Accident Analysis and Prevention 73 (2014) 187–199

Maintenance: organizational modes, activities and health and safety.Use of a French national survey and in-situ analyses

Corinne Grusenmeyer *French National Institute of Research and Safety, Preventive Ergonomics and Psychology Laboratory, 1 Rue du Morvan, CS 60027, Vandoeuvre les Nancy54519, France

A R T I C L E I N F O

Article history:Received 20 November 2013Received in revised form 2 April 2014Accepted 8 September 2014Available online xxx

Keywords:Maintenance activitiesOrganizational modesOutsourcingHealth and safetyInter-organizational interactionsNetwork of companies

A B S T R A C T

Maintenance activities are identified as critical both to operator safety and to systems safety andreliability. However, it is still difficult to identify maintenance workers in French occupational accidentand disease statistics. Moreover, few analyses of these activities and of organizational changes in this fieldhave been conducted. This paper presents two different approaches to this same issue. Analyses wereaimed firstly at identifying the occupational exposures of these operators and at comparing them withoccupational exposures of production staff and, secondly at developing understanding of normal realmaintenance activities, i.e. maintenance activities that are normally actually carried out, while takinginto account the socio-technical system and maintenance organization within which they lie.The use of the French SUMER 2003 survey shows that occupational exposures of maintenance staff to

various constraints are more frequent than occupational exposures of their production colleagues.However, maintenance staff appear to have greater independence. Analyses were also conducted in asubcontracting urban public transport company, who outsources some maintenance work. Thoseanalyses highlight a complex network of companies involved in maintenance activities, a substantialnumber of work interruptions and a significant fragmentation of the internal technicians’ activities thatcan be cognitively costly, reduce anticipation possibilities and lead to incidents or accidents. Above allthey underline internal technicians’ contributions to the completion of outsourced interventions andinterdependent relationships between the activities of the internal and the external technicians.Outsourcing maintenance interventions thus raises the question of risks associated with the

interdependence of actual work activities undertaken by the different types of staff, since they contributeto the same maintenance intervention. This study therefore pinpoints the need to integrateinter-organizational interactions in order to understand the variability of maintenance activities andits relationships with reliability and safety. In this respect, some suggestions are provided with aprevention aim.

ã 2014 Elsevier Ltd. All rights reserved.

Contents lists available at ScienceDirect

Accident Analysis and Prevention

journal homepage: www.else vie r .com/ locate /aa p

1. Introduction

Maintenance concerns the combination of all technical,administrative and managerial actions during the life cycle of anitem that are intended to retain it and secure it in, or restore it to astate in which it can perform the required function (Rosqvist et al.,2009). The maintenance function is then cross-cutting, concerningall activity sectors and essential to a company's ability to staycompetitive. At the same time, maintenance activities areidentified as critical both to operator safety and to systems safetyand reliability. As regards operator safety, several studies have

* * Tel.: +33 383502166; fax: +33 383502182.E-mail address: corinne.grusenmeyer@inrs.fr (C. Grusenmeyer).

http://dx.doi.org/10.1016/j.aap.2014.09.0090001-4575/ã 2014 Elsevier Ltd. All rights reserved.

emphasized a high level of maintenance-related accidents (Haleet al., 1998; Batson et al., 1999; Ray et al., 2000; Farrington-Darbyet al., 2005; Grusenmeyer, 2005; Lind, 2008). According to theEuropean Agency for Safety and Health at Work (2010a),maintenance was associated with 10–15% of fatal occupationalaccidents and 15–20% of all accidents in Europe in 2006. As regardssystems safety and reliability, maintenance failings may contributeto the occurrence of major accidents, such as the Piper Alphaplatform explosion, the Embraer 120 aircraft crash in Texas, andthe Bhopal or the three mile island accidents (Reason and Hobbs,2003; European Agency for Safety and Health at Work, 2009).Several accident investigations have identified inadequate or faultymaintenance as one of the main contributors to unanticipatedevents in various critical safety domains (Reiman, 2011). Thus,relationships between maintenance and safety are contradictory:

1 Ergonomic analysis refers here to the analysis of operators’ actual activities asinserted in a specific context, in order to improve health and safety conditions atwork.

188 C. Grusenmeyer / Accident Analysis and Prevention 73 (2014) 187–199

maintenance contributes to risk control by preventing andcorrecting non-optimal operating modes of equipment, whichmay cause accidents; at the same time, maintenance activitiesexpose maintenance workers naturally to many risks, not leastbecause they interact directly with equipment or installations.Then, as emphasized by Reiman (2011), “maintenance activities canbe considered as having a highly significant positive or negativeimpact on the effectiveness of the entire socio-technical system,including safety” (p.339).

Different elements may contribute to the fact that maintenanceactivities are critical to operator safety:

� nature of maintenance activities: direct interactions withdangerous equipment, installations or products (Lind, 2008);high uncertainty and variability of maintenance work (Bourrier,1996; Bounot et al., 1996; Garrigou et al., 1998; Grusenmeyer,2002; Reiman, 2011): task variability, difficulties with planningmaintenance intervention times and human resources; highnumber of unplanned events; no workstation and high mobilityof maintenance staff; manual tasks requiring complex reasoningand often informal know-how;

� conditions under which maintenance activities are carried-out:variability of the technological and environmental conditions(De La Garza and Weill-Fassina, 1995); temporal constraints(Vidal-Gomel, 2007); difficult material conditions of mainte-nance interventions depending on the postures and effortrequired, for example (Lind, 2008);

� organization of maintenance within but also beyond thecompany: for example, maintenance tasks allocated to produc-tion operators while the number of maintenance operators isdecreasing; maintenance workers being overloaded with workdue to maintenance shift work teams being removed; greatcoactivity between numerous maintenance operators duringscheduled outages of installations (Bourrier, 1996); but also,outsourcing of maintenance interventions likely to harmcooperation between the different types of staff and conse-quently to be detrimental to safety: unsteadiness and splitting ofwork teams (Perraudin et al., 2006), little mutual knowledge ofthe different operators, various corporate cultures, heterogeneityof employee status (Desriaux, 2005), no sharing of safety rules(Duhamel, 2005), loss of know-how and of overall knowledge ofthe systems by internal maintenance workers (Largier, 2008;Herrera et al., 2009).

However, it is still difficult to know the number of maintenanceworkers in France, to characterize this population, or even toidentify it in occupational accident and disease statistics. Inparticular, there are fewer studies focused on professionalexposures and diseases of these operators. Nevertheless, theFrench national survey, named SUMER, which assessed theexposure of employees to the main workplace risks in France,identified a short time ago “maintenance” as a professional domain(Equipe SUMER, 2006).

Otherwise, as emphasized by Reason and Hobbs (2003),maintenance activities received little attention and few preventionstudies were actually dedicated to these activities (Ray et al.,2000). Moreover, the most relevant studies on maintenance havefocused on human errors and performance. According to Reiman(2011), studies on maintenance were mainly interested in humanerrors and individual-level issues, even though social andorganizational factors have received increasing attention in recentyears. In such research, maintenance is considered as a majorsource of latent errors in generally high-risk socio-technicalsystems and one of the principal causes of major accidents (Reasonand Hobbs, 2003). These studies are focused on post-identificationof the most common types of errors, the factors lying behind such

errors and ways of preventing these errors (Reason and Hobbs,2003; Hobbs and Williamson, 2002, 2003; Hobbs and Kanki, 2008;Suzuki et al., 2008), with a rather individual view of maintenancework. A second type of studies examines maintenance interven-tions as dangerous for humans (Lind, 2008; Lind and Nenonen,2008). Then, most often, maintenance audits are conducted orpost-analyses of accidents having occurred during maintenanceoperations are carried out in order to identify types of accidents, aswell as local and organizational factors contributing to them, andin order to suggest prevention measures (Hale et al., 1998; Mainet al., 2002; Lind, 2008; Lind and Nenonen, 2008; Tazi, 2008). Thusergonomic literature on real maintenance activities and on normalwork, practices and cultures of maintenance is rather scarce(Reiman, 2011). Few analyses of these activities have beenconducted. Similarly, the consequences on safety of organizationalchanges in this field, such as outsourcing of these activities, haveseldom been studied (see however Carballeda, 1999; Tazi, 2008).Yet “maintenance activities have been under various organizationalchanges and restructuring initiatives” (Reiman, 2011; p.340). Thislack of literature and of studies can probably be explained to someextent by the fact that the nature and conditions under whichmaintenance activities are carried out make observation of themdifficult, and even more difficult if they are outsourced, since, inparticular, the workplaces are constantly changing (EuropeanAgency for Safety and Health at Work, 2010b).

For these reasons, this paper presents two different approachesof the relationships between maintenance and health and safety:

� the first analyses were aimed at better identifying occupationalexposures of maintenance staff and comparing them with thoseof production staff, by using the French SUMER 2003 survey. Thegeneral assumption was that, in many respects, maintenancestaff will be more frequently exposed than production staff. Wealso hoped to be able to distinguish exposures of maintenancestaff according to maintenance organization (internal vs.outsourced maintenance notably). Finally, the results of theuse of the survey should guide the second type of analyses;

� the second analyses were aimed at developing the understand-ing of normal real maintenance activities, i.e., normal mainte-nance activities that are actually carried out, their progress andmanagement, as well as risks associated with them, taking intoaccount the socio-technical system and maintenance organiza-tion within which they lie. To do this, ergonomic analysis1 of thereal maintenance activities were conducted in a public transportcompany. More precisely, the analyses should make it possibleto understand how maintenance interventions are carried outas inserted into various organizational modes (internal vs.outsourced), how tasks and activities are allocated betweenthese various types of staff and what this implies for prevention.

First of all, the analytical framework used in particular for thein-situ observations will be presented. Then, the material andmethods of each of the two phases of the study will be described indetail. Next the main results from the use of the survey and somefindings from in-situ analyses conducted in the public transportcompany will be given. Finally those main results will be discussed.

2. Analytical framework

The analytical framework, on which the in-situ observations arebased, is inspired both by the maintenance management

Fig. 1. Analytical framework of maintenance organization and activities.

2 Specialised maintenance refers to situations in which maintenance tasks areonly carried out by maintenance personnel, who might be multi-skilled orspecialised. Shared maintenance concerns situations where these tasks arecompleted by both maintenance and production staff. As for integrated mainte-nance, the tasks are conducted by production operators or by a combined team ofmaintenance and production operators.

C. Grusenmeyer / Accident Analysis and Prevention 73 (2014) 187–199 189

modelproposed by Hale et al. (1998) and by the Reason and Hobbs(2003) model of the organizational accident in their book“Managing maintenance error”. Each of these models applies tomaintenance and tries to capture the complexity of the socio-technical systems and the combination of organizational andmanagement contributing factors that are necessary but notsufficient to occurrence of a human accident, in the case of Hale’smodel, or of a major accident, in the case of Reason’s model. Thebenefit of combining these two models was then the opportunityof considering maintenance in the analytical framework, both as apotential source of risks for the operators assigned to maintenanceand, in the case of maintenance failings, also as a potential sourceof risks for others and of major accidents. The relationshipsbetween maintenance and operator safety on the one hand andmaintenance and systems safety on the other hand can then betaken into account.

Furthermore, both models consider three very similar levels:the management and policy level, the level of maintenance workorganization (i.e., the planning, procedures and resources levelaccording to Hale; level of conditions in the workplace for Reasonand Hobbs), and finally the level of carrying out the maintenancework (execution and feedback level in Hale’s model; individuallevel for Reason, i.e., local conditions combined with personal acts).They also both estimate that risks have root causes in maintenancework organization and in the system at large. According to Reasonand Hobbs (2003), major accidents involve an interaction betweenlong-standing system weaknesses, i.e., latent conditions, at themanagement and policy level and local triggering events, i.e.,active failures, at the individual level, having a direct impact uponthe system. For Hale et al. (1998), safety results from a series ofinterlocking problem-solving cycles focused on the prevention andcorrection of deviations, i.e., undesired outputs, at each of the threelevels of the model. An accident sequence may therefore begin atmaintenance management and policy level. The conditions createdat this level are subsequently transferred along organizationalpaths to the real maintenance achievement conditions. Suchapproaches should then enable maintenance organization to betaken into account.

This analytical framework has been enriched on two points(Fig. 1). First, given the aim of the analyses, it has been adapted toallow real maintenance activities, normal work and practices to be

processed. Referring to Hollnagel (2012), failures and successes areviewed as having the same source and being “two sides of the samecoin”. In the same way, Rasmussen (1997) considers that often theunexpected side effects from daily and locally rational decisions atdifferent levels of a socio-technical system pave the way foraccident. The stage for an accidental course of events is then verylikely prepared through time by normal activities of manystakeholders in their respective daily work context and a quitenormal variation in behavior can trigger an accident. Thus, theseare the mechanisms generating behaviors in the actual and normal,dynamic work context, that have to be understood. Therefore thedeviations or active failures will be of less concern than theintrinsic variability in usual maintenance activities. Usual mainte-nance activities and their results are then also considered in theanalytical framework (Fig.1). Second, in order to gain a better graspof the complexity of the socio-technical systems and in order tohave a comprehensive and thorough understanding of mainte-nance activities, the analytical framework had to take into accountthe different maintenance organizational modes, which determinemaintenance task allocation, and the way real maintenanceactivities are distributed between the different types of staff(Fig. 1). Indeed, neither of the two previous models explicitlyaddresses maintenance task allocation or the real distributionof maintenance activities, as if they were individual or theundifferentiated result of collective work. These different mainte-nance organizational modes, such as specialized, integratedor shared maintenance,2 or else internal, partially or totallyoutsourced maintenance, would however seem useful elements tobe integrated for a deeper understanding of complex socio-technical systems and their various constituent interactions. Theyinfluence the task allocation and the real distribution of mainte-nance activities between the different internal or external,specialized or multi-skilled, production or maintenance operatorsetc., and, consequently, their potential maintenance interventions

4 A public service delegation is a set of contracts, by which a legal entity governedby public-sector law entrusts the management of a public service, for which it isresponsible, to a private or public delegatee whose remuneration is substantiallyrelated to the results of operating the service.

5 Outsourcing is considered here as the mere fact that some needs of the

190 C. Grusenmeyer / Accident Analysis and Prevention 73 (2014) 187–199

on the same work equipment or installations, simultaneously orsuccessively. These places, where there are changes of departmentand/or supervision, and which separate successive activities(“border areas”) are, for example, known as having potentialeffects on reliability and safety (Faverge, 1970). Particular attentiontherefore needs to be paid to maintenance organizational modes,task allocation and real distribution of maintenance activitiesbetween the different operators, in particular for the ergonomicanalysis conducted in the public transport company. They havethen been added to the analytical framework.

3. Material and methods

3.1. Brief overview and use of the SUMER 2003 survey

The SUMER survey, organized by the French Ministry of Labor,and conducted in conjunction with the French Department ofWork Relations and the Ministry’s Research and StatisticsDepartment (DARES) was designed in 1981 to identify thenuisances to which French employees are exposed during theirwork, at local and national levels. It should allow a mappingof French employees’ exposures and to track their changesover time, in order to define research priorities and appropriateprevention policies. The first survey was conducted in1987 and was repeated every 7–8 years. Data are collectedduring periodic medical interview of the employees with theoccupational physician. The physicians evaluate employees'occupational exposures during their last worked week and theirorganizational and relational constraints in their usual worksituation. In 2003, the survey was based on 49,984 interviews andis representative, due to its scope, of 17.5 million employees,i.e., 80% of all French employees.

The use of the SUMER 2003 survey3 was aimed at describing theexposures of the maintenance population and at comparing themwith the exposures of the production population. The aim was alsoto identify the maintenance staff at greatest risk, according to theactivity sectors and mostly to maintenance organizational mode:exposures of internal and external (subcontractor) maintenancestaff especially. Unfortunately, the survey data did not allow totreat this last point. This use has however contributed to the choiceof the activity sector of the company studied in a second phase andto guide ergonomic analyses (see below).

3.1.1. Identification of the maintenance staff and constitution of thesample of production staff

To identify maintenance staff, a pre-selection was firstperformed using variables of the survey that may be linked tomaintenance. This pre-selection provided 8412records. Second, acategorization of these records was done in particular on thenaming of the occupation and main task (free variables, thereforelikely to be closest to the real respondent's activity or occupation).2829 records out of the 8412 preselected were then selected (33.6%of them).

The sample of production staff was constituted from the“production, manufacturing, building site” modality of the mainfunction variable and by pairing with the maintenance personnelsample, in terms of employee's sex and age and size of theinstitution. The activity sectors could not allow pairing.

3.1.2. Data processingData processing consisted mainly in comparing the occupa-

tional exposures of maintenance staff and production staff, andtheir scores in the Job Content Questionnaire (Karasek’s

3 A data transfer agreement was established with the DARES.

questionnaire) included in the survey (Guignon et al., 2008).Those comparisons use the Chi-Square test for qualitative variablesand student t test for quantitative variables.

3.2. In-situ analyses of maintenance organizational modes andactivities in a public transport company

In a second phase, in-situ analyses were carried out in an urbanpublic transport company of an agglomeration of a large Frenchregion. It operates and manages the bus network of theagglomeration (28 regular bus routes and two shuttles). Thechoice of the company is explained below.

3.2.1. Presentation and characteristics of the companyThe company, a subsidiary of a large private group, is a

subcontractor of the agglomeration community. The agglomera-tion community, which has the legal obligation to provide a publictransport service, entrusts that private company with operatingand managing the public transport network, under a contract ofpublic service delegation.4 The agglomeration community definesand implements the transport and travel policy, specifies thenumber and type of routes, bus frequency and daily service hourson each route, and owns the premises, equipment and vehicles.Only the personnel belong to the private company. This is internalsubcontracting, i.e., operations entrusted to the external firm areperformed at the user company site, its outbuildings or construc-tion sites (INRS, 2004). This internal subcontracting is, however,specific since almost no user company staff work on the premisesof this company.

3.2.2. Organization of vehicle maintenanceUnder the contract of public service delegation, maintenance

of the vehicles and of their interior equipment (seats,ticket-validation equipment, etc.) is handled entirely by theprivate company. Maintenance interventions are in part performedin-house (i.e., by the private company), and in part outsourced.5

An internal maintenance department with fifteen employeesincludes notably, a department manager, two team leaders,8 multi-skilled mechanics, working in pairs and an electronicstechnician. One of the two mechanics on duty (morning orafternoon) is a field operations technician or “flying technician”,the mechanics taking it in turns to perform such duties: the flyingtechnician has a radio, which connects him or her to theregulation unit, and he or she intervenes on the bus network,when defects are reported by drivers to regulators (immediatecorrective interventions). The internal department also performspreventive and corrective maintenance interventions (e.g.,immediate troubleshooting and larger corrective operations).

In addition, a number of maintenance interventions areoutsourced. Depending on an outsourcing budget envelope, thevehicle fleet and the human resources of his or her departmentnotably, the maintenance manager chooses whether or not tooutsource some of the interventions.

Analyzing technicians’ activities in this situation thus presentedseveral interests, in particular due to:

company are provided by an external firm, but other definitions exist. The usercompany is the company that outsources the tasks. The outside/external company isthe company that provides the services (European Agency for Safety and Health atWork, 2012). Subcontracting is considered as a form of outsourcing among others.

C. Grusenmeyer / Accident Analysis and Prevention 73 (2014) 187–199 191

� the fact that vehicles maintenance staff will be identified amongthe most exposed to the various constraints in the use of theSUMER survey (see below);

� the subcontracting relationship between the private companyand the agglomeration community;

� the coexistence of different vehicle maintenance organizations,internal and outsourced. This should therefore enable carryingout of interventions to be observed with different organizationmodes;

� the opportunity of observing corrective maintenance interven-tions, in particular immediate ones (bus network interventionsor troubleshooting), shown by several studies as being moreaccident-prone (Grusenmeyer, 2005; European Agency for Safetyand Health at Work, 2010a,b).

3.2.3. MethodologyThe methodology used was aimed firstly at describing and

characterizing the vehicle maintenance organization of thecompany. It was intended secondly to conduct ergonomic analysisof, in particular, corrective maintenance activities by internal andexternal stakeholders. As we will see in the main results section,the network of companies, within which this firm lies, fromthe viewpoint of vehicle maintenance, turned out to be verycomplex. Many external companies (and therefore a multiplicityof stakeholders) form the network, which did not enableoutsourced interventions to be observed. Such observationwould, at the least, have required negotiating our interventionwith each of the companies. This methodological difficulty wasincreased by the choice of analysing immediate correctivemaintenance interventions by means of observations insofar as,in such situations, the malfunction is not known a priori, and norare the technicians who are likely to intervene. As mentioned byDe La Garza (2000), troubleshooting activities, by natureunannounced, make it impossible to define accurate observationconditions. Consequently, the ergonomic analysis of maintenanceinterventions only concerned the internal staff. Nevertheless,particular attention was paid, whenever possible, to the potentialconsequences of outsourcing on the internal technicians’activities, which is a poorly studied mechanism; and a fewmaintenance interventions, during which internal technicianswere in contact with outside personnel, were the subject of clinicalanalyses(i.e., thorough qualitative analyses of some cases or situations,besides non-specific of ergonomics).

The methods used were the following:

� semi-structured interviews with the maintenance manager.They were designed to identify the network of companiesinvolved in vehicle maintenance and the nature of inter-firmrelationships;

� equipped observations (by audio–visual recordings6) of theactivities of internal “flying technicians” and their interactionswith others. Monitoring these technicians’ activities shouldallow immediate corrective maintenance interventions to beobserved. However, all their activities were observed, sincetroubleshooting activities are unannounced;

6 Two types of audiovisual recordings were performed simultaneously: a"handheld camera" recording by an observer gave an overview of the technicians’activities and their immediate context, as well as their potential contacts; a micro-camera positioned on the technicians' shoulder and a lapel microphone on theircollar enabled their proximal work area to be observed (that area being most oftennot observable from the preceding recording, because of low access to equipment tobe maintained, for example) and a better audibility of their verbalizations andcommunications to be obtained, the environment being likely to be very noisy.

� concomitant verbalizations of the technicians throughout theobservations. They should make it possible to obtain operators'details and explanations on their activities in progress.

These observations and verbalizations were expected to enablethe time course of different flying technicians’ activities to betracked. More specifically, they should enable their workinterruptions to be identified. Indeed, as the results of the useof the SUMER survey will show, maintenance operators' tasks arenotably characterized by frequent interruptions for anotherunexpected task.

In addition, diachronic analyses7 of maintenance interventionswere conducted in order to comprehend the distribution of realwork activities among various types of staff. The use of theSUMER survey will indeed point high trends in fragmentation andlack of predictability of the maintenance staff tasks. For thispurpose, and with the diachrony of interventions going beyondthe activities of one stakeholder, the previous methods have beensupplemented by:

� audio recordings in the regulation room. They were expected tohelp gather evidence related to detection and reporting ofmalfunctions by drivers to regulators, and to how they dealt withthese malfunctions;

� equipped observations (by audio–visual recordings) of othermaintenance technicians’ activities, if an intervention on thebus network, after being dealt with initially by flyingtechnicians, is taken up by or conducted jointly with anotherinternal colleague (these observations have not always beenpossible).

The main constitutive phases of maintenance interventions onthe bus network should thus be monitored and clinical analyses ofsome of them carried out.

Regarding the data collected, four interviews were conductedwith the maintenance manager. The activities of flying technicians,their interactions with others and concomitant verbalizationswere recorded during 6 shifts (3 in the morning, 3 in theafternoon). This represented a little more than 20 hours ofobservation. Verbal communications between drivers andregulators when reporting a malfunction could only be collectedon 4 of the 6 shifts. Data collected concerned 5 different “flyingtechnicians”, 22 maintenance interventions and 17 different buses.

Data processing consisted of:

� content analyses of the interviews with the maintenancemanager (identification of the network of outside companies);

� a description and characterization of flying technicians’ activitieson different levels:

- activity analyses with respect to the interventions diachrony. Foreach intervention, the maintenance phases (failure detection,troubleshooting, pre-diagnosis, diagnosis, repair, equipmenttests and vehicle return to the regulation), to which flyingtechnicians activities contributed, were identified. If thesephases were not carried out by flying technicians, the staffperforming them were identified, thanks to all data collected.Diachrony of each intervention (different phases) was thenplotted based on the distribution of real work activities amongvarious stakeholders (and coordination phases needed couldthen be identified);

7 That is clinical analyses whose particularity is to give a major place to thehistory of events in order to grasp their sequential, scalable and constructivistnature (they are non-specific of ergonomics).

Table 1Maintenance staff and production staff exposures to ambient and physical constraints.

Ambient and physical constraints

Nature of exposure Maintenancestaff sample

Production staff sample All professional families

Noise 69% (1957) 66% (1871) 32%a

x21 = 5.9, p < .02

Thermal constraints 35% (989) 33% (937) 21%x2

1 = 2.1, p > .05Radiance and radiation 14% (394) 9% (252) 4%

x21 = 35.2, p < .0005

Situations with visual constraints 56% (1570) 42% (1197) 55%x2

1 = 98.4, p < .0005Manual handling of loads 57% (1623) 56% (1589) 43%

x21 = 0.8, p > .05

Postural and articular constraints 88% (2502) 89% (2522) 72%x2

1 = 0.7, p > .05Working with vibrating tools and machinery 45% (1259) 23% (652) 12%

x21 = 291.1, p < .0005

Driving situations 58% (1652) 33% (938) 33%x2

1 = 363, p < .0005

a Results concerning all professional families are taken from Equipe SUMER (2006).

8 We should remember that the higher the psychological demands score is, thehigher the psychological job demands (quantity-rapidity, complexity-intensity,splitting-predictability) are. The lower the job decision latitude score is, the moreinsufficient the resources for coping with the demands (margin of maneuver,current use and development of skills) are. In these situations (high psychologicaldemands/low decision latitude, i.e., "job strain"), the risks of deteriorations in healthare considered high. Social support modulates this imbalance: it worsens thenegative impacts on health if weak ("isostrain").

192 C. Grusenmeyer / Accident Analysis and Prevention 73 (2014) 187–199

- analyses of the succession of these activities for the technicians.Based on observations, interventions to which each techniciancontributed successively in time, and the nature and durationof their successive activities were graphed. Their workinterruptions could thus be identified, then categorized as afunction of their nature (occurrence of interfering activity/postponement) and their sources (nature of the interferingactivity/postponement reasons);

� clinical and diachronic analyses of some maintenance inter-ventions, for which a temporal reconstruction of their coursehas been done on the basis of all the data (observations,verbalizations, communications, etc.). This reconstructioncombined a temporal structure (history of the intervention)and a functional structure (real work activities of the variousstakeholders), at which the crossings of different activities orevents and their relationships were placed. Various elementscontributing to this history were then searched (in verbalizationsnotably).

4. Main results

4.1. Main results from the use of the SUMER 2003 survey

4.1.1. 1 Comparison of employees’ occupational exposures to ambientand physical constraints

From the view point of the ambient and physical constraints(Table 1), maintenance staff are generally significantly morefrequently exposed than production staff to driving situations,working with vibrating tools and machinery, situations withvisual constraints, radiance and radiation, noise and, to a lesserextent (non significant difference), thermal constraints andmanual handling of loads, in descending order of percentagedifference. Production personnel appear slightly more frequentlyexposed than maintenance personnel to postural and articularconstraints, when these exposures are evoked in a general way inthe survey. When mentioned in an instantiated way, they arealso more frequently standing and shuffling about and submittedto repetitive movements at high speed. Maintenance staff arehowever significantly more frequently exposed to walking,kneeling positions, fixed positions of the head and neck,keeping arms up and other postural constraints (squatting,twisting, etc).

4.1.2. Comparison of employees’ occupational exposures toorganizational and relational constraints

Several dimensions related to organizational and relationalconstraints in the survey are hardly suited to maintenanceactivities. Consequently, part of the results will be presented.They mainly show (Table 2) that task interruptions for anotherunexpected task are estimated to be much more frequent formaintenance personnel than for production staff. However, thestated impossibility of varying deadlines is far greater forproduction staff. The results highlight greater independence andgreater margins of initiative, but also more serious consequencesof an error for maintenance staff. Means of working aresignificantly more often deemed by maintenance operators tobe insufficient for proper completion of work: insufficientinformation, insufficient number of colleagues, inadequate andinsufficient material resources, inappropriate and insufficienttraining.

Finally, staff involved in aircraft, agricultural and vehicle (cars,trucks, etc.) maintenance appear the most exposed to the variousphysical and organizational constraints.

4.1.3. Comparison of employees’ scores in the job contentquestionnaire

The results of the job content questionnaire8 (Karasek et al.,1998), underscore psychological demands and decision latitudescores that are, on average, significantly higher for maintenancestaff compared to production staff (psychological demands:21.5 vs. 21.1 respectively, t2647 = �2.4, p < .02; decision latitude:73.8 vs. 68.5, t2681 = �12.6, p < .0005). Concerning the decisionlatitude scores, the same trend is observed for its three constitutivedimensions (margin of maneuver, current use and developmentof skills). Conversely, the scores of job demands in terms ofquantity–rapidity are higher for production staff, while thoseconcerning complexity–intensity and splitting–predictability arehigher for maintenance staff. Maintenance staff are classified as

Table 2Maintenance staff and production staff organizational and relational constraints.

Organizational and relational constraints

Nature of constraintsa Maintenance staff sample Production staff sample All professional families

Pace of work constraintsTasks interruptions for another unexpected task 66% (1878) 47% (1317) 58%b

x21 = 226.3, p < .0005

Impossibility to vary deadlines 25% (711) 44% (1233) 36%x2

2 = 315, p < .0005Independence and margins of initiativeAction in case of incidentDealing with it personally 63% (1782) 51% (1438)Dealing with it personally in specific cases 21% (607) 24% (687)Call upon others 15% (434) 25% (697) 22%

x22 = 102.8, p < .0005

Consequences of an errorSerious for the product/service quality 89% (2491) 82% (2323) 69%

x21 = 42.9, p < .0005

Significant financial costs 84% (2341) 73% (2052) 58%x2

1 = 88.2, p < .0005Serious for safety 77% (2159) 51% (1426) 37%

x21 = 420.4, p < .0005

Sanctions 58% (1601) 48% (1331) 47%x2

1 = 54.6, p < .0005Collective workTo complete work properly, do you have:Insufficient information 23% (645) 15% (420) 19%

x21 = 58.5, p < .0005

An insufficient number of colleagues 27% (750) 20% (551) 26%x2

1 = 38.2, p < .0005Impossibility of cooperation 4% (106) 7% (183) –

x21 = 21.8, p < .0005

Inadequate and insufficient material resources 27% (747) 22% (628) 20%x2

1 = 13.2, p < .0005Inadequate and insufficient training 25% (695) 17% (485) 20%

x21 = 47.4, p < .0005

a Not all the organisational and relational constraints are itemised here.b Results concerning all professional families are taken from Equipe SUMER (2006).

C. Grusenmeyer / Accident Analysis and Prevention 73 (2014) 187–199 193

“active”: psychological demands score and decision latitude scoreabove the medians observed for all employees (respectively,medians of 21 and 70.3); while production staff are “tense” (highstrain), and therefore more subject to “job strain”. Both havesimilar social support scores (23.8 for maintenance staff, 23.5 forproduction staff; t2620 = �1.9, p > .06); but while productionpersonnel receive more support from their superiors, maintenancestaff are more supported by their colleagues. Thus, 19.3% of theformer are in “isostrain” situations vs. 12.9% of the latter (x2

1 = 19.1,p < .0005).

4.2. Some findings of in-situ analyses of maintenance organizationalmodes and activities

4.2.1. Identification of the network of companies involved in vehiclemaintenance

The content analysis of interviews with the maintenancemanager of the private company made it possible to identify thenetwork of companies to which it belongs for bus maintenance.This network is shown in Fig. 2. This graphical representation ofthe network of companies is drawn from studies about thereticular organization of companies (Rorive, 2003; Mariotti, 2005).These studies consider that the boundaries of firms becomeblurred due to multiple relationships with other companies:alliances, subsidiarization, partnerships, outsourcing, provision ofservices, subcontracting, etc. The network of companies thenbecomes a set of nodes interlinked by connections, in which theconnections between the entities are more important than theentities themselves. These connections have been characterizeddepending on the nature of the companies’ relationships:

� duration of the relationship (occasional/annual or multiannual);� site where maintenance interventions are carried out (on-site atclient firm/on-site at outside firm);

� company determining the content of the intervention: contentdefined by the user company (i.e., highly personalized servicesfor the user company)/by the outside company (standardizedservices, i.e., same services regardless of the user company).

The distinctions are based on the terminology used in Franceand are our suggestions, boundaries between different forms ofoutsourcing still remaining vague, open to debate and being notalways defined (see for example Mayhew et al., 1997; Tazi, 2008;Nenonen, 2011; Quinlan et al., 2013).

Fig. 2 shows that the network of companies involved in vehiclemaintenance is complex. This complexity stems from:

� the number of involved companies, since the internal mainte-nance department has relationships with 16 different firms;

� the diversity of the nature of the inter-firm relationships:

- internal (on-site at client firm) or external (on-site at outsidefirm) subcontracting relationships (Grusenmeyer, 2009), i.e.,highly personalized services, that are not occasional, and whosecontent is to a large extent defined by the user company;

- provision of a service (“sale of service”), i.e., more standardizedservices, that are occasional and whose content is a specific taskrequiring special expertise that the user company is not able toprovide internally (it takes place, as appropriate, on the outsidefirm’s site or on the client firm’s site, i.e., on the agglomerationcommunity site);

Fig. 2. Representation of the network of companies involved in bus maintenance.

194 C. Grusenmeyer / Accident Analysis and Prevention 73 (2014) 187–199

- guarantee, i.e., highly standardized services that are generallyoccasional and whose content is very specific and determined bythe seller;

� the fact that different types of relationships can be maintainedwith the same outside company (e.g., guarantee through themanufacturers, then external subcontracting and/or provision ofa service);

In addition, some vehicle maintenance interventions areoutsourced, not by the maintenance department of the privatecompany, but directly by the agglomeration community (e.g., busrenovation, ameliorative maintenance). Furthermore, mainte-nance outsourcing does not necessarily mean exclusive comple-tion of the involved interventions by the external staff only. Forinstance, the bodywork is subject to full outsourcing, while majormechanical interventions that are mostly outsourced may also beperformed in-house, depending on their nature and duration (cf.Fig. 2).

4.2.2. Brief general description of flying technicians’ activitiesThe flying technicians’ main activities contribute to the

completion of the 22 maintenance interventions. Most of them(21 out of 22) involve corrective interventions. The remainingactivities include subsidiary activities (e.g., refueling of vehicles),other activities (such as end-of-shift activities), breaks, but also“waiting” periods or, more accurately, periods during which thetechnicians cannot intervene. Most of these 22 interventions(77.3%) are not planned or only partially planned (planning of arepair after prior immediate troubleshooting). Their main task is to

intervene on the bus network to deal quickly with anymalfunctions.

4.2.3. Fragmented activities with regard to intervention diachronyObservations of the real activities of internal flying technicians

should notably enable the fragmentation of their activities andtheir work interruptions to be analysed. On the first point, theanalyses of these activities with regard to maintenance interven-tion diachrony highlight the contributions from these operators tocompletion of such interventions. Thus, tracking these technicians’activities rarely enabled interventions to be observed in theirentirety. Out of the 22 interventions, to which these technicianscontributed, only 6 (27.3%) were observed in their entirety and 16(72.7%) were observed only partly. Either the operation triggeringcould not be tracked (3 cases; e.g., detection by external staff ofanomalies, which are dealt with by internal staff). Or the rest/endof the intervention could not be observed (6 cases; e.g., the rest ofthe intervention, depending on its nature, being carried out byanother internal or external stakeholder). Or these two phasescould not be tracked, for the reasons just mentioned (7 cases).Flying technicians’ activities therefore appear fragmented withregard to the time course of maintenance interventions. The onesthey complete in full are essentially troubleshooting interventionsof very short duration, not resulting in a repair in a second stage.

Various factors explaining this fragmentation have beenidentified:

� execution in a first stage, of immediate corrective maintenancework by flying technicians, whoever the internal/externalstakeholders are who carry out the intervention in a second

C. Grusenmeyer / Accident Analysis and Prevention 73 (2014) 187–199 195

stage. The internal maintenance department organization andalso the operations outsourcing, regardless of its form, thuscontribute to this fragmentation9;

� the technicians’ rotation system, the need to plan correctiveactions that are unscheduled, and management of spare parts.These situations have to be seen in the context of difficultiesencountered by the internal department for performing morepreventive maintenance (hence unscheduled corrective inter-ventions are all the more numerous), and a stock of spare partsconsisting mainly of consumables and common spares (due tothe financial cost of a more substantial stock and the very highvariability of the bus fleet and therefore of parts which dependon the agglomeration community, as vehicle owner);

� service constraints. They may lead to partial unavailabilities ofvehicles (i.e., limited in time) for maintenance, not alwaysallowing technicians to complete interventions. They may alsolead to opt for troubleshooting by drivers (especially at peakhours), to the detriment of proper vehicle operation (no finalsolutions, high stresses on equipment).

These different distributions of activities between the stake-holders with regard to the intervention diachrony make it possibleto avoid fault occurrence encroaching on the service provided.However they contribute to the dissociation of the differentcomponent phases of interventions from the various stakeholders.Such dissociations are therefore likely to harm the course ofintervention and require significant phases of coordinationbetween the different stakeholders. In addition, these resultsindicate that outsourcing of maintenance interventions does notmean that internal technicians do not contribute to carrying themout (phases preparatory and subsequent to their completionespecially).

4.2.4. Interruptions in the flying technicians’ activitiesThe analyses of the succession of the flying technicians’ activities

emphasize a substantial number of work interruptions, as defined byBoehm-Davis and Remington (2009). Two types of interruptionswere distinguished between: first, those related to the occurrenceof an interfering task (Boehm-Davis and Remington, 2009),i.e., unexpected occurrence of tasks, whose execution either requiresinterruption of the task inwhichthe technician is alreadyengaged,orwill take place simultaneously to it (Eyrolle and Cellier, 2000);second, interruptions of the “postponement” type: the technician isengaged in an “initial” intervention, but events do not allow him orher to pursue it, and it has to be postponed. 24 interruptions areobserved over the entire intervention time of flying technicians, i.e.,1 every 47 min on average or 1.3 per hour. Nearly two-thirds (15)concern the occurrence of an interfering task, e.g., immediatecorrective maintenance interventions, casual detection of anothermalfunction. The remaining third (9) are of the “postponement”type. Many are related to the lack of spare parts required; but mainly,some arise from the joint regulation of several constraints by theflying technicians and team leaders:

� constraints of service; hence decisions of partial unavailabilityfor some maintained buses, not always allowing technicians tocomplete the interventions before the vehicles are put back intoservice;

� constraints of capacity and constitution of the fleet (under theresponsibility of the agglomeration community), insofar as theydetermine the number of reserve buses, and therefore the

9 Conversely a failure, whose repair is ensured by the internal maintenance staff,may be detected during an operation carried out by an outside operator.

possibility of taking one of them out of service for maintenancepurposes;

� constraints related to vehicle reliability (depending on theirdesign, the choices made by the agglomeration community, theiruse and maintenance). They determine the number of unavail-able and reserve buses;

� and also, constraints related to the outsourcing of interventionson-site with outside firms. Since the buses in question are not onthe spot, the only possible margin of maneuver, with regard tovehicle availability and operation constraints, is with buses onwhich internal interventions are carried out: greater reactivity;schedules and deadlines easier to control or modify.

Thus, different regulations may be implemented by team leadersand internal technicians: changes in the degree of availability ofvehicles that undergo internal interventions, periods when carrying-out interventions is hindered, reports of repairs to be done,technicians’ activities interruptions. These regulations are opera-tional to the extent that they contribute to ensuring continuity ofservice. The latter nonetheless illustrate the great uncertainty andthe low work predictability associated with maintenance activities(Reiman, 2011). These adjustments have consequences for the flyingtechnicians’ activities: e.g., contributions to the same interventionsthat are themselves fragmented, more limited time to performinterventions, performing reassembly operations in emergencysituations, interventions on a bus that has just been running (warmengine and parts). Activity fragmentation, identified as one of thecharacteristics of time pressure (Coeugnet et al., 2011), is likely to becognitively costly for operators, especially when they are carryingout complex maintenance. The low predictability is likely to leadoperators to adapt as much as possible, while reducing theiranticipation possibilities (Coeugnet et al., 2011) and preparation ofinterventions (e.g. fault diagnosis; Sykes 2011). Work interruptionsmay not allow technicians to complete interventions beforeequipment is brought back into service (hence a risk of non-optimaloperation of the equipment). They are considered as a potentialsource of stress (European Foundation for the Improvement of Livingand Working Conditions, 2007) and hardness likely to lead toincidents, errors, omissions or accidents (Eyrolle and Cellier, 2000;Reason and Hobbs, 2003; Boehm-Davis and Remington, 2009;Kositts and Jones, 2011).

More generally, this illustrates the difficulties for the internaldepartment to have vehicles in order to conduct interventions onits own initiative.

4.2.5. Example of clinical and diachronic analyses of some bus networkinterventions

Such analyses were conducted for a few maintenanceinterventions, during which flying technicians interacted withexternal staff. An example is related to an “unthought-of” situation(i.e., whose unexpectedness relates to the nature of the event, e.g.,exceptional malfunctioning; Cuvelier et al., 2012) involvingexternal staff and causing an incident and the transmission of adangerous situation.

In this situation, a fault of the joint system of a jointed busoccurred. During previous manifestations of the same fault, nomalfunction had been identified by the external staff in charge ofthe maintenance of these systems. Consequently, this fault wasconsidered as an untimely failure of the signal, without realmalfunction, by the external, and then internal maintenance staff,and a decision to continue the tour without intervention wastaken. This decision led to the occurrence of an incident, withconsequences on driving safety: swerving of the bus trailer,causing a deviation from its trajectory and a collision with thesidewalk. A flying technicians' intervention was then set up(vehicle swap) and the bus was deposited by them with theexternal company, dealing with such interventions. (cf. Fig. 3).

Fig. 3. More detailed reconstruction of events causing an incident and the transmission of a dangerous situation.

196 C. Grusenmeyer / Accident Analysis and Prevention 73 (2014) 187–199

Several elements contribute to this situation. Among the mainones, the complete outsourcing of interventions on these systems(except for troubleshooting) should, in particular, be highlighted.The lack of identification of the malfunction by external operatorsand the lack of internal expertise on these systems, coupled withthe absence of obvious malfunction, lead to a situation in whichinternal staff are not in a position to re-examine the diagnosismade by the external staff. The fault was considered as a falsealarm (signal fault only) both by external and internal staff, whichled to the incident. Finally, the analysis of this situation shows thetransmission of the diagnosis from external company to the usercompany and, in a second stage, transmission of the malfunctionand its associated risks, from operation (driver) to the internalmaintenance staff, and probably to the outside company staff.These transmissions have to be seen in the context of the internalmaintenance organization modes and the maintenance organiza-tion modes beyond the firm. They reflect the interdependencerelationships between operation and internal maintenance(Grusenmeyer, 2002), and also such relationships betweeninternal and external maintenance staff, at least in some cases.

5. Discussion

5.1. Maintenance organizational modes and safety: relationshipschallenging to study

Firstly, this study shows that the relationships between themaintenance organizational modes and safety remain a challengeto study. The use of the SUMER 2003 survey was expected allowing

grasping occupational exposures of maintenance staff according tothe maintenance organization, in particular exposures of internalvs. external maintenance staff. However, specialized maintenancecompanies, as well as outside operators, are difficult to identify.Furthermore, several dimensions within organizational andrelational constraints in the survey are not suited to maintenance.

In the same way, difficulties in observing the outsourcedinterventions were highlighted. Such observations imply negoti-ating with each company likely to be involved. Such observationsalso imply the possibility for observers to anticipate whichstakeholders are likely to be concerned, as many observers asthere are stakeholders, and to coordinate of these observations.The initial objective aimed at following the story of a maintenanceintervention and the real network of different contributingstakeholders (internal vs. subcontractor) has thus not beenpossible.

However, the use of the SUMER survey showed that, on manylevels, maintenance personnel are significantly more frequentlyexposed than their production colleagues to various constraints.This is notably the case for most of the physical and ambientconstraints, but also for some organizational and relationalconstraints. Maintenance operators appear to have a greaterindependence and larger margins of initiative than productionstaff. On the other hand, the psychological demand scoresunderline higher job demands for maintenance staff. At last,results notably show frequent task interruptions and highfragmentation and lack of predictability of the maintenance tasks.These latter results were consequently used to guide analyses oftechnicians' activities in the company.

C. Grusenmeyer / Accident Analysis and Prevention 73 (2014) 187–199 197

As for ergonomic analyzes, they finally focused on the activitiesof the internal maintenance staff, by paying particular attention tothe consequences of outsourcing on these activities and tointeractions of these personnel with external staff. Neverthelessthis work raises the question of studying activities in inter-relatedcompanies. These difficulties in conducting observations of theactivities of the different stakeholders within systems whose sizeand complexity are increasing (Wilson and Carayon, 2014), or evenwithin systems of systems (Siemieniuch and Sinclair, 2014), exceedthe issue discussed here, but are likely to be increasingly met.

5.2. A complex network of companies involved in maintenanceactivities

One of the main outcome of this study concerns the particularlycomplex network of companies involved in maintenance activities:number of companies involved, nature of inter-firm relationships,possibilities of multi-nature relationships with the same company,

Fig. 4. New analytical framework of main

outsourcing by level 1 (agglomeration community) and level 2(private company) user companies.

Yet studies that have sought to identify and characterizepragmatically the networks of companies involved in one (ormore) function(s) are, to our knowledge, scarce. Nonethelessreticular organization of companies has increased considerably inrecent years and knowledge in this field should be developed. Theterminology concerning the nature of the various inter-firmrelationships remains poorly stabilized and is open to debate.Definitions of or indeed distinctions between different forms ofoutsourcing are hardly congruent. This is why we distinguishedhere between several types of relationships, although it is clearthat these distinctions are neither exhaustive nor unquestionable.In any case, the networked operation of companies invites us toquestion their consequences in terms of health and safety. Forexample, what we defined as provision of services could beparticularly critical to health and safety to the extent that it can beperformed at the user company’s site, but in a much more

tenance organization and activities.

198 C. Grusenmeyer / Accident Analysis and Prevention 73 (2014) 187–199

occasional way, than would have been possible in an internalsubcontracting relationship.

5.3. Interdependence relationships between internal and externaltechnicians’ activities

Above all analyses of flying technicians' real activities hasshown that outsourcing interventions, whatever the form ofoutsourcing, does not mean a lack of contribution by internaltechnicians to completing such interventions. Regarding correc-tive maintenance, and whatever the malfunction, it is indeedflying technicians who will conduct the troubleshooting andfailure pre-diagnosis or diagnosis, prior to the execution of theintervention itself by external technicians. Conversely, a mal-function that is repaired internally may be detected anddiagnosed by an outside company. Outsourcing thus led, amongother factors, to a fragmentation of flying technicians’ activitieswith regard to the intervention diachrony and therefore to such afragmentation for external stakeholders. This dissociation of thedifferent constituent phases of interventions from these variousstakeholders is likely to be detrimental to interventions takingtheir proper course, and to safety and reliability, if it is notaccompanied by significant coordination or cooperation phases.Similarly, the clinical analyses of some interventions havehighlighted the interrelationships between the internal andexternal maintenance staff activities. In the same way, inter-ventions being outsourced on external sites may lead to changesin the degree of availability of vehicles for internal maintenancestaff, to work interruptions, intervention schedule adjustments,intervention reports of the latter or periods where internalmaintenance activities are hindered.

Thus maintenance outsourcing in various forms raises notonly the question of interference risks between activities,facilities, and equipment of different companies on the sameworkplace (INRS, 2004), and risks related to the coactivity ofthese various types of staff or risks for external staff's healthand safety (Tazi, 2008; Nenonen, 2011). It also appears to raisethe question of risks associated with the interdependence ofreal work activities undertaken by internal and external staff, orexternal technicians from different companies, since theseactivities contribute to the same goal, the same maintenanceintervention. The issue is then that of coordination betweensuccessive activities in which maintenance operators initiate orcontinue others' work, given that these activities do notnecessarily take place in the same workplace, or involvestakeholders belonging to the same structures, which may ormay not have formal relationships.

For these reasons, and those mentioned previously, theanalytical framework initially proposed needed to be supple-mented (Fig. 4; for easy readability, feedback links are notpresented), integrating inter-organizational interactions (Boholmet al., 2012; Cedergren, 2013), and therefore:

� firstly, the structure (group, responsible authority or company)using the maintenance function or which owns the company incharge of the latter (in the case studied here, the agglomerationcommunity), to the extent it delegates (in various forms) themaintenance function, defines the political, strategic andenvironmental context, in which this function will be inserted,and may also outsource some interventions to other companies,that may maintain interdependence relationships with theinterventions carried out by the company in charge of themaintenance function;

� secondly, the network of companies involved in maintenanceinterventions and their relationships, as well as the planned andactual coordination means between the various types of staff, if

they contribute to the same successive interventions. Situationsobserved here seem indeed to be found in other networks ofcompanies, at least because immediate corrective maintenanceinterventions require high reactivity that the internal mainte-nance staff are probably alone in being able to provide, even if theinterventions in question are outsourced.

5.4. Resulting measures for prevention

In terms of prevention, this study leads to formulation ofvarious adjustment suggestions or recommendations to compa-nies and prevention specialists; for example:

� involving maintenance staff and/or the manager in the designchoices or acquisition choices for the equipment;

� taking the tools and means of work required for maintenance ofthe equipment into account when choosing or acquiring it;

� avoiding excessive variability of the equipment to maintain andtherefore excessive variability of the required tools and means ofwork;

� informing the internal maintenance staff of any improvements orrenovations of equipment, even when such ameliorativemaintenance interventions and the usual equipment mainte-nance are outsourced;

� systematically informing flying technicians of the reasons andsigns of faults leading to immediate corrective maintenanceinterventions;

� estimating the number of items of equipment needed foroperation or service, and also for coping with unavailabilitiesgenerated by internal and outsourced interventions for optimummaintenance;

� promoting preventive maintenance as far as possible, whichshould reduce immediate corrective maintenance interventions,and thus the fragmentation of the technicians' activities; etc.

More specifically, this study suggests considering, as farupstream as possible, the potential interrelationships betweenthe various maintenance tasks outsourced and between those thatare performed in-house and those that are outsourced.

6. Conclusions

In conclusion, this study combined several approaches tomaintenance work: statistical treatment of results of a survey onoccupational exposures, interviews to identify the network ofcompanies involved in maintenance, analyses of actual realmaintenance activities as inserted in a particular context. Theseapproaches have not always kept all their promises and mayappear quite different. However, some have helped to guide or toenlighten others. More importantly, maintenance activities arenow inserted in systems, whose size and complexity are such, thatonly a combination of several approaches seems able to claim toapprehend them according to our view.

Furthermore this study has stressed that outsourcing ofmaintenance activities impacts not only the external stakeholders'health and safety, as highlighted in the literature, but also the work,health and safety of internal staff. Above all it has revealedinterdependence relationships between the activities of differentcorporate stakeholders and, in this way, the importance ofcoordination of these activities for health and safety, given thatthese activities do not necessarily take place in the same workplaceor involve stakeholders belonging to the same structures. This isthe reason why, associated with these results, preventativemeasures have been proposed.

C. Grusenmeyer / Accident Analysis and Prevention 73 (2014) 187–199 199

Acknowledgments

The author wishes to thank Pascal Wild (INRS, V/DS) andNathalie Guillemy (INRS, P/DCP) for their very constructivecontributions, the DARES for its consent for use of the SUMER2003 Survey and Prof. Christine Chauvin (Université de BretagneSud Lorient, France) for her very relevant feedback on this paper.

References

Batson, R.G., Ray, P.S., Wan, Q., Weems, W.H., 1999. How preventive maintenanceimpacts plant safety. Proceedings of the Annual Conference on Maintenanceand Reliability., Gatlinburg TN, Maintenance and Reliability Center, Universityof Tenessee.

Boehm-Davis, D.A., Remington, R., 2009. Reducing the disruptive effects ofinterruption: a cognitive framework for analysing the costs and benefits ofintervention strategies. Accid. Anal. Prevent. 41, 1124–1129.

Boholm, A., Corvellec, H., Karlsson, M., 2012. The practice of risk governance: lessonsfrom the field. J. Risk Res. 15 (1), 1–20.

Bounot, J., Mazeau, M., Jules, D., 1996. La maintenance des bus: analyse des sourcesd'accidents. Perform. Hum. Techniq. 83, 20–30.

Bourrier, M., 1996. Organizing maintenance work at two American nuclear powerplants. J. Conting. Crisis Manag. 4 (2), 104–112.

Carballeda G.,1999. La contribution des ergonomes à l'analyse et à la transformationde l'organisation du travail: l'exemple d'une intervention relative à lamaintenance dans une industrie de processus continu. Presses de l'Universitéde Bordeaux 2, Bordeaux, 289 p.

Cedergren, A., 2013. Implementing recommandations from accident investigations:a case study of inter-organizational challenges. Accid. Anal. Prevent. 53,133–141.

Coeugnet, S., Charron, C., Van de Weerdt, C., Anceaux, F., Naveteur, J., 2011. Lapression temporelle: un phénomène complexe qu'il est urgent d'étudier. LeTravail Hum. 74 (2), 157–181.

Cuvelier, L., Falzon, P., Granry, J.C., Moll, M.C., 2012. Managing unforeseen events inanesthesia: collective trade-off between understanding and doing. Work 41,1972–1979.

De La Garza, C., 2000. L'activité de diagnostic dans un système dynamique: le cas dudépannage d'installations de signalisation ferroviaire. Rev. Eur. Psychol. Appl. 50(1), 39–49.

De La Garza C., Weill-Fassina A.,1995. Les modalités de gestion collective des risquesferroviaires sur des chantiers d'entretien des voies. Recherche, Transports,Sécurité 49, 73–84.

Desriaux, F., 2005. Les coopérations dans le travail construisent la santé. SantéTravail 52, 23.

Duhamel, C., 2005. La sous-traitance: maillon faible des savoir-faire de prudence.Santé Travail 52, 36–37.

Equipe SUMER, 2006. Les expositions aux risques professionnels par familleprofessionnelle. Résultats SUMER 2003. Document d'Etudes de la DARES 121,274 p.

European Agency for Safety and Health at Work, 2009. Healthy workplaces. AEuropean Campaign on Safe Maintenance. Publication of the European Agencyfor Safety and Health at Work, Luxembourg, 20 p.

European Agency for Safety and Health at Work, 2010a. Safe Maintenance inPractice. Publication Office of the European Union, Luxembourg, 102 p.

European Agency for Safety and Health at Work, 2010b. Safe Maintenance inPractice. Publication Office of the European Union, Luxembourg, 102 p.

European Agency for Safety and Health at Work, 2012. Procurement of maintenanceservices and health and safety at work. E-facts 63, 10 p.

European Foundation for the Improvement of Living and Working Conditions, 2007.Working conditions in France. Electronic report. http://www.eurofound.europa.eu/ewco/surveys/FR0410SR01/FR0410SR01_4.htm.

Eyrolle, H., Cellier, J.M., 2000. The effects of interruptions in work activity: field andlaboratory results. Appl. Ergonom. 31, 537–543.

Farrington-Darby, T., Pickup, L., Wilson, J.R., 2005. Safety culture in railwaymaintenance. Safe. Sci. 43 (1), 39–60.

Faverge, J.M., 1970. Plenary Session: the operator's reliability and safety in industry.Ergonomics 13 (3), 301–327.

Garrigou, A., Carballeda, G., Daniellou, F., 1998. The role of ‘know-how' inmaintenance activities and reliability in a high-risk process control plant.Appl. Ergonom. 29 (2), 127–131.

Grusenmeyer C., 2002. Interactions maintenance–exploitation et sécurité. Etudeexploratoire. Cahiers de Notes Documentaires Hygiène et Sécurité du Travail186, 53–66.

Grusenmeyer C., 2005. Les accidents liés à la maintenance. Etude bibliographique.Les Notes Scientifiques et Techniques de l'INRS 248, 68 p.

Grusenmeyer, C., 2009. Les accidents liés à la sous-traitance. Exploitation de la basede données EPICEA. In: Héry, M. (Ed.), La Sous-Traitance Interne. EDP Sciences,Les Ulis, pp. 98–115.

Guignon, N., Niedhammer, I., Sandret, N., 2008. Les facteurs psychosociaux autravail. Une évaluation par le questionnaire de Karasek dans l'enquête SUMER.Docum. Médec. Travail 115, 389–398.

Hale, A.R., Heming, B.H.J., Smit, K., Rodenburg, F.G.Th., Van Leeuwen, N.D., 1998.Evaluating safety in the management of maintenance activities in the chemicalprocess industry. Safe. Sci. 28 (1), 21–44.

Herrera, I.A., Nordskag, A.O., Myhre, G., Halvorsen, K., 2009. Aviation safety andmaintenance under major organizational changes, investigating non-existingaccidents. Accid. Anal. Prevent. 41, 1155–1163.

Hobbs, A., Williamson, A., 2002. Unsafe acts and unsafe outcomes in aircraftmaintenance. Ergonomics 45 (12), 866–882.

Hobbs, A., Williamson, A., 2003. Associations between errors and contributingfactors in aircraft maintenance. Hum. Fact. 45 (2), 186–201.

Hobbs, A., Kanki, B.G., 2008. Patterns of error in confidential maintenance incidentsreports. The International Journal of Aviation Psychology 18 (1), 5–16.

Hollnagel, E., 2012. FRAM: the Functional Resonance Analyses Method: ModellingComplex Socio-Technical Systems. Ashgate, Burlington, 142 p.

INRS (French National Institute of Research and Safety), 2004. Interventiond'entreprises extérieures. Aide-mémoire pour la prévention des risques. Paris,INRS, ED 941, 72 p.

Karasek, R., Brisson, C., Kawakami, N., Houtman, I., Bongers, P., 1998. The job contentquestionnaire (JCQ): an instrument for internationally comparativeassessments of psychosocial job characteristics. J. Occup. Heath Psychol. 3(4), 322–355.

Kositts, L.M., Jones, K., 2011. Interruptions experienced by registered nurses workingin the emergency department. J. Emerg. Nurs. 37 (1), 3–8.

Largier A., 2008. Dispositif de gestion des compétences et logique métier. Socio-Logos 3, March 23 2008. http://socio-logos.revues.org/document1323.html.

Lind, S., 2008. Types and sources of fatal and severe non-fatal accidents in industrialmaintenance. Int. J. Ind. Ergonom. 38, 927–933.

Lind, S., Nenonen, S., 2008. Occupational risks in industrial maintenance. J. Qual.Mainten. Eng. 14 (2), 194–204.

Main B.W., Cloutier D.R., Manuele F.A., Bloswick D.S.,2002. Maintenance riskassessment. Risk assessment for maintenance activities. Preventing injuries before they happen. Michigan, Occupational Safety and HealthAdministration, Survey report.

Mariotti, F., 2005. Qui Gouverne L'entreprise En Réseau? Presses de la FondationNationale des Sciences Politiques, Domont, 261 p.

Mayhew, C., Quinlan, M., Ferris, R., 1997. The effects of subcontracting/outsourcingon occupational health and safety: survey evidence from four Australianindustries. Safe. Sci. 25, 163–178.

Nenonen, S., 2011. Fatal workplace accidents in outsourced operations in themanufacturing industry. Safe. Sci. 49, 1394–1403.

Perraudin C., Thévenot N., Valentin J., 2006. Sous-traiter ou embaucher? Uneanalyse empirique des comportements de substitution des entreprises del'industrie en France entre 1984 et 2003. Document de Travail du Centred'Etudes de l'Emploi 78, 34 p.

Quinlan, M., Hampson, I., Gregson, S., 2013. Outsourcing and offshoring aircraftmaintenance in the US: implications for safety. Safe. Sci. 57, 283–292.

Rasmussen, J., 1997. Risk management in a dynamic society: a modelling problem.Safe. Sci. 27 (2–3), 183–213.

Ray, P.S., Batson, R.G., Weems, W.H., Wan, Q., Sorock, G.S., Matz, S., Cotnam, J., 2000.Impact of maintenance function on plant safety. Profess. Safe. (August),45–48.

Reason, J., Hobbs, A., 2003. Managing Maintenance Error. A Practical Guide. AshgatePublishing Company, Hampshire (UK), 183 p.

Reiman, T., 2011. Understanding maintenance work in safety-critical organisations.Managing the performance variability. Theoret. Issue Ergonom. Sci. 12 (4),339–366.

Rorive B. 2003. L'entreprise réseau: unicité de la formule, diversité des situations.XIVème Congrès de l'Association de Gestion des Ressources Humaines, Grenoble,20–22 novembre 2003.

Rosqvist, T., Laasko, K., Reunanen, M., 2009. Value-driven maintenance planning fora production plant. Reliab. Eng. Syst. Safe. 94, 97–110.

Siemieniuch, C.E., Sinclair, M.A., 2014. Extending systems ergonomics thinking toaccommodate the socio-technical issues of systems of systems. Appl. Ergonom.45 (1), 85–98.

Suzuki, T., Von Thaden, T.L., Geibel, W.D., 2008. Coordination and safety behaviors incommercial aircraft maintenance. Proceedings of the Human Factors andErgonomics Society, 52nd Annual Meeting, New York, pp. 89–93.

Sykes, E.R., 2011. Interruptions in the workplace: a case study to reduce their effects.Int. J. Inform. Manag. 31, 385–394.

Tazi D., 2008. Externalisation de la maintenance et ses impacts sur la sécurité dansles industries de procédés. Thèse en systèmes industriels. Toulouse, InstitutNational Polytechnique de Toulouse, Décembre 2008. 176 p.

Vidal-Gomel, C., 2007. Compétences pour gérer les risques professionnels: unexemple dans le domaine de la maintenance de systèmes électriques. TravailHumain 70 (2), 153–194.

Wilson, J.R., Carayon, P., 2014. Systems ergonomics: looking into the future. Editorialfor special issue on systems ergonomics/human factors. Appl. Ergonom. 45 (1),3–4.