Embed Size (px)
Citation preview
Journal of Hospital Infection (2008) 69, 101e113
Available online at www.sciencedirect.com
www.elsevierhealth.com/journals/jhin
REVIEW
Matching criteria in caseecontrol studies onpostoperative infections
M.E. Falagas a,b,*, E.G. Mourtzoukou a, F. Ntziora a,G. Peppas a, P.I. Rafailidis a
a Alfa Institute of Biomedical Sciences (AIBS), Athens, Greeceb Department of Medicine, Tufts University School of Medicine, Boston, Massachusetts, USA
Available online 10 March 2008
KEYWORDSSurgical-siteinfections;Postoperativeinfections; Caseecontrol study;Matching
* Corresponding author. Address: AlfTel.: þ30 694 611 0000; fax: þ30 210
E-mail address: [email protected]
0195-6701/$ - see front matter ª 200doi:10.1016/j.jhin.2008.01.013
Summary Matching is commonly used in caseecontrol studies to controlfor the effect of major confounding factors. We evaluated the available ev-idence from caseecontrol studies regarding postoperative infections toidentify how frequently matching was performed and with what specificvariables. We searched for relevant caseecontrol studies in PubMed untilAugust 2006 and further evaluated those that used individual matching be-tween cases and controls. We identified and evaluated 42 relevant studies.Age was used as a matching criterion in 27 of these 42 (64.3%) caseecontrolstudies. The specific type of surgical procedure was the second most com-monly used criterion in 17 of 42 studies (40.5%). Gender was used in 14/42studies (33.3%) as a matching criterion between case and control patients.The period at risk for development of surgical site and/or other postoper-ative infections, i.e. time from surgery to the diagnosis of infection, wasused in nine of 42 studies (21.4%), as was date of operation, and the pri-mary diagnosis that led the case and control patients to surgery. The samesurgeon or surgical team was used in seven studies (16.7%); matching ac-cording to the National Nosocomial Infection Surveillance system risk scorewas performed in five studies (11.9%). The findings of our analysis suggestthat various characteristics are used for matching in caseecontrol studiesof postoperative infections. A more consistent use of matching with thespecific type of surgical procedure may help in increasing the internalvalidity of a caseecontrol study in this field of clinical research.ª 2008 The Hospital Infection Society. Published by Elsevier Ltd. All rightsreserved.
a Institute of Biomedical Sciences (AIBS), 9 Neapoleos Street, 151 23 Athens, Marousi, Greece.683 9605.
8 The Hospital Infection Society. Published by Elsevier Ltd. All rights reserved.
102 M.E. Falagas et al.
Introduction Methods
Postoperative infections, including surgical siteinfections, are possible complications of all typesof surgical procedures that influence both morbid-ity and mortality.1,2 Delayed healing, disability,deformity, prolonged hospital stay, increasedcost of healthcare, and even death can be the con-sequences of such infections. Thus, it is importantto identify the risk factors associated with thedevelopment of postoperative infections in orderto apply appropriate preventive and therapeuticstrategies to decrease the incidence and conse-quences of these infections.3
Although postoperative infections constituteone of the most important problems related toinfection in clinical practice and there are compel-ling reasons to further decrease their incidence,the proportion of patients who develop such in-fections is relatively small from the point of view ofthe selection of the most appropriate and cost-effective study design to study risk factors.4 For ex-ample about 2% of patients after coronary arterybypass grafting develop postoperative infectionsand this proportion is even smaller in patients un-dergoing other types of surgical operations.5
Thus, caseecontrol studies have been used exten-sively for the investigation of factors that may po-tentially predispose to the development ofpostoperative infections. This study design is com-monly used when the occurrence of the studiedoutcome is relatively rare.6 It should be emphas-ised that a retrospective or prospective cohortstudy of postoperative infections would involvedata collection and analysis for all patients who un-derwent a surgical operation, irrespective ofwhether a patient developed or not such an infec-tion. In contrast, a caseecontrol study would in-volve data collection and analysis for therelatively small subset of patients who developeda postoperative infection and a small subset of ran-domly selected patients or those matched for se-lected characteristics who did not develop suchan infection (usually the number of control patientsis equal to or double that of the number of cases).7
There is a dearth of studies focusing on theimportant clinical research decision regarding thecharacteristics that should be used for the match-ing of cases and controls in studies of postopera-tive infections. We sought to identify the matchingcriteria that focus on the development of surgicalsite or other postoperative infections and suggestrecommendations for the selection of the appro-priate control subjects.
Literature search and study selection
We searched in the PubMed database (the periodexamined was 1950e2006) to identify caseecon-trol studies that used matching methodology andfocused on the development of surgical site orother postoperative infections. The key wordsused were ‘caseecontrol study’ combined with‘matched’ and with ‘surgical site infection orpostoperative infection’. Studies were limited tothose that were written in English, referred tohumans, and used patients who experienced a sur-gical site infection (SSI) or postoperative infectionin an area other than the surgical site (POI) aftera surgical procedure (cases) and patients whounderwent a surgical procedure but did not de-velop SSI or other POI (controls).
Data extraction
Data regarding the first author, the year of publi-cation, study type, hospital setting, focus of study,definition and number of case patients and con-trols, and the matching criteria used were ex-tracted by two of the authors (E.G.M. and F.N.)and tabulated (Table I).
Results
The abstracts of the initially identified articleswere reviewed in order to select the relevantstudies for more detailed review. Out of the 119articles, eleven were excluded because the focuswas not relevant to our study, six because theywere written in language other than English, twobecause they were not caseecontrol studies, andone because it gave the same data with another.The complete texts of the remaining originalarticles were read. Fifty-seven were excluded:forty-six that did not use patients who developedpostoperative infection as case patients, four thatused patients who developed postoperative in-fection as controls, three of them that did notuse matching methodology, one that used match-ing methodology only for a subgroup of patientswho developed post-partum fever, two that hadstudied patients whose infections were the resultof the immunosuppression induced for transplantoperation, and one in which only half of the casepatients were operated; this left forty-two articlesrelevant to our study.
Table I Matching criteria in caseecontrol studies focusing on surgical site and other postoperative infections
Study Design, hospital setting Focus of study Patients Controls Matching
Surgical site infectionsFarinas et al. (2006)8 Retrospective casee
control study. HospitalUniversitario Marquesde Cantabria, Spain
Risk factors of PVE 81 patients withdefinite or possibleinfective endocarditisafter PVR
162 controls whounderwent PVR and didnot develop PVE
Gender, age atoperation (� 5 years),surgery of one or morevalves in the sameanatomic position
Wu et al. (2006)9 Retrospective caseecontrol study.Kaohsiung Chang GungMemorial Hospital,Taiwan
Risk of endophthalmitisafter cataract surgery
12 patients whodevelopedpostoperativeendophthalmitis
120 patients who didnot develop underwentsurgery and did notdevelopendophthalmitis
Age, gender
Kasatpibal et al.(2005)10
Prospective caseecontrol study.SongklanagarindHospital, Thailand
Extra charge and excesspostoperativehospitalisationattributable to SSI in sixsurgical operativeprocedures(appendectomy,herniorrhaphy,mastectomy,cholecystectomy,colectomy, andcraniotomy)
140 patients with SSIafter major operation
140 patients who didnot develop SSI aftermajor operation
Final diagnosis,operative procedure,American Society ofAnesthesiologists score.Additional matchingcriteria if more thanone control found:emergency operation,preoperative stay,wound class, age.
Griffiths et al. (2005)11 Retrospective caseecontrol study. RoyalAlexandra Hospital,Edmonton, Canada
The incidence ofpostoperative SSIfollowing elective CS
124 women with SSIafter elective CS
218 women without SSIafter elective CS
Type of surgery, date ofsurgery (selection ofwomen who underwentCS immediately beforeand after the case)
Upton et al. (2005)12 Retrospective caseecontrol study. GreenLane Hospital, NewZealand
The additional costattributable toStaphylococcus aureusPSM followingsternotomy for cardiacsurgery
9 patients with S.aureus PSM
9 patients without S.aureus PSM
Gender, age, type ofsurgical procedure,presence of diabetesmellitus
Rey et al. (2005)13 Retrospective caseecontrol study. BonSecours Cottage HealthServices, USA
SSI following clean andclean-contaminatedambulatory surgery(breast biopsy)
15 patients who hadambulatory breastbiopsy and developedSSI
45 patients who hadambulatory breastbiopsy and did notdevelop SSI
Date of operation,procedure
(continued on next page)
Case
eco
ntro
lin
fectio
nstu
die
s103
Table I (continued)
Study Design, hospital setting Focus of study Patients Controls Matching
Nateghian et al.(2004)14
Caseecontrol study.Stollery Children’sHospital, Tehran, Iran
Risk factors for SSIfollowing open-heartsurgery in paediatricheart patients
38 children with SSIafter cardiac surgery
3 ildren without SSIa cardiac surgery
NNIS system risk scores,age (newborns <30days, infants 1 to 12months, toddlers, >1year and <4 years, andchildren >4 years)
Minnema et al. (2004)15 Prospective caseecontrol study.Sunnybrook andWomen’s CollegeHealth SciencesCentre, Canada
Risk factors associatedwith the developmentof SSI following TKA
22 patients with an SSIafter TKA
6 tients without anS fter TKA
Date of surgery
Talbot et al. (2004)16 Retrospective caseecontrol study.Vanderbilt UniversitySchool of Medicine,Tennessee, USA
Risk factors for thedevelopment of adultpost-sternotomy SSI
38 cases with post-sternotomy SSI
1 ithout post-s otomy SSI
Age (�5 years), date ofprocedure
Abboud et al. (2004)17 Retrospective caseecontrol study. DantePazzanese CardiologyInstitute, Brazil
Risk factors formediastinitis aftercardiac surgery
39 cases withmediastinitis aftercardiac surgery
7 ses withoutm iastinitis afterc iac surgery
Age (�10 years)
Wong et al. (2004)18 Retrospective caseecontrol study. NationalUniversity of Singapore
Risk factors of acuteendophthalmitis aftercataract extraction
34 patients with acuteendophthalmitis aftercataract surgery
1 ontrols withoute phthalmitis afterc ract surgery
Date of operation
Labbe et al. (2003)19 Retrospective caseecontrol study. MontrealChildren’s Hospital,McGill UniversityHealth Centre, Canada
Rates of SSIs afterspinal surgery and riskfactors associated withinfection
13 patients whodeveloped an SSI afterspinal surgery
2 tients who did notd lop an SSI afters l surgery
Presence or absence ofmyelodysplasia
Schollin-Borg et al.(2003)20
Retrospective caseecontrol study. UppsalaUniversity Hospital,Sweden
Septic arthritis afterACLR: presentation,outcome and cause
10 patients whodeveloped septicarthritis after ACLR
1 tients who did notd lop septic arthritisa ACLR
Age, gender,preoperative level ofactivity, date ofsurgery, surgeon, typeof graft, concomitantprocedures performedat the time ofreconstruction
104M
.E.
Fala
gas
et
al.
8 chfter
6 paSI a
14 wtern
8 caed
ard
02 cndoata6 paevepina
0 paevefter
Whitehouse et al.(2002)21
Prospective pairwise-matched (1:1) caseecontrol study. DukeUniversity MedicalCenter and DurhamRegional Hospital,North Carolina, USA
The impact oforthopaedic SSIs onquality of life, length ofhospitalisation, andcost
59 patients whounderwent orthopaedicsurgery and developedan SSI
59 patients whounderwent orthopaedicsurgery and did notdevelop an SSI
Type of surgicalprocedure, NNIS riskindex, age (�5 years),surgeon
Hollenbeak et al.(2002)22
Reanalysis of aprospective caseecontrol study, usingtwo-stage methodology(unmatched andmatched). Penn StateCollege of Medicine,USA
Selection bias inestimates of theattributable cost ofdeep chest SSIfollowing CABG surgery
41 patients who had anSSI following CABG andCABG and valve surgery
41 patients who did nothave an SSI followingCABG and CABG andvalve surgery
Age, gender, diabetesmellitus, renalinsufficiency, length ofsurgical procedure
Reichert et al. (2002)23 Prospective caseecontrol study. HospitalSao Paulo, Brazil
Incidence ofpostcraniotomymeningitis
50 patients with post-craniotomy meningitis
50 patients withoutpost-craniotomymeningitisa
Age (�5 years), surgicaldiagnosis, surgicalprocedure
Jenney et al. (2001)24 Retrospective caseecontrol study. AlfredHospital, PrahranAustralia
The financial impact ofSSI following CABGsurgery
108 patients with SSIafter CABG surgery
108 patients without SSIafter CABG surgery
Gender, age (�5 years),risk index (CDC, NNISsystem) and number ofprincipal comorbidities(non-cardiac or cardiacconditions)
Jampel et al. (2001)25 Retrospective caseecontrol study.Multicentre, Baltimore
Risk factors for late-onset infectionassociated with GFS
131 patients with late-onset infection afterGFS
500 patients whounderwent GFS and didnot develop late-onsetinfection
Surgeon
Roy et al. (2000)26 Prospective caseecontrol study.University of IowaHospitals and Clinics,Iowa City, USA
Determines whetherthe NNIS system riskindex adequatelystratifies a populationof cardiothoracicsurgery patients by therisk of developing SSI
201 patients with SSIsafter cardiothoracicoperative procedures
398 patients withoutSSIs aftercardiothoracicoperative procedures
Age (�5 years), gender,type of procedure
Brewer et al. (2000)27 Retrospective caseecontrol study.University of TennesseeMedical Centre, USA
Risk factor analysis forbreast cellulitiscomplicating breastconservation therapy
17 patients with breastcancer who developedbreast cellulitis afterpartial mastectomies
34 patients with breastcancer who did notdevelop cellulitis afterpartial mastectomy
Primary surgeon
(continued on next page)
Case
eco
ntro
lin
fectio
nstu
die
s105
Table I (continued)
Study Design, hospital setting Focus of study Patients Controls Matching
Soltau et al. (2000)28 Retrospective caseecontrol study. EyeInstitutes in USA
Risk factors for bleb-related ocularinfections afterglaucoma filteringsurgery
55 eyes of 55 patientswith glaucoma filteringbleb-related infections
55 eyes of 55 patientswho underwentglaucoma filteringsurgery and did notdevelop infection
Surgeon, type ofglaucoma surgery, typeof antifibrotic agent
Kirkland et al. (1999)29 Retrospective caseecontrol study. TheDurham RegionalHospital, NorthCarolina, USA
Mortality, morbidity,and costs attributableto SSIs in the 1990s
255 patients with SSI 255 patients without SSI Age (�10 years), NNISprocedure code, NNISrisk index, surgeon
Klekamp et al. (1999)30 Retrospective caseecontrol study.Vanderbilt UniversitySchool of Medicine,Tennessee, USA
Risk factors associatedwith methicillin-resistantstaphylococcal woundinfection after spinalsurgery
35 patients whodeveloped meticillin-resistantstaphylococcal woundinfection after spinalsurgery
35 patients who did notdevelop infection afterspinal surgery
Indication for initialsurgery
Berbari et al. (1998)31 Retrospective caseecontrol study.Department ofOrthopedics, MayoClinic, USA
Risk factors for thedevelopment ofprosthetic jointinfection
462 patients withprosthetic hip or kneejoint infection
462 patients whounderwent total hip orknee arthroplasty anddid not developprosthetic jointinfection
Age, gender, prosthesislocation, date ofimplantation, length offollow-up (equal to orgreater than the timefrom prosthesisimplantation todiagnosis of prostheticjoint infection in casepatient)
Bertin et al. (1998)32 Retrospective caseecontrol study.Cleveland ClinicFoundation, USA
Determinants of SSIsafter breast surgery
18 patients who had abreast procedure anddeveloped SSI
37 who had a breastprocedure and did notdevelop SSI
Surgical procedure
Munn et al. (1998)33 Retrospective caseecontrol study.University of Alabamaat Birmingham, USA
Intraoperativehypothermia and post-caesarean woundinfection
18 women whounderwent caesareandelivery and developedwound infection
18 women whounderwent caesareandelivery and did notdevelop woundinfection
Age (<21 or �21 years),weight, presence ofgestationalhypertension, durationof surgery
Fridkin et al. (1996)34 Retrospective caseecontrol study, at anambulatory surgicalcentre
Acremonium kilienseendophthalmitis aftercataract extraction
4 patients whodeveloped A. kilienseendophthalmitis aftercataract extraction
16 patients whounderwent cataractsurgery and did notdevelopendophthalmitis
Day, time and type ofsurgery
106M
.E.
Fala
gas
et
al.
Ehrenkranz et al.(1995)35
Caseecontrol study.Florida Consortium forInfection Control,University of Miami,USA
Potential consequencesof habitual over-diagnosis of SSIs
18 patients with SSIs(12 presumptive and 6documented) afterlaminectomy
18 without SSI afterlaminectomy
Surgeon, indication foroperation
Poulsen et al. (1995)36 Retrospective caseecontrol study.Frederiksberg Hospital,Denmark
Survival of patientswith surgical woundinfection
291 patients with apostoperative woundinfection
291 patients whounderwent a surgicalintervention and didnot develop woundinfection
Type of operation, age,gender
Libombo et al. (1995)37 Retrospective caseecontrol study. MaputoCentral Hospital,Mozambique
Post-caesareanendometritis-myometritis (PCEM) inMozambique
49 women with PCEM 47 women without signsof PCEM aftercaesarean section
Age (�5 years), parity(0, 1e3, >4), days postpartum (�1 day)
Bryan et al. (1992)38 Retrospective caseecontrol study.University Hospital ofWales, UK
Risk factors andoutcome of mediansternotomy wounddehiscence (MSD)
44 patients whounderwent open heartsurgery and developedMSD
88 patients whounderwent open heartsurgery and did notdevelop MSD
Date of operation
Kappstein et al.(1992)39
Prospective caseecontrol study.University Hospital ofFreiburg, Germany
Prolongation of hospitalstay due topostoperative woundinfections followingcardiac surgery
22 patients withpostoperative woundinfection
132 patients withoutpostoperative woundinfection
Surgical procedure, age(�10 years), durationof hospital stay (atleast as long as timeinterval until infectionof the case patient)
Boyce et al. (1990)40 Retrospective caseecontrol study. MiriamHospital, Rhode Island,USA
Effect of surgicalwound infectionfollowing open heartsurgery on hospitalcosts andreimbursementpatterns
15 patients withsurgical woundinfection after openheart surgery
15 patients withoutsurgical woundinfection after openheart surgery
Diagnosis-relatedgroup, age (�7 years),gender, urgency ofsurgery, type of cardiacsurgery
Other postoperative infectionsAskarian and Gooran(2003)41
Prospective caseecontrol study. ShirazMedical School, Iran
The extra hospital stayattributable tonosocomial infections(urinary tract infection,surgical site infection,bloodstream infection,pneumonia) forpatients undergoingsurgery
69 patients withnosocomial infectionsafter surgery
69 patients who did notdevelop nosocomialinfections after surgery
Age (�7 years), gender,type of operation,elective or emergencyprocedure, NNIS riskindex category, lengthof hospital stay (fromthe time of operationto the time of infectiondiagnosis in the casepatients group)
(continued on next page)
Case
eco
ntro
lin
fectio
nstu
die
s107
Table I (continued)
Study Design, hospital setting Focus of study Patients Controls Matching
Husain et al. (2003)42 Prospective caseecontrol study. Six livertransplant (LT) centresin North America andEurope
Risk factors for invasivecandidiasis in LTrecipients
35 LT recipients withinvasive candidiasis
69 LT recipientswithout invasivecandidiasis
35 controls werematched for duration ofhospitalisation and theother 34 for duration ofbroad-spectrumantibiotics to the timeof diagnosis
Chakrabarti et al.(2003)43
Retrospective caseecontrol study. Twotertiary-care hospitalsin New York, USA
Prolonged candidaemiain infants followingsurgery for complexcongenital heartdisease (CCHD)
6 infants with CCHDwho underwent cardiacsurgery and developedcandidaemia
6 infants with CCHDwho underwent cardiacsurgery and did notdevelop candidaemia
Age (�2 months),congenital heartdisease lesions,postoperative hospitalstay comparable withthe time from surgeryto onset ofcandidaemia in thecase patient
Nieto-Rodriguez et al.(1996)44
Time-matched caseecontrol study,department of Surgery,University ofPittsburgh, USA
Factors associated withcandidaemia andcandidaemia-relateddeath among adult livertransplant recipients
26 liver transplantrecipients withcandidaemia
52 liver transplantrecipients withoutcandidaemia
Duration of follow-upto the day of the firstblood culture
Schwenzer et al.(1994)45
Retrospective caseecontrol study. ICU,University of VirginiaHealth SciencesCentre, USA
Prediction ofbacteraemia in surgicalICU
24 surgical ICU patientswith one or morepositive blood cultures
48 surgical ICU patientswith negative bloodcultures
Diagnosis, procedure,age
Chwals et al. (1994)46 Retrospective caseecontrol study. Neonataland paediatric ICUs,Wake Forest UniversityMedical Centre, NC,USA
Detection ofpostoperative bacterialinfection in infants withthe use of metabolicstress monitoring
13 infants withbacterial infection(sepsis) after majorabdominal or thoracicoperative procedure
27 infants in whompostoperative bacterialinfection did notdevelop
Diagnosis, type ofoperative procedure,time interval from dateof operation, postnatalage, gestational age
Coello et al. (1993)47 Retrospective caseecontrol study.Wycombe GeneralHospital, UK
The cost of hospitalacquired infection insurgical patients
67 surgical patientswith hospital acquiredinfection(gynaecological,urological, orthopaedicor general surgery)
67 uninfected surgicalcontrols
First operativeprocedure, primarydiagnosis, age (�10years), gender, surgicalservice
108M
.E.
Fala
gas
et
al.
Lapchik et al. (1992)48 Retrospective caseecontrol study. HospitalSao Paulo, Brazil
Risk factors fornosocomial urinarytract and postoperativewound infections inrenal transplantpatients
36 renal transplantrecipients whoacquired urinary tractand/or wound infection
36 renal transplantrecipients withoutinfection
Age (�10 years),gender, risk period(interval betweenhospitalisation andinfection diagnosis)
Givens and Wenzel(1980)49
Retrospective caseecontrol study.University of VirginiaMedical Centre, USA
Excess morbidity andcosts in catheter-associated UTIs insurgical patients (bowelresection, coronaryartery bypass graft,total hip replacement,laminectomy)
24 patients withpostoperative Foleycatheter-associatedUTIs
24 catheterisedpatients whounderwent surgery anddid not develop UTI
Age (�6 years), gender,operative procedure,clinical serviceperforming theoperation, duration ofcatheter indwellingbetween catheterinsertion and UTIdiagnosis in casepatients
ACLR, anterior cruciate ligament reconstruction; APACHE II, Acute Physiology and Chronic Health Evaluation II; ASA, American Society of Anesthesiologists; CABG, coronary arterybypass graft; CDC, Centers of Disease Control and Prevention; COPD, chronic obstructive pulmonary disease; CS, caesarean section; 3GCs: third-generation cephalosporins; GFS, glau-coma filtration surgery; HIV, human immunodeficiency virus; IABP, intra-aortic balloon pump; ICU, intensive care unit; NNIS, National Nosocomial Infection Surveillance; NWHA, NewYork Heart Association; PSM, poststernotomy mediastinitis; PVE, prosthetic valve endocarditis; PVR, prosthetic valve replacement; SSI, surgical site infection; TKA, total knee arthro-plasty; UTI, urinary tract infection; VRE, vancomycin-resistant enterococcus.
a Controls had an 88.8% successful matching rate in this study.
Case
eco
ntro
lin
fectio
nstu
die
s109
110 M.E. Falagas et al.
Of the total of forty-two studies that we in-cluded in our analysis, 12 focused on cardiacsurgery, five on cataract or glaucoma filtrationsurgery, five on orthopaedic surgery, two on anysurgical procedure performed in a hospital, threeon caesarean section, three on breast surgery orbiopsy, two on spinal surgery, one on generalsurgery, one on craniotomy, three on renal or livertransplantations, and five on miscellaneous surgi-cal procedures.8e49
Age was used as a matching criterion in 27 of thetotal 42 (64.3%) caseecontrol studies. Eighteenarticles gave information on the age range (cen-tred on the age of cases) required for the selectionof controls. Specifically, in seven of them the agerange was �5 years, in five �10 years, in two �7years, in one �6 years, in one <21 or >21 years, inone �2 months, and in one the age groups were<30 days, 1e12 months, 1e4 years, and >4 years.
The specific type of surgical procedure was thesecond most commonly used criterion, in 17 of 42studies (40.5%). Gender was used in 14 of 42studies (33.3%) as a matching criterion betweencase and control patients. The period at risk fordevelopment of SSIs and/or POIs, that is, the timefrom surgery to the diagnosis of infection, wasused in nine of 42 studies (21.4%), as was the dateof operation, and the primary diagnosis that ledthe case and control patients to surgery. The samesurgeon or surgical team was used in seven studies(16.7%), while matching according to the NationalNosocomial Infection Surveillance system riskscore was performed in five studies (11.9%).
Due to the variety of surgical procedures stud-ied in the reviewed articles, there was also a widerange of matching criteria, other than thosepresented above, relevant to each procedure,that were occasionally used: urgency of surgeryin three of 42 studies (7%), type of surgery inthree, length of surgery in two (4.7%), surgicalservice in two, presence of diabetes in two, typeof antifibrotic agent (for glaucoma surgery) in onestudy (2.4%), preoperative level of activity (fororthopaedic surgery) in one, type of graft (inorthopaedic surgery) in one, concomitant proced-ures performed at the time of surgery in one,weight of the patients in one, gestational hyper-tension in one, parity (0.1e3, or >4) in one, dayspost-partum in one, gestational age in one, theAmerican Society of Anesthesiologists score in one,preoperative stay in one, presence of myelodys-plasia in one, renal insufficiency in one,comorbidities (defined as non-cardiac or cardiacenon-ischaemic) in one, and duration of broad-spectrum antibiotics to the time of diagnosis inone.
In the total of the 42 caseecontrol studies, 10studies used one matching criterion, nine studiesused two matching criteria, 11 studies used threematching criteria, four studies used four matchingcriteria, five studies used five matching criteria,two studies used six, and one study used eightmatching criteria. In one of the studies that usedthree matching criteria, four additional matchingcriteria were applied if more than one potentialcontrol existed.
Discussion
The main finding of our analysis is that age is themost frequently used matching criterion in thecaseecontrol studies that focus on surgical siteand other postoperative infections. In most of theevaluated studies more than one matching cri-terion was used. However, it is noteworthy thatthe specific type of surgical procedure was used inonly about 40% of the reviewed caseecontrolstudies of postoperative infections. We believethat a more consistent use of matching with thespecific type of surgical procedure may help inincreasing the internal validity of a caseecontrolstudy focusing on risk factors for development ofpostoperative infections.
The selection of an appropriate comparisongroup is perhaps one of the most difficult andcritical issues in the design of a caseecontrolstudy.6,50 The control group must be comparableto the source population of the cases and any ex-clusions or restrictions made in the identificationof cases should apply equally to the controls.Matching is a technique used to control confound-ing in the design of analytic epidemiological stud-ies, and has been widely used over the years.Despite the substantial scientific limitations andlogistical difficulties, as the inability to directlyevaluate the effect of a factor that has beenmatched on the risk of the outcome, or disadvan-tages in terms of time, money and loss of potentialstudy subjects, there are some circumstances inwhich matching is desirable and occasionallyeven necessary. For example, matching is neces-sary for any factors for which there would other-wise be insufficient overlap between the studygroups (e.g. neighbourhood and sibship if such fac-tors are considered vital for the specific studiedresearch question), or when the case series isvery small.51,52
Caseecontrol studies that involve matchingmethodology have been used as a tool to identifypreoperative, intraoperative and postoperativerisk factors that are associated with the
Caseecontrol infection studies 111
development of postoperative infections in surgi-cal patients, to study the burden that postopera-tive infections add on morbidity and mortality andthe increased cost of care they create, and tosuggest preventive measures for the reduction ofthese infections.
In general the application of various matchingcriteria in caseecontrol studies may introducebiases in the analysis.53 On the other hand, incaseecontrol studies regarding postoperative in-fections it may be necessary to match for variousimportant criteria to avoid misinterpretation ofthe study findings. In our review, age was foundto be the most commonly used criterion as a char-acteristic for matching between cases and con-trols. Age is one of the most important factorsdetermining morbidity and mortality in general.For instance in the Acute Physiological Assessmentand Chronic Health Evaluation (APACHE) II scoresystem the predicted death rate for intensivecare unit patients is 2.9% for patients aged <44years, whereas it is 6.7% for those aged >75 years.Thus, it does not come as a surprise that most ofthe caseecontrol studies in postoperative infec-tions used age as a matching criterion.
The role of gender on the development ofpostoperative infections has not been evaluatedfully and only scarce data are available.54 Third infrequency of use as a characteristic for matchingbetween cases and controls in the evaluatedcaseecontrol studies was the period at risk for de-velopment of SSIs and/or POIs (time from surgeryto the diagnosis of infection) as well as the primarydiagnosis that led the case and control patients tosurgery. Length of hospital stay is known to be as-sociated with colonisation and hospital infections.The longer the patient’s stay in hospital, thegreater is the probability that the patient will ac-quire an infection. Thus, the use of length of hos-pital stay seems to be substantiated as a matchingcriterion.55 Also, matching for the primary diagno-sis is useful for an informative analysis of the com-parison of the distribution of various parameters inthe case and control groups. In addition, to betterevaluate risk factors for development of postoper-ative infections, ideally the surgeon/team shouldhave operated on cases and controls.
The use of healthy controls is logical and iscommonly used in caseecontrol studies in otherfields, e.g. antimicrobial resistance.6 In contrast,comparison of patients that underwent an opera-tion with healthy controls probably introduces sig-nificant bias. The use of the National NosocomialInfection Surveillance (NNIS) system risk score asa matching criterion offers probably the most pre-cise view of how close are the cases and controls in
terms of comparability.26 On the other hand, thepossible bias introduced by adjusting for the pa-rameters defining the NNIS risk score should alsobe examined. For example, surgical risk factorsaccording to the NNIS risk score include woundclass, American Society of Anesthesiologists classi-fication, and whether surgery was an emergency.Matching for these parameters may make the twostudied groups more comparable but the re-searcher will lose the opportunity to directly eval-uate and detect whether these particularcharacteristics are associated with the develop-ment of postoperative infections.
In general, the use of more than one matchingcriterion in caseecontrol studies may be limited bypractical difficulties in identifying appropriatecontrols for inclusion in the study. However, itseems that for caseecontrol studies focusing onpostoperative patients, matching for several cri-teria may be necessary to draw correct conclu-sions. Thus, in this field of clinical research the useof multiple matching criteria (age, time at risk fordevelopment of infection, same surgeon/surgicalteam, primary diagnosis, and NNIS risk score) isprobably justified. In addition, the need for a moreconsistent use of matching of cases and controlswith the specific type of surgical procedure cannotbe overemphasised because it may help increasethe internal validity of a caseecontrol study on riskfactors for postoperative infections.
Conflict of interest statementNone declared.
Funding sourcesNone.
References
1. Barie PS, Eachempati SR. Surgical site infections. Surg ClinNorth Am 2005;85:1115e1135.
2. NicholsRL.Postoperative infections intheageofdrug-resistantgram-positive bacteria. Am J Med 1998;104(5A):11Se16S.
3. Itani KM. Infections in the surgical patient: an update ontrends and treatment. Introduction. Am J Surg 2003;186(5A):1Se3S.
4. Falagas ME, Bliziotis IA. Cost-effectiveness of differentstudy designs: an overlooked parameter in clinical research.J Clin Epidemiol 2007;60:215e216.
5. Falagas ME, Rosmarakis ES, Rellos K, Michalopoulos A,Samonis G, Prapas SN. Microbiologically documented noso-comial infections after coronary artery bypass surgerywithout cardiopulmonary bypass. J Thorac Cardiovasc Surg2006;132:481e490.
6. Harris AD, Karchmer TB, Carmeli Y, Samore MH. Methodo-logical principles of caseecontrol studies that analysed riskfactors for antibiotic resistance: a systematic review. ClinInfect Dis 2001;32:1055e1061.
112 M.E. Falagas et al.
7. Michalopoulos A, Geroulanos S, Rosmarakis ES, Falagas ME.Frequency, characteristics, and predictors of microbiologi-cally documented nosocomial infections after cardiacsurgery. Eur J Cardiothorac Surg 2006;29:456e460.
8. Farinas MC, Perez-Vazquez A, Farinas-Alvarez C, et al. Riskfactors of prosthetic valve endocarditis: a caseecontrolstudy. Ann Thorac Surg 2006;81:1284e1290.
9. Wu PC, Li M, Chang SJ, et al. Risk of endophthalmitis aftercataract surgery using different protocols for povidone-iodine preoperative disinfection. J Ocul Pharmacol Ther2006;22:54e61.
10. Kasatpibal N, Thongpiyapoom S, Narong MN, Suwalak N,Jamulitrat S. Extra charge and extra length of postoperativestay attributable to surgical site infection in six selectedoperations. J Med Assoc Thai 2005;88:1083e1091.
11. Griffiths J, Demianczuk N, Cordoviz M, Joffe AM. Surgical siteinfection following elective Caesarian section: a caseecontrolstudy of postdischarge surveillance. J Obstet Gynaecol Can2005;27:340e344.
12. Upton A, Smith P, Roberts S. Excess cost associated withStaphylococcus aureus poststernotomy mediastinitis. N ZMed J 2005;118:U1316.
13. Rey JE, Gardner SM, Cushing RD. Determinants of surgicalsite infection after breast biopsy. Am J Infect Control2005;33:126e129.
14. Nateghian A, Taylor G, Robinson JL. Risk factors forsurgical site infections following open-heart surgery ina Canadian pediatric population. Am J Infect Control2004;32:397e401.
15. Minnema B, Vearncombe M, Augustin A, Gollish J, Simor AE.Risk factors for surgical-site infection following primarytotal knee arthroplasty. Infect Control Hosp Epidemiol2004;25:477e480.
16. Talbot TR, D’Agata EM, Brinsko V, Lee B, Speroff T,Schaffner W. Perioperative blood transfusion is predictiveof poststernotomy surgical site infection: marker formorbidity or true immunosuppressant? Clin Infect Dis2004;38:1378e1382.
17. Abboud CS, Wey SB, Baltar VT. Risk factors for mediastinitisafter cardiac surgery. Ann Thorac Surg 2004;77:676e683.
18. Wong TY, Chee SP. Risk factors of acute endophthalmitisafter cataract extraction: a caseecontrol study in Asianeyes. Br J Ophthalmol 2004;88:29e31.
19. Labbe AC, Demers AM, Rodrigues R, Arlet V, Tanguay K,Moore DL. Surgical-site infection following spinal fusion:a caseecontrol study in a children’s hospital. Infect ControlHosp Epidemiol 2003;24:591e595.
20. Schollin-Borg M, Michaelsson K, Rahme H. Presentation,outcome, and cause of septic arthritis after anteriorcruciate ligament reconstruction: a case control study.Arthroscopy 2003;19:941e947.
21. Whitehouse JD, Friedman ND, Kirkland KB, Richardson WJ,Sexton DJ. The impact of surgical-site infections followingorthopedic surgery at a community hospital and a universityhospital: adverse quality of life, excess length of stay, andextra cost. Infect Control Hosp Epidemiol 2002;23:183e189.
22. Hollenbeak CS, Murphy D, Dunagan WC, Fraser VJ. Nonran-dom selection and the attributable cost of surgical-siteinfections. Infect Control Hosp Epidemiol 2002;23:177e182.
23. Reichert MC, Medeiros EA, Ferraz FA. Hospital-acquiredmeningitis in patients undergoing craniotomy: incidence,evolution, and risk factors. Am J Infect Control 2002;30:158e164.
24. Jenney AW, Harrington GA, Russo PL, Spelman DW. Cost ofsurgical site infections following coronary artery bypasssurgery. Aust N Z J Surg 2001;71:662e664.
25. Jampel HD, Quigley HA, Kerrigan-Baumrind LA, Melia BM,Friedman D, Barron Y. Glaucoma surgical outcomes studygroup. Risk factors for late-onset infection followingglaucoma filtration surgery. Arch Ophthalmol 2001;119:1001e1008.
26. Roy MC, Herwaldt LA, Embrey R, Kuhns K, Wenzel RP,Perl TM. Does the centers for disease control’s NNIS systemrisk index stratify patients undergoing cardiothoracic oper-ations by their risk of surgical-site infection? Infect ControlHosp Epidemiol 2000;21:186e190.
27. Brewer VH, Hahn KA, Rohrbach BW, Bell JL, Baddour LM.Risk factor analysis for breast cellulitis complicating breastconservation therapy. Clin Infect Dis 2000;31:654e659.
28. Soltau JB, Rothman RF, Budenz DL, et al. Risk factors forglaucoma filtering bleb infections. Arch Ophthalmol 2000;118:338e342.
29. Kirkland KB, Briggs JP, Trivette SL, Wilkinson WE, Sexton DJ.The impact of surgical-site infections in the 1990s: attribut-able mortality, excess length of hospitalization, and extracosts. Infect Control Hosp Epidemiol 1999;20:725e730.
30. Klekamp J, Spengler DM, McNamara MJ, Haas DW. Riskfactors associated with methicillin-resistant staphylococcalwound infection after spinal surgery. J Spinal Disord 1999;12:187e191.
31. Berbari EF, Hanssen AD, Duffy MC, et al. Risk factors forprosthetic joint infection: caseecontrol study. Clin InfectDis 1998;27:1247e1254.
32. Bertin ML, Crowe J, Gordon SM. Determinants of surgicalsite infection after breast surgery. Am J Infect Control1998;26:61e65.
33. Munn MB, Rouse DJ, Owen J. Intraoperative hypothermiaand post-cesarean wound infection. Obstet Gynecol 1998;91:582e584.
34. Fridkin SK, Kremer FB, Bland LA, Padhye A, McNeil MM,Jarvis WR. Acremonium kiliense endophthalmitis thatoccurred after cataract extraction in an ambulatory surgicalcenter and was traced to an environmental reservoir. ClinInfect Dis 1996;22:222e227.
35. Ehrenkranz NJ, Richter EI, Phillips PM, Shultz JM. An appar-ent excess of operative site infections: analyses to evaluatefalse-positive diagnoses. Infect Control Hosp Epidemiol1995;16:712e716.
36. Poulsen KB, Wachmann CH, Bremmelgaard A, Sorensen AI,Raahave D, Petersen JV. Survival of patients with surgicalwound infection: a caseecontrol study of common surgicalinterventions. Br J Surg 1995;82:208e209.
37. Libombo A, Folgosa E, Bergstrom S. A caseecontrol study onpost-caesarean endometritis-myometritis in Mozambique.Gynecol Obstet Invest 1995;39:180e185.
38. Bryan AJ, Lamarra M, Angelini GD, West RR,Breckenridge IM. Median sternotomy wound dehiscence:a retrospective case control study of risk factors andoutcome. J R Coll Surg Edinb 1992;37:305e308.
39. Kappstein I, Schulgen G, Fraedrich G, Schlosser V,Schumacher M, Daschner FD. Added hospital stay due towound infections following cardiac surgery. Thorac Cardio-vasc Surg 1992;40:148e151.
40. Boyce JM, Potter-Bynoe G, Dziobek L. Hospital reim-bursement patterns among patients with surgical woundinfections following open heart surgery. Infect ControlHosp Epidemiol 1990;11:89e93.
41. Askarian M, Gooran NR. National nosocomial infectionsurveillance system-based study in Iran: additional hospitalstay attributable to nosocomial infections. Am J InfectControl 2003;31:465e468.
42. Husain S, Tollemar J, Dominguez EA, et al. Changes in thespectrum and risk factors for invasive candidiasis in liver
Caseecontrol infection studies 113
transplant recipients: prospective, multicenter, case-controlled study. Transplantation 2003;75:2023e2029.
43. Chakrabarti C, Sood SK, Parnell V, Rubin LG. Prolongedcandidemia in infants following surgery for congenitalheart disease. Infect Control Hosp Epidemiol 2003;24:753e757.
44. Nieto-Rodriguez JA, Kusne S, Manez R, et al. Factorsassociated with the development of candidemia andcandidemia-related death among liver transplantrecipients. Ann Surg 1996;223:70e76.
45. Schwenzer KJ, Gist A, Durbin CG. Can bacteremia bepredicted in surgical intensive care unit patients? IntensiveCare Med 1994;20:425e430.
46. Chwals WJ, Fernandez ME, Jamie AC, Charles BJ,Rushing JT. Detection of postoperative sepsis in infants withthe use of metabolic stress monitoring. Arch Surg 1994;129:437e442.
47. Coello R, Glenister H, Fereres J, et al. The cost of infectionin surgical patients: a caseecontrol study. J Hosp Infect1993;25:239e250.
48. Lapchik MS, Castelo Filho A, Pestana JO, Silva Filho AP,Wey SB. Risk factors for nosocomial urinary tract and post-operative wound infections in renal transplant patients:a matched-pair caseecontrol study. J Urol 1992;147:994e998.
49. Givens CD, Wenzel RP. Catheter-associated urinary tractinfections in surgical patients: a controlled study on theexcess morbidity and costs. J Urol 1980;124:646e648.
50. Harris AD, Samore MH, Lipsitch M, Kaye KS, Perencevich E,Carmeli Y. Control-group selection importance in studies ofantimicrobial resistance: examples applied to Pseudomonasaeruginosa, Enterococci, and Escherichia coli. Clin InfectDis 2002;34:1558e1563.
51. Hennekens CH, Buring JE. Methods to control confoundingin the design. In: Mayrent SL, editor. Epidemiology inmedicine. Boston: Little, Brown; 1987. p. 293e304.
52. Rothman KJ, Greenland S. Matching. In: Rothman KJ,Greenland S, editors. Modern epidemiology. 2nd edn. Phila-delphia, PA: Lippincott/Williams & Wilkins; 1998. p. 147e161.
53. Schwaber MJ, Carmeli Y. Antimicrobial resistance andpatient outcomes: the hazards of adjustment. Crit Care2006;10:164.
54. Bundy JK, Gonzalez VR, Barnard BM, Hardy RJ, DuPont HL.Gender risk differences for surgical site infections amonga primary coronary artery bypass graft surgery cohort:1995e1998. Am J Infect Control 2006;34:114e121.
55. Falagas ME, Kopterides P. Risk factors for the isolation ofmulti-drug-resistant Acinetobacter baumannii and Pseudo-monas aeruginosa: a systematic review of the literature. JHosp Infect 2006;64:7e15.
