and findings from the two assessmentphases. For easier reading, they are set out following thephases of the INAIL methodological path (Table 2).

6.1. The Preliminary Phase. Of the 124 companies inter-viewed, 97.4% (115) confirmed they had completed the pre-liminary phase. Although only nine stated they had notcompleted it (2.6%), we compared the levels of risk obtainedin the preliminary assessment phase with those of the respon-ders that had completed the preliminary phase, to checkfor significant differences (Table 3). All the tests indicated asignificant difference between the two groups; in particular,the preliminary assessment appeared more positive, that is,tending towards low risk, in the companies that had notcompleted the preliminary phase.

Table 4 shows the main indicators investigated in theinterviews in the preliminary phase of the INAIL method-ological path.There was a high level of workers’ participation.The companies chose to involve a representative sample ofworkers in 32.2% or all workers in 39.3%; in 27.4%, onlythe workers’ representative for safety was involved. Nearlythree quarters of the companies (74%) interviewed providedspecific training for those involved in interventions. Themajority of respondents considered this extremely useful indeveloping the risk assessment andmanagement process, andonly 3.2% rated it as of little or no use.

6.2. Preliminary Assessment Phase. As part of the preliminaryassessment, we investigated indicators of the involvement ofOSH professionals in the planning phase and in completingthe checklist and checked for any difficulties encounteredin completing the checklist (Table 5). Personnel involvedincluded those responsible for health and safety managementand then the employer, in keeping with the approach takenwhen assessing other risks in the company. Workers (60% ofrespondents) and/or their safety representative (68%) werefrequently involved. In particular, workers were involved in

the briefing for communicating the measures taken by thecompany, but also gathering, analysing, and discussing thedata from the checklist.

Just over a third of companies (35%) stated they didnot involve their workers or workers’ safety representatives;22% engaged an external consultant for implementing thisphase. In keeping with the objectives of the study, we madedetailed analyses to check for significant differences in thefindings of the preliminary assessment phase linked to theparticipation of workers or their safety representatives, andto the involvement of an external consultant in the process(Table 6).

There was a significant tendency towards higher levelsof risk in companies that involved workers and/or theirrepresentatives. In contrast, assessments with lower levels ofrisk tended to come from companies that engaged an externalconsultant for implementing the phase.

Although the majority of companies stated they hadno difficulty in completing the checklist (57%), those thatdid encounter some problem referred in particular to itsapplicability to their business context for all three families ofstress indicators.

6.3. In-Depth Assessment Phase. Among the reasons toimplement the in-depth assessment phase, around 43% ofthe companies wanted to analyze workers’ perceptions ofrisks related to work-related stress, 33.8 % wanted to obtaindetails of the preliminary assessment findings to define riskmore clearly, and 20.5% wanted to better identify the cor-rective measures to be put in place. Only 1.5% of companiesimplemented this phase as a result of the ineffectiveness ofcorrective measures taken following the preliminary assess-ment, a process required in order to comply with regulatoryrequirements. However, detailed analysis did not bring tolight any direct link between the results of the preliminaryassessment and the reasons that prompted companies toimplement in-depth assessment.

In 56%, the use of further tools in addition to the IndicatorTool offered in the INAIL method was confirmed. In 25.3%,focus groups were formed for samples of employees, in18.7% for detailed meetings and in 12.0% for semistructuredinterviews.

Lastly, the interviews showed a high level of appreciation(M 3.48, SD 0.854) regarding the comprehensiveness of theresults from the in-depth phase for clearly defining the risksrelated to work-related stress.

6.4. CorrectiveMeasures andMonitoring. Developing correc-tive measures and actions is a key step in this methodologicalprocess. The conditions of risk that emerge from the previ-ous phases must be managed by defining and implement-ing corrective or preventive measures and verifying their


Table 4: Descriptive indicators related to the preliminary phase ofthe INAIL’s methodology to assess and manage the risks related towork-related stress (number of companies interviewed = 124).

Preliminary phaseImplemented

Yes 97.4%No 2.6%

100.0%Perceived usefulness

Mean (ranking 1–5) 3.57Standard deviation 0.785

InvolvementWorkers’ involvement

Yes 85.7%No 14.3%

100.0%Target of the strategy

Workers’ representatives for safety 27.4%Trade union representatives 1.0%All workers 39.3%A sample of workers 32.3%

100.0%Way of involvement

Meetings 65.3%E-mail 7.8%Intranet alerts 5.4%Posts on the bulletin board 5.8%Brochure 15.7%

100.0%Training

Specific trainingYes 73.8%No 26.2%

100%Type of training

Traditional course 85.9%E-learning 14.1%

100.0%Perceived usefulness of training


Identification of homogeneousgroups of workers

Use of ad hoc toolsYes 74.4%No 25.6%

100.0%Perceived usefulness of ad hoc tools


effectiveness, on the basis of risk assessment. The majority ofrespondents had adopted or were currently adopting (51.6%

and 20.7%, resp.) corrective action or measures to prevent,reduce, or eliminate conditions of psychosocial risks. Table 7illustrates the action taken by the respondent companies,based on the types classified in the European FrameworkAgreement on work-related stress (2004).

As regards the companies’ perceptions of the usefulness ofthe INAILmethodology, most of them reported they found ituseful in assessing and managing the risks related to work-related stress (M 3.62, SD 0.771). Companies also reportedpositive perceptions of the usefulness for the preliminaryphase (M 3.57, SD 0.785) and the use of ad hoc tools foridentifying the homogeneous groups (M 3.18, SD 0.724).Lastly, the interviews indicated a high level of appreciationof the exhaustiveness of the results gathered in the in-depthphase for clearly defining the risks related to work-relatedstress (M 3.48, SD 0.854).

7. Discussion

This study makes a process evaluation of interventions forthe assessment and management of risks related to work-related stress, using a methodological path offered by INAILinvolving the investigation of (1) factors that contribute toits effective implementation, (2) the impact of these factorson changes in the level of risk, and (3) perceptions of theusefulness of the methodological path and the assessmentphases.

In the follow-up interviews on a sample of companies thatfollowed the entire methodological path, several recurringfactors help explain the differences arising during the assess-ment phases.

In terms of the process, significant differences emerged inthe levels of risk resulting from the preliminary assessmentphase in companies that completed the preliminary phase,which was the majority of the sample. There was a tendencytowards higher levels of risk compared to those that didnot complete this phase. Although it is clear that this resultmust be interpreted with caution, given the small number ofcompanies not completing the preliminary phase, the abilityof a company to plan and manage the process from theperspective of projectmanagement is recognized as one of thefeatures that is often associated with the success of interven-tions, in the context of work-related stress and wellbeing inthe workplace [24].Themethodological path offered stronglyrecommends setting up a steering group for planning andmanaging the preparatory work for actual implementation ofthe intervention as well as for the involvement of the OSHprofessionals.

It can be assumed that the differences in the tendencytowards risk depend partly on how accurately the assessmentis conducted from the initial stages. Crucial steps in theprocess, such as identifying homogeneous groups of workers,verifying internal skills, the involvement and participation ofworkers, and their safety representatives, are completed dur-ing the preliminary phase. Therefore, thorough preparationof the organisation for risk assessment and a participatoryapproach probably make it easier to make the best use of thetools and fully recognise any issues related to risk factors.


Table 5: Descriptive indicators related to the preliminary assessment phase of the INAIL’s methodology to assess andmanage the risks relatedto work-related stress (number of companies interviewed = 124).

Preliminary assessment phaseInvolvement

Personnel involvedEmployer 14.6%Manager as employer’s delegate 8.2%Personnel assigned 7.8%Health and safety manager 18.9%Workers’ safety representatives 15.7%Health and safety workers assigned 4.3%Company physician 11.6%Workers 13.8%External consultant 5.0%Total 100.0%

Workers’ involvementIn the information meetings 34.1%In planning the assessment 15.2%In the collection, analysis, and discussion of data from the checklist 33.8%In identifying corrective measures 16.9%

Personnel completing the checklistEmployer 15.2%Manager as employer’s delegate 11.1%Health and safety manager 23.8%Company physician 10.7%Health and safety workers assigned 5.3%Workers’ safety representatives 18.1%Workers 15.8%Total 100.0%

Problems in completing the checklistYes 42.6%No 57.4%

100.0%Concerns that emerged

Type of concern in completing the checklistSentinel event 37.7%Content of statements not clear 2.8%Concerns about the application to different business contexts 37.3%Concerns about data availability 22.2%

100.0%Work content factors 38.6%Content of statements not clear 12.0%Concerns about the application to different business contexts 36.9%Concerns about data availability 12.4%

100.0%Work context factors 36.1%Content of statements not clear 10.2%Concerns about the application to different business contexts 40.3%Concerns about data availability 13.4%

100.0%


Table 6: Comparison of the preliminary assessment for companies that involved workers or their safety representative and for those thatengaged external consultants.

M SD 𝜒2 Shapiro-Wilk 𝑝 ANOVA 𝑡-est Kruskal-Wallis

Workers involvedYes 1.34 0.513 3.941 (0.139) 0.000 0.047 0.014 0.05No 1.16 0.374 0.000

Consultants engagedYes 1.22 0.450 4.101 (0.129) 0.000 0.053 0.038 0.04No 1.35 0.517 0.000

Table 7: Descriptive indicators relating to the corrective measures and monitoring phase of the INAIL’s methodology to assess and managethe risks related to work-related stress (number of companies interviewed = 124).

Corrective measures and monitoringImplementation of interventions

Yes 51.6%No 20.7%In implementation 20.7%

After which phase the interventions were implementedPreliminary assessment 9.8%In-depth assessment 20.7%Both of these 69.5%

100.0%Type of measure

Organizational 20.3%Communication 19.3%Training 19.5%Procedural 24.5%Technical 16.4%

Time from the last assessmentFrom 1 to 6 months 19.3%From 6 to 12 months 42.9%Over 12 months 37.9%

Monitoring planYes 61.4%No 38.6%

Method for implementing the monitoring planPeriodic monitoring of sentinel events 21.9%Periodic monitoring of workers’ perceptions 41.8%Other ways for assessing the effectiveness of the measures 36.3%

Internal skills and expertise are another primary aspectof the process, acknowledged in the literature [27]. Themajority of companies that completed the preliminary phaseconsidered it necessary to provide specific training on thisissue, probably to compensate for a lack of internal expertise.In most cases, this was perceived as extremely useful forcompleting the process of risk assessment and management.In some cases, this lack was compensated by engaging anexternal consultant, although the INAIL methodology wasdeveloped with a view to ensuring that companies coulduse it autonomously. There appeared to be a tendencytowards assessments with lower risk during the preliminaryassessment phase in the companies that engaged external

consultants. Therefore, in future, it would be interesting toanalyse the impact of external professionals on the assessmentprocess and its outcomes; for example, the types of pro-fessionals involved (psychologists, physicians, employmentconsultants, etc.) and the support methods offered in thesteering group could be investigated.

As noted in previous studies [21, 26] and in the indicationsof the Permanent Consultative Commission regarding theminimum legal requirements for assessing andmanaging therisks related to work-related stress, the direct participation ofworkers and/or their representatives is considered decisive inimplementing this type of intervention.The companies inter-viewed attributed substantial importance to the participation


of employees and generally included as many as possibleright from the early phases of the process, using a strategyof communication and updating. The study findings confirmits importance, indicating a tendency to lower levels of riskin cases where this participation was lacking, particularlyin essential steps such as completing the checklist. In thiscase, it can also be assumed that the greater the participationof workers in the preliminary assessment phase, the moreaccurate the information on working conditions. In any case,the majority of companies appear to recognise the role ofemployees in assessing this risk, as they are a valuable sourceof information about work context and work content.

The findings of the in-depth assessment confirmed this,regardless of the results from the previous phase, with themain aim of analysing workers’ perceptions related to organ-isational risk factors. Most of companies reported low risksrelated to work-related stress in the preliminary assessmentfindings. This already indicates a willingness to establish thepresence of risk in the most comprehensive manner possiblein order to plan any necessary targeted and preventivemeasures. In this regard, the majority of companies hadadopted or were adopting corrective measures to prevent orreduce work-related stress risks, as well as a monitoring plan.

The tools offered appear to provide full information andindications for identifying the different types of intervention,in compliance with the requirements of the European Frame-work Agreement. However, a limiting factor in this study isthe lack of data, in this first phase, on the actualmethods usedto manage risk, as well as verification of the effectiveness ofcorrective measures because of the impossibility of followingthe process step by step in an observational way. This lack ofdata also reflects a scarcity of specific information about thecontext in which each company developed the interventionfor assessing and managing work-related stress risks, accord-ing to the literature on process intervention evaluation [29,30]. To address this, case studies are nowunderway in specificcontexts (e.g., social and health care, public administration,and small and medium enterprises) designed to analyse indetail the applicability of the methodological path, using thesupport provided to the steering group throughout all thesteps.

The main issue related to content concerns the appli-cability of the checklist to different business contexts. Theproposed checklist was developed as a tool that could be usedacross a range of different contexts to identify risk factors.Ongoing studies have, in actual fact, brought to light signif-icant differences between the levels of risk emerging fromthe use of the checklist and the Indicator Tool [31]. Duringthe in-depth assessment phase, additional tools can still beused alongside the Indicator Tool, such as focus groups,which are useful for obtaining more detailed information,as the results show. The majority of the sample consideredthe tools offered during this phase exhaustive enough toclearly define risk. This confirms the need for a bottom-uptype of methodological approach [18] where workers mustbe involved to successfully identify risk and the relativemeasures for improvement.

To further develop themethodology, we are now runningstudies on the contextualisation and adaptability to specific

occupational features and needs. As part of a project fundedby the Ministry of Health, additional tools will be developed,specifically tailored on the basis of companies’ characteristicsto take account of their sector and size.



Acknowledgments

The authors are grateful to Tiziana Castaldi and Sara Vitalifor their help in contacting companies and collecting datathrough the telephone interviews and to Antonia Ballottin inparticular for her collaboration in developing the interviewquestionnaire. Finally, the authors wish to thank StavroulaLeka for her valuable suggestions.

References

[1] EU-OSHA, ESENER—European Survey of Enterprises on Newand Emerging Risks, European Agency for Safety and Health atWork, Bilbao, Spain, 2010.

[2] EU-OSHA, Calculating the Costs of Work-Related Stress andPsychosocial Risks, A Literature Review, European Agency forSafety and Health at Work, Bilbao, Spain, 2014.

[3] EU-OSHA, European Opinion Poll on Occupational Safety andHealth, EuropeanAgency for Safety andHealth atWork, Bilbao,Spain, 2013.

[4] INAIL, Indagine Nazionale sulla Salute e Sicurezza sul Lavoro,INAIL, Milan, Itlay, 2014.

[5] WHO, Health Impact of Psychosocial Hazards at Work: AnOverview/Stavroula Leka, Aditya Jain, World Health Organiza-tion, Geneva, Switzerland, 2010.

[6] WHO, “Preamble to the constitution of the World HealthOrganization as adopted by the International Health Confer-ence, New York, 19–22 June, 1946; signed on 22 July 1946 bythe representatives of 61 states,” Official Records of the WorldHealth Organization 2, 1948.

[7] P. Deitinger, C. Nardella, R. Bentivenga, M. Ghelli, B. Per-sechino, and S. Iavicoli, “D.Lgs. 81/2008: conferme e novita intema di stress correlato al lavoro,” Giornale Italiano di Medicinadel Lavoro ed Ergonomia, vol. 31, no. 2, pp. 154–162, 2009.

[8] B. Persechino, A. Valenti, M. Ronchetti et al., “Work-relatedstress risk assessment in Italy: a methodological proposaladapted to regulatory guidelines,” Safety and Health at Work,vol. 4, no. 2, pp. 95–99, 2013.

[9] INAIL,Managing and Assessing the Risk forWork-Related Stress.Guide for Companies, in Compliance with Leg. Decree 81/2008and Subsequent Integrations and Modification, Research Area,2013.

[10] T. Cox, A. Griffiths, C. Barlow, R. Randall, L. Thomson, and E.Rial-Gonzalez, Organisational Interventions for Work Stress: ARisk Management Approach, HSE Books, Sudbury, UK, 2000.

[11] M. K. Langenhan, S. Leka, and A. Jain, “Psychosocial risks:is risk management strategic enough in business and policymaking?” Safety and Health at Work, vol. 4, no. 2, pp. 87–94,2013.


[12] S. Leka and T. Cox, PRIMA-EF, Guidance on the EuropeanFramework for Psychosocial Risk Management: A Resource forEmployers and Worker Representatives, World Health Organi-zation (WHO), Geneva, Switzerland, 2008.

[13] S. Leka, J. Hassard, A. Jain et al., Towards the Development of aEuropean Framework for Psychosocial Risk Management at theWorkplace, Saltsa, I-WHO, 2008.

[14] B. M. Rondinone, B. Persechino, T. Castaldi et al., “Work-related stress risk assessment in Italy: the validation study of thehealth safety and executive indicator tool,” Giornale Italiano diMedicina del Lavoro ed Ergonomia, vol. 34, no. 4, pp. 392–399,2012.

[15] S. Leka, A. Griffiths, and T. Cox, Work Organisation andStress, Protecting Workers’ Health Series no.3, WHO LibraryCataloguing, 2003.

[16] C. Biron and M. Karanika-Murray, “Process evaluation fororganizational stress and well-being interventions: implicationsfor theory, method, and practice,” International Journal of StressManagement, vol. 21, no. 1, pp. 85–111, 2014.

[17] T. Cox,M. Karanika, A. Griffiths, and J. Houdmont, “Evaluatingorganizational-level work stress interventions: beyond tradi-tional methods,”Work & Stress, vol. 21, no. 4, pp. 348–362, 2007.

[18] N. Mellor, C. Mackay, C. Packham et al., “’Management Stan-dards’ andwork-related stress inGreat Britain: progress on theirimplementation,” Safety Science, vol. 49, no. 7, pp. 1040–1046,2011.

[19] M. Egan, C. Bambra,M. Petticrew, andM.Whitehead, “Review-ing evidence on complex social interventions: appraisingimplementation in systematic reviews of the health effects oforganisational-level workplace interventions,” Journal of Epi-demiology and Community Health, vol. 63, no. 1, pp. 4–11, 2009.

[20] K. Nielsen, H. Fredslund, K. B. Christensen, and K. Albertsen,“Success or failure? Interpreting and understanding the impactof interventions in four similar worksites,”Work and Stress, vol.20, no. 3, pp. 272–287, 2006.

[21] K. Nielsen and R. Randall, “The importance of employee par-ticipation and perceptions of changes in procedures in a team-working intervention,” Work & Stress, vol. 26, no. 2, pp. 91–111,2012.

[22] R. Randall, K.Nielsen, and S.D. Tvedt, “Thedevelopment of fivescales to measure employees’ appraisals of organizational-levelstress management interventions,”Work & Stress, vol. 23, no. 1,pp. 1–23, 2009.

[23] S. Leka and T. Cox, The European Framework for PsychosocialRisk Management: PRIMA-EF, I-WHO Publications, Notting-ham, UK, 2008.

[24] E. Donaldson-Feilder, J. Yarker, and R. Lewis, “Managementcompetencies for preventing and reducing stress at work,”in Occupational Health Psychology: European Perspectives onResearch, Education and Practice, J. Houdmont and S. Leka,Eds., pp. 229–256, Nottingham University Press, Nottingham,UK, 2008.

[25] Coordinamento Tecnico Interregionale della Prevenzione neiLuoghi di Lavoro, Decreto Legislativo 81/2008 s. m. i.. Stresslavoro-correlato indicazioni per la corretta gestione del rischioe per l’attivita di vigilanza alla luce della lettera circolare del 18novembre 2010 delMinistero del Lavoro e delle Politiche Sociali,2012.

[26] K. Nytrø, P. Ø. Saksvik, A. Mikkelsen, P. Bohle, and M.Quinlan, “An appraisal of key factors in the implementation ofoccupational stress interventions,”Work & Stress, vol. 14, no. 3,pp. 213–225, 2000.

[27] J. Yarker, R. Lewis, and E. Donaldson-Feilder, ManagementCompetencies for Preventing and Reducing Stress at Work: Iden-tifying and Developing the Management Behaviours Necessary toImplement the HSE Management Standards, Goldsmiths, Uni-versity of London for the Health and Safety Executive, London,UK, 2007.

[28] A. Marinaccio, P. Ferrante, M. Corfiati et al., “The relevance ofsocio-demographic and occupational variables for the assess-ment of work-related stress risk,” BMC Public Health, vol. 13,no. 1, article 1157, 2013.

[29] C. Biron, M. Karanika-Murray, and C. L. Cooper, ImprovingOrganizational Interventions for Stress and Well-Being. Address-ing Process and Context, Routledge, Taylor & Francis Group,London, UK, 2012.

[30] G. F. Bauer and G. J. Jenny, “From fidelity to figuration: currentand emerging approaches to organizational health interventionresearch,” in Salutogenic Organizations and Change: The Con-cepts Behind Organizational Health Intervention Research, G. F.Bauer and G. J. Jenny, Eds., pp. 1–16, Springer, Dordrecht, TheNetherlands, 2013.

[31] M. Ronchetti, C. Di Tecco, S. Russo et al., “An integratedapproach to the assessment of work-related stress risk: compari-son of findings from two tools in an Italianmethodology,” SafetyScience, vol. 80, pp. 310–316, 2015.

Documents

Psychosocial Factors and Workers’ Health and Safetydownloads.hindawi.com/journals/specialissues/909274.pdfelevated basal systolic blood pressure, reduced basal heart rate variability,