Prevalence and patterns of major depressive disorder in the United state labor force - PDF document

TheJournalofMentalHealthPolicyandEconomicsJ.MentalHealthPolicyEcon.,123±131(1999)PrevalenceandPatternsofMajorDepressiveDisorderintheUnitedStatesLaborForceDaveE.Marcotte,VirginiaWilcox-Go*andD.PatrickRedmonPolicySciencesGraduateProgram,UniversityofMarylandBaltimoreCounty,Baltimore,MD21250,USADepartmentofEconomics,NorthernIllinoisUniversity,DeKalb,IL60115,USAMarylandHealthServicesCostReviewCommission,Baltimore,MD21215,USA *Correspondenceto:V.Wilcox-Go *Estimatesofthe12-monthprevalencerateofmajordepressionintheUSrangefromabout5to10per100personsintheadultpopulation(seeetal.forestimatesbasedontheEpidemiologicCatchmentAreaStudyandKessleretal.forestimatesfromtheNationalComorbiditySurvey). currentlyparticipatinginthelaborforcecomparedtothoseoutofthelaborforce.Thelaborforceconsistsofallpersonscurrentlyworking,oractivelyseekingwork.Amongthoseparticipatinginthelaborforce,wealsoestimateprevalenceratesseparatelyfortheemployedandtheunemployed.*Second,weexaminewhetherthelabormarketconse-quencesofdepressionvarybygenderandoverthelife-cycle.Weexpectgendertobeanimportantdimensionoftherelationshipbetweendepressionandthelabormarketbothbecausedepressionmorecommonlyaffectswomenandbecausechoicesaboutlabormarketparticipationvarybygender.Weexpectthattherelationshipbetweendepressionandlaborforcestatusislikelytovarybyageforanumberofreasons.Forexample,youngpeoplewithdepressionmaybemorelikelytodropoutofschoolandparticipateinthelaborforcethancomparablepeerswhoarenotdepressed.However,aspeopleageandleaveschool,wemightexpectdepressivestobelesslikelytoparticipateinthelaborforcethantheirhealthypeersaslongasmajordepressionincreasesthecostsordecreasestherewardsassociatedwithworking.Third,wehopetoidentifywhetherpeoplecanrecoverfromdepressionwithtime,atleastintermsoftheirabilitytoseekand®ndemployment.Majordepressivedisordermayfundamentallyalteraperson'sworkingcareer,renderinghim/herlessproductiveorlesssecurethanhealthyworkers.Alternatively,theillmaybeabletosurmountorrecoverfromdepression,sothat,overtime,theirlabormarketprospectsarenotremarkablydifferentfromtheircounterpartswhohaveneverhadtodealwiththedebilitatingeffectsofmajordepression.Identifyingtheextenttowhichmajordepressivedisorderaffectsworkersiscrucialifwehopetounderstandthefullcostsofoneofthemostcommonanddebilitatingformsofmentalillnesses.Animportantsteptowardsunderstandingthefullcostsofdepressionisidentifyingtheextenttowhichdepressionoccursinthelabormarket.Furthermore,studyingthelabormarketconsequencesofmajordepressionmayprovideinsightintotheprocessesthroughwhichenvironmen-talfactorsaggravateillnessandhelpinpredictingwhetherandhowtheaf¯ictedcanadapttoorrecoverfromtheirillness.Thelinksbetweenstressesinherentinthelabormarketandtheonsetandseverityofdepressionarewellestablished.²Byexaminingtherelationshipbetweendepressionandlabormarketoutcomes,wehopetolaythegroundworkforamorecompleteunderstandingofthetiesbetweenthelabormarketanddepression.DataandMethodsUntilrecently,workondepressionandthelaborforcehasbeenhinderedbyalackofnationallyrepresentative *Therearethreedistinctlabormarketstatesintowhicheconomistsclassifypeople.Apersoncanchoosenottoparticipateinthelaborforceatall(about41%ofwomenand25%ofmendo).Ifapersonchoosestoparticipateinthelaborforce,s/hemaybeemployedorunemployed.²Indeed,ofalllifeeventswhichhaveemergedaspredictorsofdepression,thestrongesthasbeentheoccurrenceofmarkedlystressfuleventspriortoConsistently,thisresearchprovidesevidencethatfordepressiveslifestressisunusuallyhigh.D.E.MARCOTTEETAL.1999JohnWiley&Sons,Ltd.J.MentalHealthPolicyEcon.,123±131(1999)informationaboutlabormarketexperiencesandmentalillness.DatafromtheNationalComorbiditySurvey(NCS)providethe®rstopportunitytodevelopanunderstandingoftheprevalenceofdepressioninthelabormarketandtounderstandtheconsequencesofthisdiseaseforthoseaf¯icted.TheNCSisthe®rstnationallyrepresentativesurveydesignedtostudytheprevalence,causesandconsequencesofmentalillnessesandcomorbiditybetweenvariouspsychiatricdisorders.Thedataareastrati®ed,multi-stageareaprobabilitysampleofpersons15±54yearsold,livinginthe48contiguousstates.ThesurveywasadministeredbytheSurveyResearchCenter(SRC)attheUniversityofMichiganbetweenSeptember1990andFebruary1992.Theresponseratewas82.6percentwith8098totalrespondents.TheNCSprovidesinformationaboutrespondents'employ-mentstatusaswellastheirongoingandpastexperienceswithmentalillness.Todiagnosementalillnesses,theNCSusedamodi®edversionoftheCompositeInternationalDiagnosticInterview(CIDI),astructureddiagnosticinter-viewbasedontheDiagnosticInterviewScheduleanddesignedtobeusedbytrainedinterviewerswhoarenotclinicians.§ResponsestothesequestionsareusedtodiagnosetheprevalenceofseveralDSM-III-Rpsychiatricdisorders,includingmajordepressivedisorderoversamplemembers'lifetimesandduringtheprevious12months.UsingtheNCSdata,weestimatetheprevalenceoftheselifetimeand12-monthcasesofmajordepressivedisorders.First,weestimatethefrequencywithwhichNCSrespondentsreportsymptomswhichsatisfydiagnosticcriteriaformajordepressionduringthe12monthspriortointerview.Wethenestimatethefrequencywithwhichrespondentsexperi-encedsuchsymptomsatanytimeduringtheirlifetimes,comparingthesefrequenciesforpersonsparticipatinginthelaborforceandforthosenotparticipating,andfortheemployedandunemployed.Apriori,weanticipatemajordepressionwillhaveanimpactontheworkinglivesoftheillinavarietyofways.First,thosesufferingfromdepressionmaybeunabletoparticipateinthelabormarketatall.Second,withinthelabormarket,workerswithahistoryofmajordepressionorwithacurrentdepressiveepisodemayexperiencemoreunemploymentiftheirillnesshindersjobperformanceandsubsequentlyleadstoahigherprobabilityofdismissal,oriftheyhaveahardertimeengaginginthejobsearchprocesswhenunemployedandthereforebecomeemployedataslowerrate.Alternatively,wemayobservehigherlevels ³TheresponseratefortheNCSishighcomparedtopreviousmentalhealthsurveys.Nonetheless,dataavailablefromtheoriginalNCSsurveyweresupplementedbyaspecialsurveyofnon-respondents,whichelicitedinformationfromapproximatelyhalfoftheoriginal`non-respondents'.§Diagnosesknowntohavelowprevalenceinpopulation-basedsurveys(suchassomatizationdisorder)weredeletedfromthesurvey.Othermodi®cationsincludedtheadditionofcommitmentandmotivationprobesforrecalloflifetimeepisodesandclarifyingprobesforCIDIquestionsfoundinpilotworktobeunclearorconfusingtorespondents.SeeKesslerfordetails.SeeAnthonyetal.etal.andHelzeretal.discussionofdiagnosesbasedoninterviewsconductedbylayinterviewersusingtheDiagnosticInterviewSchedule. ofdepressionamongtheunemployedifthestressesassociatedwithunemploymentitselfcanservetotriggerdepressiveepisodes.Suchprecursorsmayincludestressbroughtaboutbyeconomicinsecurityorthenegativepersonalandsocialconsequencesofunemployment.Tomorefullyunderstandthesigni®canceofdepressionintheworkplace,wealsoattemptedtoaccountforthecomplexityofthedisorder.Personsaffectedbymajordepressivedisordercanhavevastlydifferentexperienceswiththeirillnesses.Forsome,oneepisodemaybeallthatiseverendured.Forothers,repeatedcyclingintodepressiveepisodesmarksthecourseofthedisorder.Inaddition,duringanyoneepisode,differentindividualsmayexperiencedifferentlevelsofdebilitation.Someexperiencesymptomswhichonlymildlyimpedetheirdailyfunctioning.Othersexperiencesymptomswhichmoreseverelyinterferewiththeirlives.Toexaminetheseissues,weseparatelyestimatethebivariatefrequencydistributionsoflabormarketoutcomesandthefrequencyorseverityofdepressiveepisodes.Wethenexaminedifferencesindepressionandlabormarketoutcomesbygender.Wedosobecausedepressionmorecommonlyaf¯ictswomen.Womenalsomakedifferentchoicesaboutlabormarketparticipationandemployment.*Becauseofthesedifferentpatternsofillnessandemploymentstatus,weexpecttoobservedifferentpatternsofdepressionconditionalonlabormarketstatusbygender.Weanalysetherelationshipbetweenlabormarketstatus,gender,anddepressionincross-tabularform.Wedothisfortworeasons.First,aninterestingpublichealthquestionis:Inwhichlabormarketstatesdoesdepressionaffectmenandwomen?Second,aninterestingeconomicquestionis:Whatarethelabormarketchoicesmadebymenandwomenwhohaveahistoryofdepression?Wepresentsimpleanswerstobothquestions.However,thesequestionsarelikelytoberelated,andwhichquestiononeasksrevealsanassumptionaboutcausation.Infact,itmaybethecasethatlabormarketstatushasadirecteffectontheprevalenceofdepressionanddepressionhasaneffectonlabormarketstatus.Considerationofsuchacomplicatedrelationshipwillbelefttolaterresearch.Here,ouraimistopointouttheempiricalregularitieswhichfuturetheoreticalandempiricalworkwillneedtoexplain.Next,weexaminetherelationshipbetweendepressionandlabormarketstatusoverthelife-cycle.Depressionmayaffectearlydecisionsaboutwhetherornotapersonseeksemploymentorremainsinschool.Weanalyzedepressionoverthelife-cyclebydirectcomparisonoflifetimeand12-monthprevalenceratesofdifferentagegroupswithintheNCSsample.Unfortunately,thecross-sectionalNCSdatarequireustodrawinferencesaboutdepressionandthelabormarketoverthelife-cyclebycomparisonofdifferentcohorts.Wediscussour®ndingsinthecontextofpotentialcohorteffects. *Therateatwhichwomensupplytheirlaborinthemarkethasconsistentlybeenfoundtobemoresensitivetothelevelofwages.Atlowwages,womenaremorelikelythanmentostayordropoutofthelaborforce,optinginsteadtodevotetimetonon-marketactivitiessuchasraisingchildren.MAJORDEPRESSIVEDISORDERINLABORFORCE1999JohnWiley&Sons,Ltd.J.MentalHealthPolicyEcon.,123±131(1999)Our®nalanalysissupplementstheunivariateandbivariateanalysesbyseparatelyexaminingtheeffectsofvariousdimensionsofillnessonlaborforceparticipationandemploymentformenandwomen.Weestimatebivariatelogitmodelstoidentifythecontributionsofthepresence,severityandrecencyofmajordepressivedisorderontheselabormarketoutcomesseparately.Wecontrolforstandardlife-cyclechangesinlabormarketbehavior.Laborforceparticipationtendstorisewithagethroughthe40s,andthenfallasworkersstarttoretireintheir50s.²Weparticularlyinvestigatethepossibilitythat,astimesinceonsetofamajordepressiveepisodepasses,thelaborforceparticipationratesandemploymentstatusoftheaf¯ictedmaycometolookmoreandmorelikethoseoftheLifetimeand12-MonthPrevalenceofMajorDepressionintheLaborMarketBasedonlifetimeprevalence,depressionisascommoninthelabormarketasinthegeneralpopulation.Amongthepopulationparticipatinginthelaborforce,15.7percenthavesufferedfrommajordepressivedisorderatsometimeintheirlives(seeTable1).Thiscompareswith15.8percentofthetotalpopulationand16.2percentofthepopulationnotparticipatinginthelaborforce(theseratesarenotstatisticallydifferentatconventionallevels).Whilelifetimeprevalenceratesdifferlittlebylaborforcestatus,thedifferencesaremoredramaticamongthosewith12-monthdisorders.Speci®cally,8.6percentofthoseinthelaborforceenduredaboutofmajordepressionduringthe12monthspriortobeinginterviewed,comparedto11.6percentforthosenotparticipatinginthelabormarket(theseratesarestatisticallydifferent(0.001)).So,whilelifetimeprevalenceofmajordepressivedisorderisashighinthelaborforceasitisinthegeneralpopulation,12-monthdepressivedisordersaremorecommonamongthepopulationnotparticipatinginthelaborforce.Amongthoseparticipatinginthelaborforce,therelationshipbetweendepressionandemploymentstatusismuchstarker.Employedworkersaresigni®cantlylesslikelytoreportahistoryofdepression,andespeciallyadepressiveepisodewithinthe12monthspriortotheinterview,thantheunemployed(again,seeTable1).Amongworkerscurrentlyunemployed,21.9percenthadahistoryofdepression,comparedto15.2percentoftheemployed.12-monthprevalenceofdepressionamongtheunemployedismorethantwiceashighasprevalenceamongtheemployedÐ16.5percentversus7.9percent. ²Unemploymenttendstofallwithage.Timesincemostrecentonsetisnaturallycorrelatedwithage,anddevelopingasenseoftheeffectoftimesincemostrecentonsetonlabormarketoutcomesrequiressomebasiccontrolforthiseffect. Table1.PrevalenceofdepressionamongNCSsample,bygenderandbylaborforceparticipation Proportionw/majorProportionw/majorSamplesizedepression(lifetime)depression(12months) Totalsample80930.1580.092BygenderMen38460.1170.066Women42470.1980.117Byemploymentstatus1)Outofthelaborforce14650.1620.1162)Participatinginthelaborforce64760.1570.086a)Unemployed5720.2190.165b)Employed59040.1520.079TestofindependenceoflaborChi-square0.18713.725forceparticipationand-value0.6650.001TestofindependenceofChi-square16.35045.312employmentstatus(conditional-value0.0010.001onlaborforceparticipation)and MeasuresofSeverityofMajorDepressionbyLaborMarketStatusItisrarethatpeoplewithahistoryofdepressiveillnesshaveenduredonlyoneepisodeofmajordepression(seeTable2).Forexample,only24.5percentofpeopleoutofthelaborforcewhohavesufferedfromdepressionhaveexperiencedjustoneepisode.Atthesametime,two-thirdshaveexperienced®veorfewerepisodes.We®ndnosigni®cantdifferencesinthenumberoflifetimeepisodessufferedbydepressiveswhoarenotparticipatinginthelabormarketandthosewhoare(seeTable2).However,withinthelabormarket,depressiveswhoareunemployedhavesufferedfrommoreepisodesofmajordepressionthananyothergroup.Only17.7percentoftheunemployedhaveexperiencedjustoneepisodeofTable2.NumberofdepressiveepisodesamongthoseexperiencinglifetimeMDE:bylaborforcestatus Inthelaborforce OutofthelaborforceTotalUnemployedEmployed NumberofepisodesofFrequencyCumulativeFrequencyCumulativeFrequencyCumulativeFrequencyCumulativedepressionever(%)frequency(%)frequency(%)frequency(%)frequencyexperienced(%)(%)(%)(%) 124.524.527.327.317.717.728.628.6215.540.017.444.722.340.016.845.3312.352.312.256.96.346.313.058.347.760.08.965.810.256.58.767.056.967.06.572.26.362.76.573.56±1014.981.912.284.411.974.612.285.711±206.988.85.389.711.686.24.590.221±302.491.22.492.02.588.72.492.68.8100.08.0100.011.3100.07.5100.1-value0.883-value0.006 D.E.MARCOTTEETAL.1999JohnWiley&Sons,Ltd.J.MentalHealthPolicyEcon.,123±131(1999)depression,comparedto28.6percentoftheemployed(seeTable2).Thissuggeststhat,evenifthereissomecausalrelationshipfromunemploymenttodepression,depressivesobservedinunemploymenthaveahistoryofdepression,notjustasingleepisodeassociatedwiththeircurrentemploymentstate.Wealso®ndevidenceofarelationshipbetweenaseconddimensionofseverityandlabormarketstatus.Ingeneral,peoplewhohaveexperiencedatleastonemajordepressiveepisodereportsigni®cantinterferenceintheirlivesduetodepression.But,thoseoutofthelaborforcearemorelikelythanthosewhoparticipateinthelaborforcetoreportthattheirepisode(s)interfered`alot'or`some'withtheirlives(84percentversus77percent).Theyarealsolesslikelytoreportlittleornointerference.Thissuggeststhattosomedegree,depressivesparticipatinginthelaborforcehave experienced(orareexperiencing)episodesofmajordepressioninvolvinglesssevereimpairment,onaverage(9.611).Clearly,however,evenamongthoseparticipatinginthelaborforce,depressionisreportedtoresultinsubstantialimpairment.Whiledepressivesinandoutofthelaborforcedifferinthereportedinterferenceintheirlivesassociatedwithdepressiveepisodes,suchvariationdoesnotexistbyemploymentstatusamongthoseinthelaborforce.Aswesawabove,prevalenceofmajordepressionismuchhigheramongunemployedworkersthanamongemployedworkers.However,wedonot®ndasigni®cantrelationshipamongthoseparticipatinginthelabormarketbetweenemploymentstatusandthedegreeofinterferencecausedbytheillness.Theunemployedweremorelikelytoreportthehighestlevelofinterferenceduetodepression(42.7percent,versus37.1percentamongemployedworkers).Theywerealsomorelikelytoreportonlyalittleinterference(26.1percent,versus18.2percentamongtheemployed).So,whilethestressfulstateofunemploymentmaybeassociatedwithhigherprevalenceandfrequencyofmajordepressiveepisodesforavarietyofreasons,itdoesnotappeartobeassociatedwithmarkedlymoredebilitatingepisodesthanaverage.Thismaybeduetothefact,identi®edabove,thatindividualswiththemostdebilitatingboutsofdepressionaremorelikelytohavedroppedoutofthelaborforcealtogether.GenderDifferencesTheresultspresentedinTable1illustratethatlifetimeand12-monthprevalenceofdepressionisnearlytwiceashighamongwomenthanmen.Whileprevalenceratesareconsistentlyhigherforwomen,weconsiderbelowwhetherpatternsofdepressioninvariouslaborforceandemploymentstatesvarybygender.InTable3,wepresentresultsofouranalysisofgender,depressionandlabormarketoutcomes.Forbothmenandwomen,therearenolargedifferencesinlifetimeprevalenceofmajordepressionbylaborforceparticipationstatus.10.0percentofmenand19.2percentofwomenoutofthelaborforcehavesufferedfrommajordepressivedisorderatsometimeintheirlives.Theselifetimeprevalenceratesarenotsigni®cantlydifferentfromthoseofmenandwomenwhodoparticipateinthelaborforce(0.197formenand0.516forwomen).*Asfurtherevidenceoftheweakrelationshipbetweenlifetimeincidenceofmajordepressionandlaborforceparticipationforbothmenandwomen,noticethatamongthosewhohaveexperienceddepressionintheirlifetimes,11.4percentofmenand26.0percentofwomenarenotparticipatinginthelaborforce.Thiscomparesto13.6percentofmenand27.1percentofwomenwhohaveneverexperienceddepressionandarenotparticipatinginthelaborforce.² *11.44percentofmenand20.07percentofwomeninthelaborforcehaveexperienceddepressionatsometimeintheirlives.²Wecandetectnosigni®cantdifferencebetweenprevalenceoflifetimedepressivedisorderandlaborforceparticipation(0.197formen,and0.516forwomen).MAJORDEPRESSIVEDISORDERINLABORFORCE1999JohnWiley&Sons,Ltd.J.MentalHealthPolicyEcon.,123±131(1999)However,forbothmenandwomen,astrongerrelationshipexistsbetweenlaborforceparticipationandrecentexperi-enceswithdepression.Ofmenexperiencingdepressioninthe12monthspriortointerview,16.6percentwerenotparticipatinginthelaborforce.Thiscomparesto13.1percentofmenwithnosuchrecentexperiencewithdepression.Forwomen,30.3percentofthosewhohadexperienceddepressioninthe12monthsbeforeinterviewwerenotparticipatinginthelaborforce,comparedto26.5percentofotherwomen.Interestingly,whilelaborforceparticipationratesarelowerforwomenwith12-monthdepressivedisorder,thereisaproportionatelylargerdiffer-encebetweenthebehaviorofrecentlydepressedandhealthymen.Thissuggeststhatdepressedmenarerelativelylikelytodropoutofthelaborforce.Anotherexplanationfortherelativelyhighratesatwhichmenwithrecentexperiencewithdepressiondonotparticipateinthelaborforceisthatlackofparticipationdispro-portionatelyincreasestheriskofdepressionformen,.Thismayariseifmenwhohavedroppedoutofthelaborforcebeargreaterpersonaleconomicandsocialcostswhichincreasetheriskofdepression.Amongmenwhodonotparticipateinthelaborforce,8.2percenthaveexperiencedadepressiveepisodeinthe12monthspriortointerview.Thisissigni®cantlyhigherthanthe5.9percentofemployedmenwhohaveexperiencedarecentdepression.Thereisnosuchdifferenceinlifetimedepressionprevalenceamongthoseoutofthelaborforceandamongthoseemployed.Thisrelativelystrongrelationshipbetweenlackofparticipationinthelaborforceandrecent/currentepisodesofdepressionisconsistentwithanincreasedriskfordepressionamongthosenotinthelaborforce.Wenextturnourattentiontotherelationshipbetweendepressionandunemploymentandemploymentbygender.Amongbothmenandwomeninthelabormarket,lifetimeand12-monthprevalenceratesofdepressionaremuchhigheramongtheunemployedthantheemployed.Indeed,we®ndthattherelationshipbetweenunemploymentanddepressiondominatestheinteractionbetweendepressionandalllabormarketstates.Amongunemployedmen,thelifetimeprevalencerateofmajordepressionis18.3percent,andthe12-monthprevalencerateis12.0percent(comparedwith11.4percentofunemployedwomenhaveforemployedmen).Similarly,amongwomen25.6percentofunemployedwomenhaveexperienceddepressionintheirlifetime,21.0percentinthe12monthspriortointerview.Thiscomparesto19.5percentand10.2percentforemployedwomen.Clearly,comparedtoallotherstates,depressionismuchmoreprevalentamongtheunemployed.Takingadifferenttackontheproblem,ofpeopleinthelaborforcewhohaveexperiencedadepressiveepisodeinthe12monthspriortointerview,12.3percentofmenwereunemployed,and11.7percentofwomenwereunemployed.³ ³PleasenotethatthisisnotanunemploymentrateinthemannerwhichtheUSBureauofLaborStatisticsreportstheof®cialunemploymentrate.Thatrateisthepercentageofallworkerswhoareparticipatinginthelaborforce,andwhoareunemployed(withsomestipulations). Table3.Bivariaterelationshipbetweendepressionandlaborforcestatus,bygender RowpercentNomajorMajorNomajorMajorColumnpercentdepressivedepressivedepressivedepressivedisorderdisorderdisorderdisorder(ever)(ever)(12month)(12month) Outofthelaborforce89.9810.0291.768.2413.6111.4313.1216.64Unemployed81.7318.2788.0311.976.2910.606.4112.30Employed88.5611.4494.115.8980.1077.9780.4771.06Outofthelaborforce80.8519.1586.8113.1927.1426.0026.4630.37Unemployed74.4325.5778.9721.036.038.385.8111.69Employed80.4719.5389.8310.1766.8365.5267.7357.93 Testsofdifferencesinpatternsofemploymentstatus(employed/unemployed)anddepression Chi-squarep-value Employmentstatusanddepressivedisorder,lifetime(men)12.6720.002Employmentstatusanddepressivedisorder,12month(men)17.2160.001Employmentstatusanddepressivedisorder,lifetime(women)5.8460.054Employmentstatusanddepressivedisorder,12month(women)30.2420.001 Thisissigni®cantlyhigherthanthe6.4percentofmenand5.8percentofwomenwithoutarecentdepressionwhowereunemployed.Thispatternprovidesfurtherevidenceofastrongrelationshipbetweenunemploymentanddepression.Life-CycleDifferencesFigure1,weillustratethepatternofthelifetimeprevalenceratesofmajordepressivedisorderbylaborforceandemploymentstatusandbyagegroup.Amongtheveryyoung,thereisnohigherprevalenceofmajordepressivedisorderamongthosenotinthelaborforcecomparedtolaborforceparticipants,oramongthoseunemployedcomparedtotheemployed.However,amongthoseinmiddleageranges,prevalenceofdepressionissigni®cantlyhigherforthoseoutofthelaborforceorunemployed.Duringtheseageranges,inwhichpeopleareattheprimeoftheirworkinglives,depressionisassociatedwiththenegativeemploymentstatesofbeingoutofthelaborforceorunemployedifinthelaborforce.Interestingly,however,therelativelyhighlevelsofdepressionamongthoseoutofthelaborforceandunem-ployedoftenobservedintheagerangesbetween20and50donotshowupamongthoseover50.Infact,depressionismildlymorecommoninthelabormarketthanout,andamongtheemployedforthisoldestgroup.Together,these®ndingssuggestdepressionisassociatedwithpoorerlabormarketoutcomesformiddle-agedpersons.ThisconclusionissomewhatcomplicatedbypotentialcohortD.E.MARCOTTEETAL.1999JohnWiley&Sons,Ltd.J.MentalHealthPolicyEcon.,123±131(1999)effects,however.Particularly,thegrowingopennesstomentalhealthissuesandtherecentboominuseofserotoninselectivereuptakeinhibitors(e.g.ProzacandZoloft)maybeassociatedwithanincreasedpropensitytoself-reportdepressivesymptomsforyoungcohorts.However,thepresent®ndingssuggestthehighestprevalenceinthelaborforceisnotamongtheyoungestcohorts,leadingustobelievethattheobservedpatternsformiddle-agedworkersarenotduetosystematiceffects.MultivariateModelsTable4,wepresenttheresultsofmultivariatelogisticregressionsforthreegroups:theentiresampleandseparatesubsamplesofmenandwomen.Inallthree,weincludeseveralpredictorsoflabormarketoutcomes:theindividual'sageandgender;experiencewithdepression(yes/no);thenumberofepisodesexperienced;theageatwhichthe®rstepisodewasexperiencedandthenumberofyearssincethelastepisode.Forthesampleincludingbothmenandwomen,wealsoincludeaninteractionbetweenthedepressionandfemaledummyvariables,topickupanyuniqueeffectsofdepressionforwomen.The®rstthreecolumnspresentthemarginaleffectsofdepressionanditscourse,aswellasbasicpersonalcharacteristics,onlaborforceparticipation;* *Thesemarginaleffectsarethepartialderivativesoftheprobabilityoflaborforceparticipation,withrespecttoaparticularexplanatoryvariable.Thesemarginalsforindependentvariablearecalculatedbasedonlogitcoef®cientestimatesas: Figure1.Majordepressivedisorderbyageandlabormarket/employmentstatustheassociatedoddsratioandthechi-squareteststatisticforthesigni®canceofeachfactorindetermininglaborforceparticipation.The®nalthreecolumnspresentthemarginaleffectsonemployment(conditionalonlaborforceparticipation),andtheassociatedoddsratiosandchi-squarestatistics.InthetoppanelofTable4aretheestimatesforthesamplecontainingmenandwomen.Theseestimatesillustratethattherelationshipbetweendepressionandlabormarketoutcomesisnothomogeneous.Ifapersonhaseverexperiencedamajordepressiveepisode,weestimatethatthereisnosigni®canteffectonhisorherpropensitytoparticipateinthelaborforce.However,weestimatethatworkerswithahistoryofdepressionare4.8percentagepointslesslikelytobeemployed.²Thenegligiblemarginaleffectofthenumberofepisodesindicatesthatnumerousepisodesofdepressiondonotfurtherimpactlabormarket d dxikL(x9ib)5exp(x9ib) 1(x9ib)]2bkwhereiindexesindividuals.Forthispurpose,allindependentvariableswereevaluatedattheirmeans.²Alternatively,interpretingtheoddsratio,workerswithahistoryofdepressionareabouthalfaslikelyastheirhealthycounterpartstoavoidunemployment.MAJORDEPRESSIVEDISORDERINLABORFORCE1999JohnWiley&Sons,Ltd.J.MentalHealthPolicyEcon.,123±131(1999)outcomes.³However,workerswithahistoryofdepressionaremorelikelytobeemployedthelongeristheelapsedtimesincethelastepisode.Foreachyearthatpassessincethelastepisode,thechanceofbeingemployedincreasesby0.6percentagepoints.Asaresult,wepredictthat,onaverage,peoplewhohavesufferedfromanepisodeofmajordepressioncanexpectemploymentratescomparabletothosewithoutahistoryofmajordepression8yearspost-onset,intheabsenceoffurtherepisodes.Whilewomenarelesslikelytobeinthelaborforce,theestimatesforthecombinedsampleofmenandwomenindicatesthattherelationshipsbetweendepressionandlaborforceparticipationandemploymentdonotvarysigni®cantlybygender.However,toallowallthreeaspectsofdepression,aswellasage,tovarybygender,wereportestimatesfromseparatemultivariatelogisticregressionsformenandwomeninthebottomtwopanelsofTable4.These®ndingsindicatethatthereareimportantdifferencesbetweenmenandwomenintheeffectsofdepression,numberofepisodesandtimesincelastepisodeonlaborforceparticipationandemployment.Depressionsigni®cantlyreducesthelikelihoodthatamanwillbeinthelaborforce(2.7percentagepoints),butdoesnothaveasigni®canteffectonthelikelihoodthatawomanwillparticipateinthelaborforce.However,forbothmenandwomeninthelaborforce,depressionsigni®cantlyreducesthelikelihoodofemployment.Amanwhohasahistoryofdepressionis4.6percentagepointslesslikelytobeemployedthanamanwithoutahistoryofdepression.Similarly,awomanwithahistoryofdepressionis3.9percentagepointslesslikelytobeemployed.Thenumberofepisodesofdepressionhasaverydifferenteffectonlabormarketoutcomesformenandwomen.Asthenumberofepisodesrises,thelikelihoodofawomanbeinginthelaborforcefalls.Further,amongwomeninthelaborforce,thelikelihoodofemploymentfallsasthenumberofepisodesrises.Theseeffectsarestatisticallysigni®cant,althoughrelativelysmallinmagnitude.Formen,however,thenumberofepisodesissigni®cantlyrelatedtoneitherlaborforceparticipationnoremployment.Thenumberofyearssincethelastepisodealsohasastrikinglydifferenteffectformenandwomen.Formen,eachyearsincethelastepisodeisassociatedwitha1.6percentagepointincreaseinthelikelihoodofbeinginthelaborforce.Forwomen,we®ndnosigni®cantassociationbetweenthetimesincethelastepisodeandlaborforceparticipation.However,amongwomeninthelaborforce,eachyearsincethelastepisodeisassociatedwitha1.5percentagepointincreaseinthelikelihoodofemployment,whilewe®ndnosigni®cantassociationbetweentimesincethelastepisodeandemploymentformen.ConclusionsandImplicationsThe®ndingsreportedinthispaperaredrawnfromastudyoftherelationshipbetweenmajordepressivedisorderand ³Wealsotriedaquadraticspeci®cationbetweenthenumberofepisodesandlabormarketoutcomes.However,statisticaltestsindicatedthatthelinearspeci®cationsuf®cientlydescribestherelationship. Table4.Logit-estimatedrelationshipbetweenthehistoryandcourseofdepressiveillnessandlaborforcestatus:bygender Dependentvariable:laborforceparticipationdummyDependentvariable:employmentdummy MarginaleffectOddsChi-squareMarginaleffectonlaborforceratiotestonemploymentChi-squareparticipationstatisticinthelaborOddstestIndependentvariablesforceratiostatistic MenandwomenIntercept0.259237.44**0.08944.30**Depression(ever)0.0090.9180.150.0480.47513.27**Numberofepisodes0.0010.9952.180.0000.9932.75**Timesincelastepisode0.0041.0362.490.0061.0985.53**Female(1female,00.1580.236275.33**0.0100.8561.90Femaledepression0.0081.0770.100.0131.2270.70Age0.0011.0095.69**0.0021.03943.52**MenonlyIntercept0.08229.30**0.08725.65**Depression(ever)0.0270.6123.62*.0460.46711.23**Numberofepisodes0.0001.0060.350.0001.0080.74Timesincelastepisode0.0161.3274.80**0.0011.0190.17Age0.0021.03619.94**0.0021.03719.57**WomenonlyIntercept0.21358.48**0.07917.87**Depression(ever)0.0101.0570.200.0390.5578.99**Numberofepisodes0.0010.9933.64*0.0010.9886.95**Timesincelastepisode0.0031.0190.730.0151.2537.78**Age0.0001.0000.010.0031.04023.58** **Signi®cantat5%level.*Signi®cantat10%level.labormarketoutcomes.UsingdatafromtheNationalComorbiditySurvey,we®nddepressiontobejustasprevalentamongindividualswithinthelaborforceasamongthosenotinthelaborforce.Notsurprisingly,amongpeopleinthelaborforce,majordepressionismoreprevalentamongtheunemployedthanamongtheemployed.Therelationshipbetweendepressionandthelabormarketvariesoverthelifecycle.Depressiontendstohaveasigni®cantnegativerelationshipwithlaborforceparticipationandemploymentduringtheagesinwhichindividualsareinthemiddleyearsoftheircareers.Therearenosigni®cantnegativeeffectsofdepressiononlaborforceparticipationamongtheveryyoungandamongpeopleintheir®fties.Thelackofastrongrelationshipbetweendepressionandlaborforceparticipationamongtheyoungmaystemfromteenagers'participationinbothschoolingandwork.Depressionmayeitherincreasethechancethatateenagerdropsoutofschoolandentersthelaborforceorthatateenagerdropsoutofthepart-timelaborforcebutstaysinschool.Forolderpersons,ifepisodesofdepressionaresubstantiallydebilitating,ahistoryofmajordepressionmaybeassociatedwithlowerearningsthroughoutthecareerandconsequentlymayaffectworkdecisionslateinlifeasworkerssaveforretirement.Finally,we®ndthatprevalenceisuniformlyhigherforwomenthanmen.However,itisdisproportionatelyhigheramongunemployedmenthanamongunemployedwomen.Further,ourresultsindicatethattheimpactofdepression,D.E.MARCOTTEETAL.1999JohnWiley&Sons,Ltd.J.MentalHealthPolicyEcon.,123±131(1999)numberofepisodes,andtimesincethelastepisodedifferstrikinglybetweenmenandwomen.Depressionhasanegativeeffectonlaborforceparticipationonlyformen,whileitreducesthelikelihoodofemploymentforbothmenandwomen.Multipleepisodeshaveanegativeeffectonlabormarketoutcomesonlyforwomen,butforbothmenandwomenwe®ndevidencethatindividualswithahistoryofdepressionmaybeabletorecoverfromtheirillnesstosomedegree.Astimesincethelastdepressiveepisodepasses,thelabormarketprospectsoftheillapproachthoseofindividualswhodonotsufferfromdepression.Thissuggeststhateffectivetreatmentfordepressionmayhavesigni®cantlong-termeconomicbene®ts.Theaimofourresearchhasbeentodescribetheparametersoftherelationshipbetweenmajordepressionandlaborforcestatus.Theseresultspresentabasicsetoffactsabouttherelationshipbetweenmajordepressivedisorderandlabormarketoutcomes.Wehavenot,however,attemptedtosortoutthecomplexitiesofthisrelationshiphere.Thesecomplexitiesariseatalmosteveryturn.Forinstance,thehighlevelofprevalenceofdepressionamongtheunemployedmaybeduetothepossibilitythatthestressesassociatedwithunemploymenttriggerdepressiveepisodesortothepossibilitythatworkerswhoaredepressedaremorelikelytobe®redorquit.Ourcontinuingresearchattemptstoaddresstheseproblems.Understandingwhenandhowdepressionaffectslabormarketoutcomesandwhenandhowlabormarketoutcomesaffectdepressionisan importantendeavorforthoseinterestedintreatingthediseaseandunderstandingitsconsequences.1.WeissmanM.Advancesinpsychiatricepidemiology:ratesandrisksformajordepression,Am.J.PublicHealth:445±451.2.KesslerRC.BuildingontheECA:TheNationalComorbiditySurveyandtheChildren'sECA,Int.J.MethodsPsychiatr.Res.3.RegierDA,NarrowW,RaeD,ManderscheidR,LockeB,GoodwinF.ThedeFactoU.S.MentalandAddictiveDisordersServiceSystem,Arch.Gen.Psychiatry:85±94.4.KesslerRC,McGonagleKA,ZhaoS,NelsonCB,HughesM,EshlemanS,WittchenHU,andKendlerKS.Lifetimeand12-MonthPrevalenceofDSM-III-RPsychiatricDisordersintheUnitedStates:ResultsfromtheNationalComorbiditySurvey,Arch.Gen.Psychiatry:8±19.5.BrownGW,HarrisTOPetoJ.Lifeeventsandpsychiatricdisorders,PartII:natureofacausallink,Psychol.Med.:159±176.6.BrownGW,HarrisT.Socialoriginsofdepression.London:Tavistock,1978.MAJORDEPRESSIVEDISORDERINLABORFORCE1999JohnWiley&Sons,Ltd.J.MentalHealthPolicyEcon.,123±131(1999)7.SwannAC,SecundaSK,StokesPE,CroughanJ,DavisS,KoslowSMaasJ.Stress,depression,andmania:relationshipbetweenperceivedroleofstressfuleventsandclinicalandbiochemicalActaPsychiatr.Scand.:389±397.8.BeckJC,WorthenK.Precipitatingstress,crisistheoryandhospitaliz-ationinschizophreniaanddepression,Arch.Gen.Psychiatry:123±129.9.RodgersB.Modelsofstress,vulnerabilityandaffectivedisorder,Affect.Disorders1991;1±13.10.IsometsaE,HeikkinenM,HenrikssonM,AroH,LonnqvistJ.Recentlifeeventsandcompletedsuicideinbipolaraffectivedisorder:acomparisonwithmajordepressivesuicides,J.Affect.Disorders:99±106.11.AnthonyJC,FolsteinM,RomanoskiA,VonKorffM,NestadtG,ChahalR,MerchantA,BrownC,ShapiroS,KramerM,GruenbergE.ComparisonoftheLayDiagnosticInterviewScheduleandastandardizedpsychiatricdiagnosis.Arch.Gen.Psychiatry12.RobinsL,HelzerJ,CroughanJ,RatcliffK.TheNIMHDiagnosticInterviewSchedule:itshistory,characteristics,andvalidity,'Gen.Psychiatry:381±389.13.HelzerJ,RobinsL,McEvoyL,SpitznagelE,StoltzmanR,FarmerA,BrockingtonI.Acomparisonofclinicalanddiagnosticinterviewschedulediagnoses,Arch.Gen.Psychiatry:657±666.14.KillingsworthMR.Laborsupply.Cambridge:CambridgeUniversityPress,1983.