Using the GHQ-12 to screen for mental health problems ...

The General Health Questionnaire (GHQ) is a self-administered screening tool designed to detect current state mental disturbances and ... Skiptomaincontent Advertisement SearchallBMCarticles Search UsingtheGHQ-12toscreenformentalhealthproblemsamongprimarycarepatients:psychometricsandpracticalconsiderations DownloadPDF DownloadPDF Research OpenAccess Published:10August2020 UsingtheGHQ-12toscreenformentalhealthproblemsamongprimarycarepatients:psychometricsandpracticalconsiderations S.G.Anjara  ORCID:orcid.org/0000-0002-1024-48991,C.Bonetto2,T.VanBortel1&…C.Brayne1 Showauthors InternationalJournalofMentalHealthSystems volume 14,Article number: 62(2020) Citethisarticle 26kAccesses 16Citations 7Altmetric Metricsdetails AbstractBackgroundThisstudyexploresthefactorstructureoftheIndonesianversionoftheGHQ-12basedonseveraltheoreticalperspectivesanddeterminesthethresholdforoptimumsensitivityandspecificity.Throughafocusgroupdiscussion,weevaluatethepracticalityoftheGHQ-12asascreeningtoolformentalhealthproblemsamongadultprimarycarepatientsinIndonesia.MethodsThisisaprospectivestudyexploringtheconstructvalidity,criterionvalidityandreliabilityoftheGHQ-12,conductedwith676primarycarepatientsattending28primarycareclinicsrandomisedforparticipationinthestudy.Participants’GHQ-12scoreswerecomparedwiththeirpsychiatricdiagnosisbasedonface-to-faceclinicalinterviewswithGPsusingtheCIS-R.ExploratoryandConfirmatoryFactorAnalysesdeterminedtheconstructvalidityoftheGHQ-12inthispopulation.TheappropriatethresholdscoreoftheGHQ-12asascreeningtoolinprimarycarewasdeterminedusingthereceiveroperatingcurve.Priortodatacollection,afocusgroupdiscussionwasheldwithresearchassistantswhopilotedthescreeningprocedure,GPs,andapsychiatrist,toevaluatethepracticalityofembeddingscreeningwithintheroutineclinicprocedures.ResultsOfallprimarycarepatientsattendingtheclinicsduringtherecruitmentperiod,26.7%agreedtoparticipate(676/2532consecutivepatientsapproached).Theirmedianagewas46(range18–82 years);67%werewomen.ThemedianGHQ-12scoreforourprimarycaresamplewas2,withaninterquartilerangeof4.TheinternalconsistencyoftheGHQ-12wasgood(Cronbach’sα = 0.76).Fourfactorstructureswerefittedonthedata.TheGHQ-12wasfoundtobestfitaone-dimensionalmodel,whenresponsebiasistakenintoconsideration.ResultsfromtheROCcurveindicatedthattheGHQ-12is‘fairlyaccurate’whendiscriminatingprimarycarepatientswithindicationofmentaldisordersfromthosewithout,withaverageAUCof0.78.TheoptimalthresholdoftheGHQ-12waseither1/2or2/3pointdependingontheintendedutility,withaPositivePredictiveValueof0.68to0.73respectively.Thescreeningprocedurewassuccessfullyembeddedintoroutinepatientflowinthe28clinics.ConclusionsTheIndonesianversionoftheGHQ-12couldbeusedtoscreenprimarycarepatientsathighriskofmentaldisordersalthoughwithsignificantfalsepositivesifreasonablesensitivityistobeachieved.Whileitinvolvesadditionaladministrativeburden,screeningmayhelpidentifyfutureusersofmentalhealthservicesinprimarycarethatthecountryiscurrentlyexpanding. BackgroundIn2015,Indonesiahadonly773psychiatristsfor250millionresidents[1].ThisshortageofspecialistmentalhealthprofessionalsissharedbymostLow-andMiddle-IncomeCountries(LMICs).Thisisreflectedinthetreatmentgapandlowproportionofpeoplewhoreceiveadequatementalhealthcarefortheirneeds.WhilethemedianworldwideTreatmentGapforpsychosisis32.2%[2],thetreatmentgapinIndonesiaismorethan90%[3].Mentalhealthproblemsareestimatedtobepresentinaround20–36%ofpatientsattendingprimarycaresettingsandwhenuntreated,resultinsignificantsufferingandgrowinghealthcarecosts[4,5].ImprovingwaystoidentifypeopleatriskofmentalhealthproblemsisafeasiblestrategytohelpbridgetheTreatmentGapandreducetheirsuffering[6].Embeddingascreeningprocedureintoprimarycarecouldhelpearlyidentification,intervention,andpreventionofcommonmentaldisorders,includinganxietyanddepression[7].Screeningscalesallowforamoresystematicassessmentofself-reportedmentalhealthproblems.Forascreeningproceduretobeeffective,areliablescreeninginstrumentisnecessary,anditsoptimalthresholdneedstobedetermined.Screeningalonecannotandwillnotimprovetheoutcomesforcommonmentaldisorderssuchasdepression,ifresourcesforeffectiveinterventionmustalsobeinplace[8].InIndonesia,mentalhealthservicesareincreasinglyprovidedatzeroorverylowcostsinprimarycarefollowingthesystematicintroductionoftheWorldHealthOrganization(WHO)MentalHealthGapActionProgrammeto10,000primarycareclinics[9].TheGeneralHealthQuestionnaire(GHQ)isaself-administeredscreeningtooldesignedtodetectcurrentstatementaldisturbancesanddisordersinprimarycaresetting[10].TheGHQhasbeentranslatedinto38languagessinceitsdevelopment,indicatingitsfacevalidityacrosscultures[11].WhiletheGHQwasoriginallydevelopedasa60-itemquestionnaire,severalabridgedversions(30-item,28-item,20-item,and12-item)arecurrentlyavailable.The12-itemversionwasadoptedasascreeningtoolinamulti-countryWorldHealthOrganization(WHO)studyofmentaldisordersinprimarycaresetting,asitwasconsideredthebestvalidatedamongsimilarinventories[12,13,14].Thetwelve-itemGeneralHealthQuestionnaire(GHQ-12)isintendedtoscreenforgeneral(non-psychotic)mentalhealthproblemsamongprimarycarepatients[12].ItemsontheGHQ-12areratedona4-pointscaleusingatimeframeof“inthelasttwoweeks.”TherearethreewaysofscoringtheGHQ-12:thebimodalGHQscoringmethod(0-0-1-1)recommendedbythetestauthorsforuseinclinicalsettings;andtheLikertscoringmethod(0-1-2-3)whichiscommonlyusedinresearch,andtheC-GHQscoringmethodwherepositivelyphraseditemsarescored(0-0-1-1)andnegativelyphraseditems(0-1-1-1).AreviewofinternationalvaliditystudiesofGHQ-12conducted20 yearsago,includinginLMICs,reportedthattheoptimalthresholdvariedfrom1/2to6/7,withthemostcommoncut-offbeing2/3[12].Considering17moreinternationalstudiesrevealedarangeofthresholdsfrom0/1to5/6[15].Table 1showslaterstudies,andtheirdistributionofthresholds[4,7,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36].Thesedifferencesmaybetheresultofvaryingprevalenceratesofmentaldisordersandcomorbidity,aswellasthepopulationsinwhichthescalewasadministeredandculturalinfluences[37].Table 1ASampleofGHQ-12ThresholdStudiesonVariousClinicalPopulationsafter1998FullsizetableThefirstGHQ-12validityandreliabilitystudyinIndonesiawaspublishedin2006,whereGHQ-12wascomparedagainstSymptomChecklist(SCL-90)asthegoldstandard,inacommunity-basedprevalencestudy[38].AConfirmatoryFactorAnalysis(CFA)foundtheIndonesianversionoftheinstrumenttohavetwofactors:psychologicaldistressandsocialdysfunction.Sincethen,theIndonesianlanguageversionoftheGHQ-12hasbeenextensivelyusedinnumerousresearchstudies.AmorerecentstudyexaminedthevalidityoftheGHQ-12asascreeningtoolforAdjustmentDisorderinIndonesianprimarycaresetting[39].ThisstudyshowsthattheGHQ-12isvalidandreliableforusewithadjustmentdisorder,Cronbach’sα = 0.863forLikertscoringand0.841forbimodalscoring.ForAdjustmentDisorder,sensitivityandspecificityforGHQ-12were.81and0.62(fortheoptimumcut-offpoint ≥ 11inLikertscoringmethod),0.81and0.57(fortheoptimumcut-offpoint ≥ 2inbimodalscoringmethod).ThestudyfurtherconductedCFAsofthedifferentscoringmethods,eachfindingagreementwithdifferentexistingtheoreticalmodels.ThisstudyaimstoexaminethepsychometricsandpracticalityofusingGHQ-12toscreenforcommonmentalhealthproblemsamongIndonesianadultprimarycarepatients.Thefeasibilityofthescreeningprocedurewillbeevaluatedbyembeddingitintoroutinepatientflowfor2 weeksinapilotstudy,followedbyafocusgroupdiscussionwithstakeholdersinvolvedintheimplementation.Cronbach’salphawillindicatethescale’sinternalconsistency.CFAswillbeusedtodetermineconstructvalidityasusedinpreviousstudies[40].ReceiverOperatingCharacteristic(ROC)curveshavebeenwidelyusedtodescribeandcomparetheperformanceofdiagnosticalgorithms[41]andwillbeusedtodeterminethemostappropriatethresholdscore.MethodsContextThereareapproximately10,000state-ownedprimarycareclinicsinIndonesia,providingfreeaccesstomedicalanddentalcareforresidentsofeachclinic’scatchmentarea.Theseclinics,calledPuskesmas,alsoprovidecareatanominalfeefornon-residents.Thisstudyrecruitedparticipantsfrom28PuskesmasinYogyakarta,Indonesia,aspartofapre-studyofaclusterrandomisedcontrolledtrial[9].These28Puskesmasprovidementalhealthservices.AllPuskesmasintheprovincehavereceivedISOaccreditationstandardisingtheirpatientflowandadministrativeprocedures,makingitpossibletoembedauniformscreeningprocedureacrosstheclinics.DesignThisisacrosssectionalstudyconductedtotestthevalidityandscreeningaccuracyoftheGHQ-12anddeterminethepointatwhichthebalancebetweensensitivityandspecificityisoptimised.Thisstudypilotedtherecruitmentproceduresforatrialexaminingtheclinicalandcost-effectivenessoftwomentalhealthcareframeworksforprimarycare[9].ApilotstudywasconductedinJune2016totestthescreeningprocedure.EthicsEthicsapprovalforthestudyandlargertrialwasgrantedbytheUniversityofCambridgePsychologyResearchEthicsCommittee(referencenumberPRE.2015.108)andUniversitasGadjahMada(referencenumber1237/SD/PL.03.07/IV/2016).Trialinsurancefurthercoversinvestigatorsandresearchparticipants(UniversityofCambridgeTrialInsurancereferencenumber609/M/C/1510).PermissiontoconductresearchattheProvinceofYogyakartaincludingitsallfivedistrictswasobtainedfromtheProvincialGovernmentOffice(referencenumber070/REG/V/625/5/2016).Additionalpermitswerealsoobtainedfromeachofthefivedistricts.Ethicsapprovalfromindividualclinics(Puskesmas)werenotrequiredasallclinicsarefundedandmanagedbydistrictgovernments.Thetrialwhichthisstudywasembeddedinhasbeenregisteredwithclinicaltrials.govsince25February2016,NCT02700490.ParticipantsParticipantswereprimarycareattendeesrecruitedoveraperiodof2 weeksinDecember2016.ThesepatientspresentwithphysicalailmentsattheadultgeneralcareclinicofthePuskesmas.PatientspickupaqueuenumberandaGHQ-12form,whichtheyself-completedwhilewaitingforroutinebloodpressurechecks.PatientsweretheninvitedtotakepartinthestudyregardlessoftheirGHQ-12score.From2532consecutiveprimarycarepatientswhocompletedtheGHQ-12,26.7%(676)consentedtoadditionalin-depthpsychiatricinterview.Theinterviewswereconductedbyageneralmedicalpractitioner(GP)blindedtotheirpatients’GHQ-12score.MeasuresGeneralHealthQuestionnaire(GHQ-12)TheprimarymeasurebeingassessedforitsscreeningaccuracyistheBahasaIndonesiaversionoftheGHQ-12.Priortopatientrecruitment,theleadauthor(SGA)reviewedtheitemswiththe28cliniciansfromparticipatingsitestoensurecontentandsemanticvalidity.Thesameversionhadbeenusedinpreviousvalidationstudieswithvariousclinicalpopulations.IntheBahasaIndonesiaversion,items2,5,6,9,10,and11arenegativelyphrased.Thisstudytookplacein‘reallife’clinicalsetting,suggestingtheappropriatenessofthebimodalscoringmethod(0-0-1-1).AsthisstudyaimstoexaminetheadequacyoftheGHQ-12asascreeningtool,lifetimediagnoseswerenottakenintoconsideration.Instead,currentmentalhealthstatuswasevaluated.ClinicalInterviewSchedule-Revised(CIS-R)Fortheevaluationofmentalhealth,GPsusedtheClinicalInterviewSchedule-Revised(CIS-R)[42],followingtheprotocolofsimilarvaliditystudiesinItaly,England,Brazil,andChile[15].TheCIS-R[42]isafullystructureddiagnosticinstrumentthatwasdevelopedfromanexistinginstrument,theClinicalInterviewSchedule(CIS),designedtobeusedbyclinicallyexperiencedinterviewers[43].TheCISwasrevisedanddevelopedintoafullystructuredinterviewtoincreasestandardisationandtomakeitsuitabletobeusedbytrainedlayinterviewersinassessingminorpsychiatricmorbidityinthecommunity,generalhospital,occupationalandprimarycareresearch.AstheCIS-Rspecificallydiagnosesmoodandanxietydisorders,participantswithindicationofotherdisorders(psychosis,sleepdisorders,dementia)wereaskedadditionalquestionswhichenabledtheinterviewerstoestablishanICD-10diagnosis.Foroursample,interviewswereconductedbyGPs.ThepsychiatricdiagnosticcriteriaoftheICD-10arewidelyusedintheIndonesianhealthsystemastheIndonesianmanualfordiagnosingpsychiatricdisorders(PedomanPanduanDiagnosaGangguanJiwa)releasedin1993andusedbymedicaldoctorsandpsychologists,wasatranslationandadaptationoftheICD-10releasedbytheWHOin1992.DataanalysisIBMSPSSversion24.0andIBMSPSSAmosversion24.0wereusedtoconducttheConfirmatoryFactorAnalysis(CFA)andROC.Exploratoryfactoranalysis(EFA)wasfirstconductedwiththesamedataset,toexplorewhetherthedatawouldreplicateeithertheone,two,orthree-factorsolutionspreviouslyreported.TheEFAyieldedathree-factorsolution,whichwehavelabelleddistress,anxiety,andsocialfunction.ThismodelwasfurthertestedinthesubsequentCFA.ConsistentwithpreviousEFAanalysis,theprincipalcomponentsmethodwasused,withorthogonal(Varimax)rotation.FollowingtheEFA,fourmodelsweretestedforgoodnessoffit(CFA): 1. Three-dimensional:asindicatedbytheEFA,theGHQ-12wasmodelledasameasureofthreelatentvariables(distress,anxiety,andsocialfunction). 2. One-dimensional:theGHQ-12wasmodelledasameasureofoneconstruct(psychiatricmorbidity)usingall12items.Themodelindicatesonelatentvariablewithtwelveindicatorvariables,eachwithitsownerrorterm. 3. Two-dimensional:theGHQ-12wasmodelledasameasureoftwolatentvariables(psychologicaldistressandsocialdysfunction)asfoundinapreviousvalidationstudyinIndonesia[38].Themodelindicatesitems2,5,6,9,10,and11correspondtopsychologicaldistress,whiletherestcorrespondtosocialdysfunction. 4. One-dimensionalwithcorrelatederrors:theGHQ-12wasmodelledasameasureofoneconstructbutwithcorrelatederrortermsonthenegativelyphraseditems,modellingresponsebias[44].Thismodelisidenticaltomodel2,butwithcorrelationsspecifiedbetweentheerrortermsonthenegativelyphraseditems. FollowingtheCFA,aROCanalysiswasconducted.TherequiredsamplesizeforaprospectiveROCstudyofasinglediagnostictest[45]allowingatypeIerrorof0.05andapowerof0.80,withthemoreconservativeAUC1of0.80,AUC0of0.70,andtheallocationratioof4(prevalenceofcommonpsychiatricdisordersisestimatedtobe20%intheprimarycarepopulation,thustheprevalenceofnon-diseasedisestimatedat80%)was370subjects(74clinicallyconfirmedcasesand296clinicallyconfirmednon-cases).TheROCcurveanalysisisacommonlyusedmethodforvisualisingperformanceabilityandgroupingclassification[46].TheROCanalysisplotsatest’struepositiverate(sensitivity)againstitsfalsepositiverate(1-speficity)[47].TheareaunderaROCcurverepresentstheprobabilitythatarandomlychosensubjectiscorrectlyratedorrankedwithgreatersuspicionthananon-diseasedsubject[48].Theareaunderthecurve(AUC)rangesfrom0.5formodelswithnodiscriminationability,to1formodelswithperfectdiscriminationability[49].AROCcurvethatisnearthepointofperfectclassification(upperleftcorneroftheROCspace)isconsideredsuperiorfordetectionperformance[50].Inaddition,thepositivepredictivevalue(PPV)describestheproportionofallpositiveresultsthatarecorrect;whilethenegativepredictivevalue(NPV)describestheproportionofallnegativeresultsthatarecorrect.Thesepredictivevaluesaredependentontheprevalenceofmentaldisordersinthestudysample[51].TotalGHQ-12scoreswereutilisedasthetestvariablefortheROCanalysis.ThegoldstandardagainstwhichtheGHQ-12wastestedwasthepresenceofdiagnosisfollowinganin-depthpsychiatricinterviewusingtheCIS-R.Two-by-twocontingencytableswerecreatedbycross-tabulatingdiagnosticoutcomes(thepresenceorabsenceofanymentaldisorders)andtheGHQ-12screeningoutcomes(positiveornegativescreeningontheGHQ-12).PilotstudyandfocusgroupdiscussionThepilotstudywasconductedoveraperiodof1 weekinJune2016.Trainedandvettedresearchasistantscheckedinfordutyeverymorningat7a.m.AtallyofthenumberofscreeningscompletedwascheckedagainstPuskesmasattendanceattheendofeveryday,whichenabledthecalculationofthepercentageofadultprimarycareattendeesscreened.Intotal,5341patientswerescreenedwithinthepilotperiod.Attheendofthepilot,stakeholderswhowereinvolvedinthescreeningprocessandapsychiatrist(expertinculturalpsychiatry)wereinvitedtoparticipateinafocusgroupdiscussion(FGD)todiscussthechallengesofimplementingthescreeningprocedure,scoring,operationalburden,andinformingpatientsoftheoutcomes.Intotal,sixGPsandresearchassistantsparticipatedintheFGD,whichtookplaceinSeptember2016.TheFGDwassemi-structuredandexploredthefollowingtopics: Primarycarepatients’comprehensionofthescreeningquestionnaire; Feasibilityofthescreeningprocedureaccordingtotheflowofpatientsintheclinics; Commonissuesencounteredduringthescreeningprocess; Generalfeedbackaboutprovidingmentalhealthservicesinprimarycare. AstwoGPsdeclinedtohavetheFGDrecorded,aresearcherwastakingnotesduringtheFGDprocess.Thenoteswerediscussedwithotherco-authorsandanalysedforthepurposeofensuringthefeasibilityofthescreeningprocess.DuringtheFGD,itbecameclearthatwhilethescreeningprocedurelargelyworked,olderpatientsrequiredhelpwithreadingthescreeningquestionnaire.Patientspickedupthescreeningquestionnairealongsideaqueuenumberattheregistrationcounter,filledthequestionnairewhilewaitingforroutinebloodpressurecheck(alladultpatientsarerequiredtopassthroughthebloodpressurecounter).Astaffnursecheckingpatients’bloodpressurecouldassessthescreeningquestionnairevisuallyastheGHQscoringmethod(0-0-1-1)requirednoadvancedarithmetic.Theclinicsgenerallyhaddifficultykeepingtheirpensaspatientsaccidentallytookthemhome.ItwasevidentthatGPsrequiredbetween20and60 minmorewitheachpatientwhoscreenedpositive,creatingalongqueueinthewaitingrooms.GPsreportedthatastheygetusedtoaskingpatientsabouttheirmentalhealthsymptoms,theadditionalinterviewscouldbecomequicker.Whenpatientswereaskedtoreturnforanin-depthpsychiatricinterviewatalaterdate,unfortunatelymostdidnotreturn.ResultsSamplecharacteristicsParticipantswereagedbetween18and82 yearsold(median46).Fromthe2532primarycarepatientsapproached,676consentedtoparticipate(452women;224men).Medianandinterquartilerangeforwomenwere2and4,andformen2and3.Thedifferenceinmedianscoresbetweenwomenandmenwasnotsignificant(Mann–WhitneyU = 47,981.50,p = 0.253).Thetablebelowpresentsparticipants’demographiccharacteristics(age,maritalstatus,educationlevel),aswellastheirGHQ-12scoresbygender.(Table 2).Table 2Totalandbygendersocio-demographiccharacteristicsandGHQ-12scores(0-0-1-1scoring)FullsizetableAlmostoneinfive(19%)hadonlycompletedelementary-leveleducation.Afurther21%completedJuniorHighSchool,and37.9%completedahighschooldiploma.Therest(22.1%)completedundergraduateorpostgraduatedegrees.Fewerthan5%receivedlessthan6 yearsofformaleducation.Table 3showstheprevalenceofICD-10psychiatricdiagnosesandGHQ-12medianscoresforadultIndonesianprimarycarepatients.Forthosewithaseveredepressiveepisode,theGHQ-12medianscorewas10,withaninterquartilerangeof7.ForthosewithComorbidAnxietyandDepression,theGHQ-12medianscorewas3,withaninterquartilerangeof3.ForthosewithgeneralanxietydisordertheGHQ-12medianscorewas6,withaninterquartilerangeof9.Table 3TotalandbygenderprevalenceofpsychiatricdiagnosesandmedianGHQ-12scores(bimodalscoring)ofrespondentsinterviewedwithCIS-RandfurtherclinicalinterviewsFullsizetableMedianscoresforthosewithadiagnosis(cases)comparedtothosewhodonotmeettheICD-10diagnosticcriteria(non-cases)areshowninTable 4.Table 4GHQ-12meanandmedianscoresfornon-casesvs.casesmeetinganyICD-10diagnosticcriteriaduringsamplingperiod,Bimodalscoring(0-0-1-1)FullsizetableTheGHQ-12medianforcases(48%)was3,withaninterquartilerangeof3,andthemedianfornon-caseswas1,withaninterquartilerangeof2.Thegroupmeetingdiagnosticcriteriahadsignificantlyhighermedianscoresthanthosewithoutdiagnosis(Mood’sMedianTestχ2 = 111.07,df = 1,p