RESEARCHARTICLE Mental health among children and adolescents: Construct validity, reliability, and parent-adolescent agreement on the ‘Strengths and Difficulties Questionnaire’ in Chile JorgeGaete1,2*,JesusMontero-Marin3,DanielaValenzuela4,CristianA.Rojas-Barahona5, a1111111111 EsterbinaOlivares6,RicardoAraya7 a1111111111 a1111111111 1 DepartmentofPublicHealthandEpidemiology,FacultyofMedicine,UniversidaddelosAndes,Santiago, a1111111111 Chile,2 DepartmentofPopulationHealth,LondonSchoolofHygieneandTropicalMedicine,London,United a1111111111 Kingdom,3 PrimaryCarePreventionandHealthPromotionResearchNetwork(RedIAPP),Centrode Investigacio´nBiome´dicaenReddeSaludMental,CIBERSAM,Zaragoza,Spain,4 SchoolofPsychology, UniversidaddelosAndes,Santiago,Chile,5 FacultyofEducation,PontificiaUniversidadCato´licadeChile, Santiago,Chile,6 SchoolofNursing(CampusSanFelipe),UniversidaddeValpara´ıso,SanFelipe,Chile, 7 CentreforGlobalMentalHealthandPrimaryCareResearch,HealthServiceandPopulationResearch Department,InstituteofPsychiatry,Psychology,andNeuroscience,King’sCollegeLondon,London,United OPENACCESS Kingdom Citation:GaeteJ,Montero-MarinJ,ValenzuelaD, *[email protected] Rojas-BarahonaCA,OlivaresE,ArayaR(2018) Mentalhealthamongchildrenandadolescents: Constructvalidity,reliability,andparent-adolescent Abstract agreementonthe‘StrengthsandDifficulties Questionnaire’inChile.PLoSONE13(2): e0191809.https://doi.org/10.1371/journal. TheStrengthsandDifficultiesQuestionnaire(SDQ)isascreeningtoolusedtomeasure pone.0191809 psychologicalfunctioningamongchildrenandadolescents.Ithasbeenextensivelyused Editor:AndreaMartinuzzi,IRCCSE.Medea,ITALY worldwide,butitspsychometricproperties,suchasinternalstructureandreliability,seemto varyacrosscountries.Thisisthefirststudyexploringtheconstructvalidityandreliabilityof Received:March16,2017 theSpanishversionofSDQamongearlyadolescents(self-reported)andtheirparentsin Accepted:January11,2018 LatinAmerica.Atotalof1,284earlyadolescents(9–15years)andtheirparentsanswered Published:February5,2018 theSDQ.Wealsocollecteddemographicvariables.Aconfirmatoryfactoranalysiswascon- Copyright:©2018Gaeteetal.Thisisanopen ductedtoassessthelatentstructureoftheSDQ.Wealsousedthemultitrait-multimethod accessarticledistributedunderthetermsofthe analysistoseparatethetruevarianceontheconstructsfromvarianceresultingfrommea- CreativeCommonsAttributionLicense,which surementmethods(self-reportvs.parentreport),andevaluatedtheagreementbetween permitsunrestricteduse,distribution,and reproductioninanymedium,providedtheoriginal adolescentsandtheirparents.Wefoundthattheoriginalfive-factormodelwasagoodsolu- authorandsourcearecredited. tionandtheresultingsub-scaleshadgoodinternalconsistency.Wealsofoundthattheself- DataAvailabilityStatement:Dataarefromthe reportedandparentalversionsofSDQprovidedifferentinformation,whicharecomplemen- AKA-EDU/15andFONIS-EVS-13I0100studies taryandprovideabetterpictureoftheemotional,social,andconductproblemsofadoles- whosePrincipalInvestigatormaybecontactedat cents.WehaveaddedevidencefortheconstructvalidityandreliabilityoftheSpanishself- [email protected] reportedandparentalSDQversionsinaChileansample. providedrestrictionstosharethedatapublicly becauseitcontainssensitiveinformationregarding mentalhealthofchildrenandadolescents.For furtherinformationregardingrestrictionstoshare thedatapublicly,pleasecontactthePresidentof theCommitteeofEthicsofUniversidaddelos Andes,JoaquinGarcia-Huidobro:[email protected]. PLOSONE|https://doi.org/10.1371/journal.pone.0191809 February5,2018 1/18 ValidityandreliabilityoftheSDQforadolescentsinChile Funding:Thisworkwassupportedby:1.Joint Introduction project—NewLearningEnvironmentsand Mentalhealthisanimportantproblemworldwide,andChileisnoexception.Ingeneral,the Technologies—AKA-EDU/15,AcademyofFinland (http://www.aka.fi/en)andCONICYT,Chile(http:// maincauseofyearslostbecauseofdisabilityamongthepopulationaged10–24yearsoldis www.conicyt.cl)toJG;2.ChileanNationalGrantfor neuropsychiatricdisorder(45%)[1].Mentalhealthproblemshaveanimportanteffect,not ResearchandDevelopmentinHealth—FONIS- onlyonthelivesoftheaffectedindividuals,butalsoonthepeoplearoundthem(family, EVS-13I0100,CONICYT,Chile(http://www.conicyt. school,peers).Mostofthepsychiatricproblemsweseeinadultlifestartduringchildhoodor cl)toJG.Thefundershadnoroleinstudydesign, adolescence[2].RecentreportsshowthattheprevalenceofpsychiatricdisordersamongChil- datacollectionandanalysis,decisiontopublish,or eanadolescentsreaches16.5%–18.2%[3,4],butonlyaminorityofaffectedadolescentsmake preparationofthemanuscript. contactwiththespecializedservicestheyrequire. Competinginterests:Theauthorshavedeclared Detectingmentalhealthproblemsearlyincreasesthechancesofreducedburdenandfuture thatnocompetinginterestsexist. complications[5].Inrecentyears,interestindevelopinginstrumentstodetecttheseemotional andbehaviouraldisordershasbeenincreasing.Moreover,theWorldHealthOrganizationhas repetitivelyadvocatedusingscreeningtoolsinprimarycaresettingsandschools[6–8].These instrumentsshouldbevalidandreliableforassessingmentalhealthamongadolescents,but shouldalsobeshort,simple,andculturallyadapted/tested. Themostwidelyusedscreeningtestsfordetectingpsychologicalandbehaviouralproblems amongyoungpeoplearetheChildBehaviorChecklist(CBCL),thePediatricSymptoms Checklist(PSC),andtheStrengthsandDifficultiesQuestionnaire(SDQ).TheCBCLconsists of113questionsonspecificproblematicbehaviourstobeansweredbytheparentsorsurro- gatesofschool-agedchildrenbetween6and18yearsold.Theitemscanbeorganizedinto threemainscales(totalproblems,internalizingproblems,andexternalizingproblems)and eightsubscales(withdrawn,somaticcomplaints,anxious/depressed,socialproblems,thought problems,attentionproblems,delinquentbehaviour,andaggressivebehaviour)[9,10].The PSCconsistsof35questionsaboutarangeofemotionalandbehaviouralproblemsreported byparentsofchildrenandadolescentsbetween4and18yearsold[11,12].Theseitemscanbe organizedintothreesub-scales:attentionproblems,internalizingproblems,andexternalizing problems.TheSDQconsistsof25itemsaboutanumberofemotionalandbehaviouralprob- lemsexperiencedbychildrenagedbetween4and16years.Ithasfivesubscales[13]:emotional symptoms,conductproblems,hyperactivity/attentionproblems,peerproblems,andpro- socialbehaviour.Thefirstfoursub-scalescanbecombinedintoatotaldifficultiessub-scale, whilethelastisconsideredthestrengthssub-scale.Ithasthreeversions,tobeanswered, respectively,byparents/caregivers,byteachers,andbythesubjectsthemselvesifover11years old;however,theself-reportedversionhasalsobeenusedinyoungerchildren(8–13yearsold) withgoodpsychometricresults[14].TheSDQoffersseveraladvantagesovertheotherinstru- ments:1)itisabriefinstrument,andthuslessburdensome;2)itprovidesinformationfrom threedifferentsourcesusingasimilaritemstructure,allowingcomparisonofresultsanda morecomprehensiveoverallassessmentofchildrenandadolescents’mentalstatus;and3)it assessbothstrengthsanddifficultiesatthesametime. Anotherimportantconsiderationwhenselectinganinstrumentistheavailabilityofstudies exploringitsvalidityandreliabilityinapopulationofthetargetage.Forinstance,inChile, thereisastandardizedversionoftheCBCLforchildrenbetween6and11yearsold[15].Itis worthmentioningthatthiscountry-specificversionisnotbasedontheupdatedgeneralver- sionavailabletoday.AChileanversionofthenewCBCLhasbeenvalidated,butforapopula- tionof1½to5-year-oldchildren[16].ThePSChasalsobeenvalidatedinChile,ina populationofsocioeconomicallyvulnerablestudentsinthefirstgradeofprimaryschool(aged 6years)[17].TheChileanversionofthePSCconsistsof33itemsandhasbeenextensively usedforschool-agedpre-adolescentstudents(under12yearsold)[18–21].However,valida- tionstudiesforadolescentsarelacking.Finally,thepsychometricpropertiesoftheSDQhave PLOSONE|https://doi.org/10.1371/journal.pone.0191809 February5,2018 2/18 ValidityandreliabilityoftheSDQforadolescentsinChile beenevaluatedinChilefortheparent-answeredversionforchildrenbetween4and11years old.Therefore,tothebestofourknowledge,nostudieshaveexploredtheconstructvalidity, reliability,anddegreeofagreementbetweentheSDQself-reportandtheparentalreportfor Chileanadolescents.Additionally,weknowofnoSDQstudiesforadolescentsfromSpanish- speaking,Latin-Americancountries. ThepsychometricpropertiesoftheSDQself-reportandparentalformatshavebeenwell establishedforadolescentsinseveralcountries,suchastheUnitedKingdom[22],Germany [23,24],Netherlands[25],Nordiccountries[26],Russia[27],Australia[28],Austria[29], India[27],Yemen[27,30],Bangladesh[27,31],China[32,33],Brazil[27],andotherEuro- peancountries[34].Theonlystudiesexploringtheconstructvalidityandreliabilityofthe SpanishversionoftheyoungandparentalformatsofthisinstrumentcamefromSpain[35, 36]. Studieshaveconfirmedthefivetheoreticaldimensionsintheadolescentself-reportedver- sionandintheparentalversionofthequestionnaire,usingexploratoryandconfirmatoryfac- toranalyses[37,38].However,otherstudieshavefailedtoreplicatetheoriginallypostulated five-factorsolution[39–42].Additionally,theyhaveproposedathree-factorsolution[41], combiningtheconductandhyperactivity/attentionproblemsasan‘externalizing’dimension andtheemotionalandpeerproblemsasan‘internalizing’dimension,whilekeepingthepro- socialsub-scaleasaseparatefactor[43].Furthermore,despiteevidencethatthefive-factor modelfitswellacrossgenderandethnicgroupsforyoungchildren[44],astudygathering informationfromfiveEuropeancountriesfoundthatthenumberoffactorscouldbecountry- dependentinthecaseofadolescents[40],andaNorwegianstudyfoundthatfactorloadings weredifferentbetweenpre-adolescentsandadolescents[45].Regardingreliability,somestud- ieshaveshownadequateinternalconsistency[46–49],whileothershavereportedlowvalues forsomesub-scales,especiallyforconductandpeerproblems[50]. ForadolescentsinSpanish-speakingcountriesinLatinAmerica,wehavefoundnostudies exploringtheconstructvalidityoftheSDQbymeansofconfirmatoryfactoranalysis(CFA), usingparentalandself-reporteddata.Asmentionedearlier,wefoundonlytwostudiesusing theSDQinChile,oneexploringthepsychometricpropertiesoftheparentalreportsonchil- drenbetweentheagesof4and11[51],andonepresentingtheresultsofcomparingthescores betweenearly-adolescentAymara[anindigenousSouthAmericannation]andnon-Aymara students,usingtheself-reported,parental,andteacherversions[52]. Therefore,weseeaknowledgegapconcerningtheperformanceoftheSDQinSpanish- speaking,Latin-Americancountries,specifically,itsconstructvalidityandreliabilityforado- lescentpopulations.Theaimsofthepresentstudyare:i)toevaluatecompetingmodelsofthe latentstructureoftheSDQ,usingconfirmatoryfactoranalysis;ii)toexplorethereliabilityof theresultingsub-scaleshavingthebestfit;iii)tocomparethedegreeofagreementbetween adolescentself-reportsandparentalreports,andtheirrespectiveexplanatorypower;iv)topro- videnormativedatafortheSDQadolescentandparentalversions. Methods Participants Forthepurposesofthisstudy,weusedSDQdatafromtwoseparatestudiesinsimilar,school- agedpopulations.Thefirststudy(Study1)isbeingconductedinavulnerableurbanpopula- tioninSanFelipe,asmallcitynorthofSantiago.Thisisanongoinglongitudinalstudyexplor- ingthefactorsassociatedwiththedevelopmentofhealth-promotingbehavioursinearly adolescents.Aspartofthebaselineassessment,weadministeredtheSDQtothestudents(10 to15yearsold)andtheparents.Somepreliminaryresultsofthecross-sectionalanalysishave PLOSONE|https://doi.org/10.1371/journal.pone.0191809 February5,2018 3/18 ValidityandreliabilityoftheSDQforadolescentsinChile beenpublished[53].Thesecondstudy(Study2)gathereddatafromstudents9to15yearsold, aimingtotestthevalidityoftheSDQandoftheChilean-adaptedversionoftheOlweusBully/ VictimQuestionnaireRevised.Thispreliminaryvalidationstudyispartofalarger,ongoing researchcalled‘TheKiVaantibullyingprograminprimaryschoolsinChile,withandwithout thedigitalgamecomponent,arandomizedcontrolledtrial’[54].Wedecidedtopresenthere analyticalresultsfrombothstudiesbecauseStudy1gatheredinformationfromlow-income families,whileStudy2gatheredinformationfromlow-,middle-,andhigh-incomefamilies, allowingthelattersetofresultstobemorerepresentativeoftheadolescentChileanpopulation asawhole. Procedureandethics InStudy1,weinvitedallurban,municipal,state-fundedprimaryschoolsinSanFelipe (n=10)toparticipateafterobtainingauthorizationfromschoolboardauthorities.Allten schoolsagreedtoparticipate.Weinformedtheparentsormaincaregiversaboutthestudyand askedthemtosignandinformedconsenttoallowtheirchildrentoparticipate.Atotalof1,035 parentswerecontacted,682consentedandansweredtheparentalquestionnaire.Onthedayof theassessment,560studentsassentedtoparticipateandansweredthequestionnaire(10did notassent,and112wereabsentthatday).Atotalof488parent-childdyadsprovidedcomplete data.InStudy2,weinvitedfiveschoolstoparticipate.Atotalof1,945parentsormaincaregiv- erswerecontacted,and1,068consentedandansweredtheparentversionoftheSDQques- tionnaire.Onthedayoftheassessment,913studentsassentedtoparticipateandansweredthe questionnaire(50didnotassent,and105wereabsentthatday).Atotalof796parent-child dyadsprovidedcompletedata.SeeFig1:Flowchart.Inbothstudies,theparentalSDQques- tionnairewasansweredbythemaincaregiver,whichmostofthetimewasthemother.Other maincaregiverswerethefatherorothersignificantfamilymembersuchasgrandmother. BothstudieshadbeenapprovedbytheEthicsCommitteeoftheUniversidaddelosAndes. Writteninformedconsentwasobtainedfromparentsandwritteninformedassentfromado- lescents.Thestudyposednorisks. Measures Socio-demographicvariables. Sex(0=male;1=female),age(years)andsocio-economic status(0=low-incomefamilies;1=middle-income;2=high-incomefamilies)werecollected. Thesocioeconomicstatuswasbasedonthecriteriaofthe2009NationalSystemfortheMea- surementofEducationQuality,whichgathersinformationfromparentsormaincaregivers abouttheirhouseholdincome,anditwascollectedfromtheMinistryofEducation. StrengthsandDifficultiesQuestionnaire(SDQ)[55–57]. Thisquestionnairehas25 items,dividedintofivesubscales:emotionalsymptoms,conductproblems,hyperactivity-inat- tentionproblems,peerproblems,andpro-socialbehaviour.Thesefivesubscalescanbeorga- nizedintotwomajorsub-scales:strengths(pro-socialbehaviour)anddifficulties(theother foursubscales).Eachitemusesathree-pointordinalformattobeansweredwithoneofthefol- lowing:0=nottrue;1=somewhattrue;and2=certainlytrue.Fiveoftheitemsarenegatively wordedintheoriginal(i.e.isobedient,thinksbeforeacting,hasgoodattention,hasgood friends,isgenerallyliked).Therefore,forcompatibilityincombiningsubscalesintomajorsub- scales,theirscoreswerereversed.Themeanscoreforeachsubscalewasthencalculated(range 0–10).Allscoresforthedifficultiessubscaleswereaddeduptoatotaldifficultiesscore(range: 0–40).Thescoresonthepro-socialsubscalewereanalysedindependently(range:0–10).The SDQhasbeentranslatedintomorethan50languages[10].WeusedtheauthorizedSpanish PLOSONE|https://doi.org/10.1371/journal.pone.0191809 February5,2018 4/18 ValidityandreliabilityoftheSDQforadolescentsinChile Fig1.Flowchart. https://doi.org/10.1371/journal.pone.0191809.g001 versionandthescoringalgorithmsproposedbyitsauthor(formoreinformationsee:sdqinfo. com). Dataanalysis Forthepurposesofthisarticle,onlyresponderswithvalidanswersonall25itemswere includedintheanalyses. Firstly,wesummarizedthesocio-demographicvariablesandbasicpsychometriccharacter- isticsoftheitemsusingdescriptivestatistics,includingmeans,standarddeviations(SD),and, whennecessary,frequenciesandpercentages.Weusedastructuralequationmodelling PLOSONE|https://doi.org/10.1371/journal.pone.0191809 February5,2018 5/18 ValidityandreliabilityoftheSDQforadolescentsinChile approachtoCFA,toassessthestructureoftheproposedfive-factorandthree-factormodels fortheSDQ,inself-reportsaswellasinparentalreports.MultivariateMardia’scoefficients [58]andpolychoricmatriceswerecalculatedtoevaluatethedistributionoftheitems.We ensuredtheadequacyofthematricesbyassessmentofthedeterminant,bytheKMOindex, andbyBarlett’stest[59].Wealsocalculatedtheinternalconsistencyofeachfactorbyusing McDonald’sOmega(ω),whichcanbeinterpretedasthesquareofthecorrelationbetweenthe scalescoreandthelatentvariablecommontoalltheindicators[60].TheOmegaindex assumesacongenericmodel,whichmeansthatfactorloadingsareallowedtovary,anditalso takesintoaccounttheitem-specificmeasurementerror.Thus,itprovidesamorerealisticesti- mateoftruereliabilitythanclassicalCronbach’sAlphavalues,beingthatbothcanbeinter- pretedusingthesamethresholdcut-offpoints. Weusedtheunweightedleastsquares(ULS)methodforfactorextraction,inviewofits robustness[61].Specifically,TheULSmethoddoesnotprovideinferentialestimationsbased ontheχ2distribution(andthereforedoesnotprovidep-values),butitdoesnotrequireany distributionalassumption;itisrobustandusuallyconvergesbecauseofitscomputationaleffi- ciency;ittendstoprovidelessbiasedestimatesofthetrueparametervaluesthanotherproce- dures;anditshowsgoodperformancewhenworkingwithpolychoricmatrices[62–65].Froma generalperspective,weusedweusedthefitindicesthattheULSreportssuchasthegoodness- of-fitindex(GFI),theadjustedgoodness-of-fitindex(AGFI),thenormed-fitindex(NFI),and theroot-mean-squareofthestandardizedresiduals(RSMR).GFIandAGFIrefertothe explainedvarianceofthemodel,andvalues>0.90areconsideredacceptable[66].TheNFI measurestheproportionalreductionintheadjustmentfunctionwhengoingfromnulltothe proposedmodelandisconsideredacceptablewhen>0.90[67].TheRSMRisthestandardized differencebetweentheobservedandthepredictedcovariance,indicatingagoodfitforvalues <0.08[68].Fromananalyticalperspective,standardizedsaturationsandtheexplainedvariance wereconsidered.WealsousedaCFAapproachtoMTMManalysis[69].Thisapproachpermits separationofthetruevarianceontheconstructsfromvarianceresultingfrommeasurement methods(self-reportvs.parentalreport).Thelogicisthatself-reportandparentalmeasuresof thesameconstructshouldbehighlycorrelated,butthatmeasuresofdifferentconstructsshould havelowcorrelations.Wecalculatedsquaredfactorloadingstoestimatetheexplainedvariance inthesub-scalesresultingfromtheunderlyingtraitandthereportingmethod.Theunexplained variancewastermeduniqueness.Weperformedthesameanalysesaccordingtoage((cid:20)11vs >11)andsocioeconomicstatus(LowvsMiddle/Highincome)toassesspotentialdifferences. Wecalculatedthe25th,50th,75th,and90thpercentilescoresforeachgeneratedsub-scale, forthetotalsampleandforeachsex,forboththeadolescentandtheparentalSDQversions. Wealsopresentthenormativedataforagegroups(cid:20)11and>11yearsold. AllanalyseswereperformedusingSTATA-14,SPSS-19,andAmos-7. Results Thematerialsusedtoproducethefollowingresultswillbeavailableuponrequest,includinga detailedlistofdocumentsandallthedatafilesneededinorderforreplication,aswellasevery stepandthespecificsequencetheinterestedresearchersshouldtakeintoaccounttomakedata available[70].Authorswillpostthereferredmaterialsinthegroup’swebsite,and/orwillbe sendwhenaskedforthem[71]. Characteristicsofparticipants Missingdatafromitemresponsesvariedfrom0.2%to1.0%.Thefinalnumberofrespondents inchildrenandearlyadolescents(agerange:9–15years)andtheirparentswas1,284(54.0%, PLOSONE|https://doi.org/10.1371/journal.pone.0191809 February5,2018 6/18 ValidityandreliabilityoftheSDQforadolescentsinChile Table1. Descriptivefeaturesofparent-childdyadswithcompletedata. Totalsample Study1 Study2 %(95%CI)/Mean(SD) %(95%CI)/Mean(SD) %(95%CI)/Mean(SD) n=1284 n=488 n=796 Sex Girls 46.0(43.3–48.8) 50.4(46.0–54.8) 56.2(52.7–59.6) Boys 54.0(51.2–57.0) 49.6(45.2–54.0) 43.8(40.4–47.3) Age 11.3(1.4) 11.5(1.2) 11.2(1.5) Grade Year4 13.6(11.8–15.5) - 21.9(19.1–24.9) Year5 26.2(23.8–28.6) 34.2(30.1–38.6) 21.2(18.5–24.2) Year6 24.1(21.8–26.5) 32.8(28.8–37.1) 18.7(16.2–21.6) Year7 24.9(22.6–27.4) 33.0(28.9–37.3) 20.0(17.3–22.9) Year8 11.3(9.7–13.1) - 18.2(15.7–21.1) Socioeconomicstatus Low 40.6(37.9–43.3) 100 4.1(3.0–5.8) Medium 21.0(18.9–23.3) 33.9(30.7–37.3) High 38.4(35.8–41.1) 61.9(58.5–65.3) https://doi.org/10.1371/journal.pone.0191809.t001 female).Themeanagewas11.3years(SD=1.4).Studentsattended4th-(13.6%),5th-(26.2%), 6th-(24.1%),7th-(24.9%),and8th-(11.3%)yearprimaryschool.Ofthestudentsample,40.6% camefromlow-incomefamilies,21.0%frommiddle-incomefamilies,and38.4%fromhigh- incomefamilies.Table1showsthegeneralfeaturesofthewholesample,anddividedbystudy. PsychometricsandconstructvalidityoftheSDQ ThepolychoricmatrixoftheSDQitemsusingself-reportdatahadadeterminantof0.008,the KMOtesthadavalueof0.87,Bartlett´stestgave6,167.30(df=300;p<0.001),andMardia’s statisticwas48.24(p<0.001).Incontrast,thematrixoftheparental-reportSDQitemshada determinantof0.001,aKMOvalueof0.87,aBartlett´stestof8,340.10(df=300;p<0.001), andaMardia’sstatisticof82.07(p<0.001). Table2showsthattheCFAfitindicesfortheSDQwerewithinacceptablevaluesonlyin thecaseofthefive-factormodel–whichinfactwasoriginallyproposedfromatheoretical pointofview–andtheywerebetterthaninthecaseofthethree-factormodel,beingthatboth thecorrespondingself-reportandparental-reportindiceswereadequateintheformer. Table2. ConfirmatoryfactoranalysisoftheSDQ. Models CMIN NPAR GFI AGFI NFI RMSR Threefactors Self-report 594.79 53 0.924 0.910 0.875 0.092 Parental 554.35 53 0.932 0.918 0.900 0.101 Fivefactors Self-report 451.05 60 0.943 0.930 0.905 0.081 Parental 387.06 60 0.952 0.941 0.930 0.084 MTMM 704.06 41 0.997 0.988 0.993 0.026 Note:CMIN,Chi-squarestatistic,comparingthetestedmodelandtheindependencemodeltothesaturatedmodel;NPAR,numberofparametersinthemodel;GFI, goodness-of-fitindex;AGFI,adjustedgoodness-of-fitindex;NFI,normed-fitindex;RMSR,rootmeansquareofthestandardizedresiduals;MTMM,multitrait- multimethod. https://doi.org/10.1371/journal.pone.0191809.t002 PLOSONE|https://doi.org/10.1371/journal.pone.0191809 February5,2018 7/18 ValidityandreliabilityoftheSDQforadolescentsinChile Table3. DescriptiveandstandardizedfactorloadingsforSDQitems. Self-report Parental-report ω Mean SD ω Mean SD Emotionalsymptoms 0.76 3.48 2.40 0.81 2.86 2.37 Somatic(3) 0.57 0.71 0.56 0.70 Worries(8) 0.90 0.75 0.52 0.66 Unhappy(13) 0.51 0.68 0.38 0.63 Clingy(16) 0.93 0.77 0.77 0.74 Fears(24) 0.58 0.71 0.63 0.72 Conductproblems 0.76 2.21 1.92 0.85 1.98 1.85 Tempers(5) 0.58 0.71 0.80 0.74 Obedient(7) 0.70 0.59 0.57 0.61 Fights(12) 0.33 0.59 0.25 0.52 Lies(18) 0.40 0.63 0.32 0.52 Steals(22) 0.20 0.50 0.04 0.26 Hyperactivity 0.77 3.91 2.39 0.85 3.69 2.56 Restless(2) 0.79 0.75 0.62 0.72 Fidgety(10) 0.78 0.76 0.61 0.72 Distractible(15) 0.92 0.75 0.93 0.77 Thinksbeforeacting(21) 0.69 0.63 0.77 0.61 Goodattention(25) 0.74 0.62 0.75 0.68 Peerproblems 0.65 2.30 1.93 0.76 1.95 1.84 Solitary(6) 0.40 0.65 0.46 0.68 Goodfriend(11) 0.31 0.63 0.20 0.46 Generallyliked(14) 0.51 0.60 0.31 0.52 Bullied(19) 0.42 0.65 0.33 0.55 Adults(23) 0.66 0.74 0.65 0.70 Pro-socialbehaviour 0.72 7.73 1.80 0.77 8.42 1.64 Considerate(1) 1.61 0.55 1.74 0.48 Shares(4) 1.51 0.60 1.68 0.53 Caring(9) 1.55 0.56 1.70 0.51 Kindtokids(17) 1.71 0.53 1.81 0.46 Oftenvolunteerstohelp(20) 1.36 0.64 1.50 0.59 Note:Valuesaremeans,standarddeviations(SD),andMcDonald’sOmega(ω). https://doi.org/10.1371/journal.pone.0191809.t003 Table3showsthedescriptivestatisticsandMcDonald’sOmegavaluesfortheSDQitems andfactors,andFig2theweightsandcorrelationsbetweenfactorsfortheCFAbothforself- reportandparental-reportscores.Aswecansee,thevaluesandvariabilityofthe‘steal’item arelowinself-reportscores(mean=0.20,SD=0.50),butespeciallysoinparental-report scores(mean=0.04,SD=0.26).Thefactorialloadingsareadequate,althoughthe‘goodfriend’ itemisalowoutlierinthecaseofself-reportscores(w=0.25).Intermsofreliability,parental reportsaremoreconsistentthanself-reports.McDonald’sOmegavaluesforself-reportsrange from0.65(‘peerproblems’)to0.77(‘hyperactivity’)and,inthecaseofparentalreports,from 0.76(‘peerproblems’)to0.85(‘conductproblems’,and‘hyperactivity’). Inself-reportscores,thecorrelationsamongthefiveconstructsarestrongestbetween‘emo- tionalsymptoms’and‘peerproblems’(r=0.78),whileinparentalreports,thestrongestcorre- lationsarebetween‘conductproblems’and‘hyperactivity’(r=0.76).Thelowestcorrelations inself-reportsarebetween‘pro-socialbehaviour’and‘emotionalsymptoms’(r=-0.06),while PLOSONE|https://doi.org/10.1371/journal.pone.0191809 February5,2018 8/18 ValidityandreliabilityoftheSDQforadolescentsinChile Fig2.CFAparametersoftheSDQforself-andparent-reports(n=1,284). https://doi.org/10.1371/journal.pone.0191809.g002 thelowestcorrelationsinparentalreportsarebetween‘hyperactivity’and‘peerproblems’ (r=0.23).SeeFig2. TraitandmethodcomponentsintheMTMMapproach TheCFAapproachtoMTMMwithtwomethodfactorsandfivetraitfactorshasaverygood fittothemodel(Table2).Fig2showstheCFAparametersfortheMTMMapproach,Fig3 showsthestructuralparametersfortheMTMMmodel,andTable4showsthetraitand methodvariancecomponents.Aswecansee,thetraitvariancecomponentssuggestthatself- reportstendtobemorediscriminatingon‘pro-socialbehaviour’(wheretheyarepractically unaffectedbymethod)and‘conductproblems’.Ontheotherhand,parentalreportsseemto beparticularlydiscriminatingonratingsfor‘peerproblems’(wheretheyaresubjecttolow methodeffects)and‘emotionalsymptoms’.Thelargestuniquenessvalueinself-reportsisin ‘emotionalsymptoms’,whileinparentalreportsitisin‘pro-socialbehaviour’.Finally,we foundmoderatelylowcorrelationsbetweenthemethods(r=0.38). WehaveproducedadditionalresultsforCFAandMTMManalysesstratifiedbyage((cid:20)11 vs>11)andsocioeconomicstatus(LowvsMiddle/Highincome).Theseresultsareavailable inS1File. PLOSONE|https://doi.org/10.1371/journal.pone.0191809 February5,2018 9/18 ValidityandreliabilityoftheSDQforadolescentsinChile Fig3.StructuralMTMMmodeloftheSDQforChileanadolescents(n=1,284)usingCFA. https://doi.org/10.1371/journal.pone.0191809.g003 Normativedata. Regardingthepercentilesofthetotaldifficultiesscale,thevaluesare similarbetweengirlsandboys,beinglowerforgirlswithadifferenceofonepoint.SeeTables 5and6forthetotalsample,Tables7and8forparticipantsaged(cid:20)11yearsold,andTables9 and10forparticipantsaged>11yearsold.Thepercentilesintheself-reportedSDQare slightlylowerthanthoseintheparentalSDQ. Table4. TraitandmethodvariancecomponentsfortheMTMManalysis. Measuredvariables Trait Method Uniqueness Self-report Emotionalsymptoms 0.10 0.41 0.49 Conductproblems 0.36 0.34 0.30 Hyperactivity 0.42 0.13 0.45 Peerproblems 0.21 0.33 0.46 Pro-socialbehaviour 0.55 0.01 0.44 Parental Emotionalsymptoms 0.45 0.30 0.25 Conductproblems 0.14 0.52 0.34 Hyperactivity 0.29 0.34 0.37 Peerproblems 0.49 0.10 0.41 Pro-socialbehaviour 0.10 0.12 0.78 https://doi.org/10.1371/journal.pone.0191809.t004 PLOSONE|https://doi.org/10.1371/journal.pone.0191809 February5,2018 10/18