Selection bias - Institute for Work & Health

Selection bias is a kind of error that occurs when the researcher decides who is going to be studied. It is usually associated with research where the selection ... Skiptomaincontent Whatresearchersmeanby... Selectionbias Aboutthe“Whatresearchersmeanby...”series Thisresearchtermexplanationfirstappearedinaregularcolumncalled“Whatresearchersmeanby…”thatranintheInstituteforWork&Health’snewsletterAtWorkforover10years(2005-2017).Thecolumncoveredover35commonresearchtermsusedinthehealthandsocialsciences.Thecompletecollectionofdefinedtermsisavailableonlineorinaguidethatcanbedownloadedfromthewebsite. Published:May2014 Mostscientificstudiesaredesignedtopinpointtheeffectofsomething—suchastheeffectofaconditionondevelopingaproblem(disease,injury)ortheeffectofanintervention(treatment,program)onovercomingaproblem.Scientistsusuallydetermineeffectbytakingtwosimilargroups—theonlydifferencebeingthegroups’exposuretothatconditionorintervention—andmeasuringthedifferenceinoutcomesexperiencedbythem. Butwhathappenswhenthetwogroupsselectedwerenotsimilartobeginwith?Whatifkeycharacteristicsdistinguishingthetwomighthaveplayedaroleinproducingthedifferentoutcomes?That’sanexampleofwhat’scalledselectionbias. Biasisatypeoferrorthatsystematicallyskewsresultsinacertaindirection.Selectionbiasisakindoferrorthatoccurswhentheresearcherdecideswhoisgoingtobestudied.Itisusuallyassociatedwithresearchwheretheselectionofparticipantsisn’trandom(i.e.withobservationalstudiessuchascohort,case-controlandcross-sectionalstudies). Forexample,sayyouwanttostudytheeffectsofworkingnightsontheincidenceofacertainhealthproblem.Youcollecthealthinformationonagroupof9-to-5workersandagroupofworkersdoingthesamekindofwork,butatnight.Youthenmeasuretheratesatwhichmembersofbothgroupsreportedthehealthproblem.Youmightconcludethatnightworkisassociatedwithanincreaseinthatproblem. Thetroubleis,thetwogroupsyoustudiedmayhavebeenverydifferenttobeginwith.Thepeoplewhoworkednightsmayhavebeenlessskilled,withfeweremploymentoptions.Theirlowersocioeconomicstatuswouldalsobelinkedwithmorehealthrisks—duetolesshealthydiets,lesstimeandmoneyforleisureactivitiesandsoon.Soyourfindingmaynotberelatedtonightworkatall,butareflectionoftheinfluenceofsocioeconomicstatus. Selectionbiasalsooccurswhenpeoplevolunteerforastudy.Thosewhochoosetojoin(i.e.whoself-selectintothestudy)mayshareacharacteristicthatmakesthemdifferentfromnon-participantsfromtheget-go.Let’ssayyouwanttoassessaprogramforimprovingtheeatinghabitsofshiftworkers.Youputupflyerswheremanyworknightshiftsandinvitethemtoparticipate.However,thosewhosignupmaybeverydifferentfromthosewhodon’t.Theymaybemorehealthconscioustobeginwith,whichiswhytheyareinterestedinaprogramtoimproveeatinghabits. Ifthiswasthecase,itwouldn’tbefairtoconcludethattheprogramwaseffectivebecausethehealthofthosewhotookpartintheprogramwasbetterthanthehealthofthosewhodidnot.Duetoself-selection,otherfactorsmayhaveaffectedthehealthofyourstudyparticipantsmorethantheprogram. Minimizingselectionbias Goodresearcherswilllookforwaystoovercomeselectionbiasintheirobservationalstudies.They’lltrytomaketheirstudyrepresentativebyincludingasmanypeopleaspossible.Theywillmatchthepeopleintheirstudyandcontrolgroupsascloselyaspossible.Theywill“adjust”forfactorsthatmayaffectoutcomes.Theywilltalkaboutselectionbiasintheirreports,andrecognizethedegreetowhichtheirresultsmayapplyonlytocertaingroupsorincertaincircumstances. Anotherwayresearcherstrytominimizeselectionbiasisbyconductingexperimentalstudies,inwhichparticipantsarerandomlyassignedtothestudyorcontrolgroups(i.e.randomizedcontrolledstudiesorRCTs).However,selectionbiascanstilloccurinRCTs.Forexample,itmaybethatthepoolofpeoplebeingrandomlyassignedtotheinterventiongroupisnotveryrepresentativeofthewiderpopulation.Oritcouldbetheresearcher’sallocationtechniquesaren’tsorandom(e.g.whenclinicians,oftenmotivatedbygoodintentions,manipulatetheallocationmethodtogettheirpatientsinatreatmentgroupinsteadofthecontrolgroup). Often,selectionbiasisunavoidable.That’swhyit’simportantforresearcherstoexaminetheirstudydesignforthistypeofbiasandfindwaystoadjustforit,andtoacknowledgeitintheirstudyreport. Source: AtWork,Issue76,Spring2014:InstituteforWork&Health,Toronto