The Function of Personality in Suicidal Ideation from the Perspective of the Interpersonal ...

14 Pages·2017·0.38 MB·English

International Journal of
Environmental Research
and Public Health
Article
The Function of Personality in Suicidal Ideation from
the Perspective of the Interpersonal-Psychological
Theory of Suicide
MarcBaertschi1,2,* ID,AlessandraCostanza3,AlessandraCanuto4andKerstinWeber5
1 ServiceofGeneralPsychiatryandPsychotherapy,NantFoundation,AvenuedesAlpes66,
1820Montreux,Switzerland
2 FacultyofPsychology,UniversityofGeneva,BoulevardduPontd’Arve40,1205Geneva,Switzerland
3 PsychiatryDepartment,SS.AntonioeBiagioeCesareArrigoHospital,ViaVenezia16,
15521Alessandria,Italy;[email protected]
4 ExecutiveandGeneralManagementService,NantFoundation,1804Corsier-sur-Vevey,Switzerland;
[email protected]
5 DivisionofInstitutionalMeasures,MedicalDirection,GenevaUniversityHospitals,
LesVoirons—CheminduPetit-Bel-Air2,1225Chêne-Bourg,Switzerland;[email protected]
* Correspondence:[email protected];Tel.:[phone]
(cid:1)(cid:2)(cid:3)(cid:1)(cid:4)(cid:5)(cid:6)(cid:7)(cid:8)(cid:1)
(cid:1)(cid:2)(cid:3)(cid:4)(cid:5)(cid:6)(cid:7)
Received:15February2018;Accepted:29March2018;Published:30March2018
Abstract: TheInterpersonal-PsychologicalTheoryofSuicide(IPTS)hasbeenincreasinglystudied
overthelastyears,respondingtothedemandforavalidframeworkaddressingsuicidality. Yet,only
afewstudieshaveexploredthefunctionofpersonalityintheIPTSandnonewithclinicalpatients.
We aimed to contribute to fill this gap in investigating the relationship between personality as
conceptualizedbytheFive-FactorModel,theIPTSconstructs,andadimensionalmeasureofcurrent
suicidalideation. Weconductedcorrelation,multiplelinearregression,andpathanalysesbasedon
atrait-interpersonalframeworkinasampleof201individualsvisitingthepsychiatricemergency
roomofageneralhospitalwithcurrentsuicidalideation. Neuroticism(positively)andopenness
(negatively)predictedperceivedburdensomeness,whileneuroticism(positively)andextraversion
(negatively)predictedthwartedbelongingness. Higherconscientiousnessandlowerextraversion
werebothpredictorsoftheacquiredcapabilityforsuicide. However,noneofthemodelsinvolving
pathanalyseswithIPTSvariablesasmediatorsoftherelationshipbetweenpersonalitytraitsand
suicidal ideation was adequately adjusted to the data. Thus, it appears that personality plays a
significantalbeitmodestroleinsuicidalitywhenconsideredfromanIPTSperspective. Aspersonality
isfrequentlyassessedintheclinicalroutine,healthprofessionalsshouldconsideritascomplementary
todetectindividualsatriskoforpresentingsuicidalideation.
Keywords: personality; Five-Factor Model; Interpersonal-Psychological Theory of Suicide;
suicidalideation
1. Introduction
The clinical management of patients who attempted suicide or have suicidal ideation
suffers from the scarcity of comprehensive models accounting for the suicide phenomenon [1].
TheInterpersonal-PsychologicalTheoryofSuicide(IPTS)wasdesignedinthisregardasaframework
proposing that the most dangerous kind of desire for suicide stems from two constructs, namely
feelingsofnotbelongingtoone’ssocialgroup(thwartedbelongingness)andtheperceptionofbeing
aburdenforsignificantothers(perceivedburdensomeness). Additionally,thetheoryspecifiesthat
actingonthisdesireforsuicideisonlypossibleifonehasacquiredthecapacitytodoso,assessedbya
thirdconstructlabeledtheacquiredcapabilityforsuicide[1,2].
Int.J.Environ.Res.PublicHealth2018,15,636;doi:10.3390/ijerph15040636 www.mdpi.com/journal/ijerph
Int.J.Environ.Res.PublicHealth2018,15,636 2of14
Testedinanincreasingnumberofstudiesduringthepastdecade,thepredictionsoftheIPTS
havenotbeenunanimouslysupported[3]. Inthiscontext,furtherresearchonpossibledeterminantsof
thwartedbelongingness(TB),perceivedburdensomeness(PB)andtheacquiredcapabilityforsuicide
(AC)iswarranted. Specifically,ithasbeenpointedoutthattheIPTSdoesnotaccountforthepropensity
toengageinsuicidebehavior. This,inotherwords,referstopossiblepre-existingvulnerabilitiesfor
suicide,whichcouldnotablybeappraisedintermsofpersonality[4]. Toourknowledge,thefounding
textsoftheIPTSdidnotmentionpersonality,althoughitsrolehasbeenconsideredinrecentIPTS
studiesfrombothcategorical[5,6]anddimensionalperspectives[7–14].
Atraitmodelofpersonality,theFive-FactorModel(FFM)[15]hasbeenadoptedasconsensual
framework to determine each person’s individuality as a function of five broad dimensions:
neuroticism (N), extraversion (E), openness (O), agreeableness (A) and conscientiousness (C).
These dimensions have been progressively identified in the second half of the twentieth century
startingwiththepioneeringworkofEysenck[16]whoinitiallyidentifiedthe“BigTwo”N-Epair.
Neuroticismreferstoatemperamentalmeasureofemotionalhealthcomprisingcomponentssuch
as fear, anger and impulsivity. Extraversion relates to variables like warmth, assertiveness and
activity,andhasbeenviewedasapredispositiontowardpositiveeffects. Takentogether,NandE
constitute independent, nonetheless interrelated, sources of subjective well-being [17]. The “high
N/low E” pattern is notably predictive of depressive symptoms and hopelessness [18] as well as
suicidalideation[19]. Agreeablenessdescribestraitsusuallydisplayedininterpersonalrelationships,
suchascourtesy,altruism,andcompliance. Opennesshasoftenbeenassociatedwithintellectand
depictsatendencytowardavarietyofexperiences,rangingfrominnerthoughtsorfantasytoactions
and ideas (hence frequently referred to as “openness to experience”). Finally, conscientiousness
concernsvariablessuchasself-discipline,deliberationanddutifulness,andhasbeenassociatedwith
educationalachievementandvolitionalaspects.
Despitetheabsenceoftheoreticalassumptions,theexistingliteratureallowsustomakeinferences
onthewaypersonalitytraitscouldaccountforvariabilityinIPTSconstructs. DerivedfromtheFFM,
theFive-FactorTheoryofpersonality[20]postulatesthattraits“influencepatternsofthoughts,feelings,
andactions”(p. 165). Itcouldthereforebehypothesizedthatanindividual’sexpressionofPB,TBand
ACmaypartlystemfromthespecificityofhis/herpersonalitytraits.
First,PBandTBhaveastrongcognitivecomponentastheydevelopfromnon-necessarilyobjective
perceptionsoflackofconnectednessorsocialisolation[2,21].HighlevelsofNhavebeenassociatedwith
distortedcognition[22]andhighlevelsofEwithperceivedsocialsupport,notablybelongingness[23].
Additionally,highlevelsofAhavebeenrelatedtotheabilityofmaintainingpositiveinterpersonal
relationshipsandtointerpersonaladjustment[24].Atentativehypothesiswouldthusbethataconjoint
presenceofhighN,lowEandlowAcreatesfavorabledispositionsforPBandTB.
The only five studies quantitatively addressing the relationships between the FFM and the
IPTS [4,9,11,13,14] support this assumption. Indeed, N was positively associated with PB, and E
negativelyassociatedwithbothPBandTB[9,13]. NwasalsopositivelyandEnegativelyassociated
with suicide proneness [11,14], a construct including “engagement in overtly suicidal behavior as
wellasinrisk-takingandpotentiallyinjury-producingbehaviors,coupledwithalackofhealthand
safetybehaviors,and/oralackofself-worthandself-enhancingbehaviors”([25],p. 416). Additionally,
Cramerandcolleagues[11,13,14]testedthemediatingeffectsofIPTSvariablesontherelationship
between the FFM and suicide proneness using structural equation modelling (SEM). Consistently
over three different populations of pre-incarcerated offenders, members of the lesbian, gay and
bisexualcommunity,anduniversitystudents,theyfoundthat,respectively,PBpositivelymediates
theinfluenceofNandTBnegativelymediatestheinfluenceofEonsuicideproneness. Moreover,the
twostudiesincludingO,AandCintheSEManalyses[11,14]showedthatTBnegativelymediates
theinfluenceofAonsuicideproneness,andthatneitherPBnorTBhaveamediatingeffectonthe
pathway between O and suicide proneness. Other effects were also identified but inconsistently,
whichledtheauthorstosuggestthattrait-interpersonalpathwaysarticulatedifferentlyinvarious
Int.J.Environ.Res.PublicHealth2018,15,636 3of14
populationgroups[11]. ThesamehypothesiswasraisedbyIrelandandYork[4],whosesamplewas
comprisedofwomenprisoners,toexplaintheunexpectedassociationbetweenincreasedextraversion
andself-injuriousbehavior.
Second, theIPTSproposesthatACisprogressivelyacquiredthroughpainandfearexposure
viamechanismsofhabituationandopponentprocesses[2]. Thequestionarisesastowhethercertain
patternsofpersonalitytraitscouldfavortheactivationofthesemechanisms. Habituationhasbeen
traditionallydefinedasa“decrementinmagnitudeofunlearnedresponses(... )duetorepetitive
stimulatory activation” ([26], p. 385). It has been experimentally assessed through a variety of
physiologicalmeasuressuchastheacousticstartleresponseandskinconductancerecording. Toour
knowledge,therelationshipbetweenhabituationandpersonalityhasneverbeenspecificallystudied
withinthecontextofAC;however,shouldacertainpatternofpersonalitytraitsfosterhabituationin
general,thiscouldalsoapplytofearandpainhabituationinparticular,thelattertwodimensionbeing
germanetoAC.Weonlyfoundonestudyaddressingtherelationshipbetweenpersonality,painand
habituation[27],andthisstudyshowedthatindividualswithlowNhabituatefastertopainthanthose
withhighNlevels. However,inanotherstudyusingtheFFM,individualswithapersonalityprofileof
highN/lowE-O-A-CdisplayedsmalleramplitudeoftheP300componentoftheauditoryevent–related
cortical potential, which suggests faster habituation [28]. Although less directly relevant to AC,
habituationandpersonalityingeneralhavenotyieldedconstantassociations,asillustratedbystudies
usinganacousticstartleresponseparadigm.Fasterhabituationwasassociatedwithhigh-Nindividuals
byBlanchandcollaborators[29]butwithhigh-EparticipantsbyLaRoweandcolleagues[30],while
Akdag et al. [31] did not find any relation with personality traits at all. Differences in theoretical
approachesandmeasurementmethodshaveprobablyfosteredsuchinconsistency. Incomparison
withNandE,theotherthreeFFMdimensionshavebeenfarlesssubjectofinvestigations.
Forthesereasons,researchonpersonalityandtheIPTSmainlyremainsexploratoryatthecurrent
stage. Thisstudywasthereforedesignedtoaddafurthercontributiontotheexistingliteraturein;
first,investigatingthepredictivecharacterofFFMpersonalitytraitsonthethreemainconstructsof
theIPTSandameasureofsuicidalideation,and;second,assessingtheeffectsofIPTSvariablesas
mediatorsoftherelationshipbetweenpersonalitytraitsandsuicidalideation. Basedontheexisting
data,weexpectedthatNwouldpositivelypredictPBandEnegativelypredictPBandTB.Wealso
hypothesizedthatPBwouldmediatethepositiveinfluenceofN,andTBthenegativeinfluenceof
bothEandA,onsuicidalideation. Thisboilsdowntoreplicatethehypothesizedtrait-interpersonal
pathwaysprogressivelyelaboratedbyCramerandcollaborators[11,13,14]. Additionally,wedecided
toinvestigatethepossiblemediatingfunctionofACintherelationshipbetweenFFMtraitsandsuicidal
ideation. Although the IPTS posits that AC is independent from suicide desire [2], a relationship
betweenthesetwoconstructswasfoundinabout57%ofreviewedstudies[3]. Becauseoftheabsence
of consistent findings in the habituation literature, we included the five FFM personality traits as
predictorsinouranalyses.
2. MaterialsandMethods
2.1. ParticipantsandProcedure
Thestudyprocedurewasmadeofself-andclinician-administeredquestionnairesincludinga
rangeofsociodemographicinquiries. Atotalof368individualsvisitingthepsychiatricroomofthe
emergency department at the Geneva University Hospitals, Switzerland, for a suicide attempt or
suicidalideationwereofferedtoparticipatetothestudy. Amongthem,129refusedforvariousreasons
suchasdifficultyconcentrating,lackofinterest,andproblemsunderstandingFrench. Additionally,
38wereexcludedbythepsychiatristinchargeofstudyinclusionduetoserioussuspicionregarding
dataqualityafternoticingobviouslackofmotivation—illustratedbyrepeatedpatternsofanswers
or systematic refusal to fill in questionnaires—or procedural problems (e.g., absence of signed
consent form). Thus, the final sample was comprised of 201 participants, aged 33.4 ± 14.5 years
old,predominatelycomposedofwomen(60.7%),Swisscitizens(57.7%),currentlynotinarelationship
Int.J.Environ.Res.PublicHealth2018,15,636 4of14
(65.7%),withoutchildren(60.2%),andeitherworkingorstudying(57.7%). Mostparticipantshada
history of suicide attempts (63.2%) and a current psychiatric diagnosis (88.6%) according to the
Mini-International Neuropsychiatric Interview [32] with the most prevalent diagnosis being an
episode of major depression (69.7%), followed by alcohol dependence (16.9%) and non-alcohol
substancedependence(10%).Additionalsociodemographicinformationonthesampleandrecruitment
procedurehasbeenpublishedelsewhere[33]. Thestudyprocedureandcontentswereapprovedbythe
localResearchEthicsCommitteeandallparticipantssignedaninformedconsentbeforestudyentry.
2.2. Measures
ThepersonalityprofilewasevaluatedwiththeFrenchversionoftherevisedNEO-Five-Factor
Inventory(NEO-FFI-R)[34,35],whose60itemswereselectedfromthelongerNEOPI-R[36]inorder
toprovideascoreforthefivepersonalitydimensionsoftheFFM(i.e.,N,E,O,AandC).Eachitem
is calculated with a 0–4 Likert score, leading to a maximum score of 48 per dimension. Internal
consistenciesweregoodforC(α=0.849,95%CI(0.816,0.878)),acceptableforN(α=0.758,95%CI
(0.705,0.805))andE(α=0.705,95%CI(0.641,0.762)),andquestionableforO(α=0.681,95%CI(0.612,
0.741))andA(α=0.680,95%CI(0.611,0.742)).
Thedesireforsuicide,asconceptualizedbyPBandTB,wasassessedwiththerevisedversion
of the Interpersonal Needs Questionnaire (INQ-R) [37], which proposes a subscale to evaluate PB
(6items,scorerange: 6–42)andasubscaletoevaluateTB(9items,scorerange: 9–63). Forbothscales,
higherscoresindicatestrongerfeelingsoftheconstructmeasured. Internalconsistencywasgoodfor
PB(α=0.866,95%CI(0.835,0.893))andacceptableforTB(α=0.749,95%CI(0.693,0.798)).
The acquired capability for suicide was assessed with the German Capability for Suicide
Questionnaire(GCSQ)[38],whichproposesthreesubscalesoffearlessnessofdeath(5items,score
range: 5–25),paintolerance(fiveitems,scorerange: 5–25),andperceivedcapabilityforsuicide(1item,
scorerange: 1–5). ThesummationofthesesubscoresprovidesatotalscoreforAC.Inallsubscales,
higherscoresindicategreatercapabilityforsuicide. TheinternalconsistencyoftheGCSQwasgood
(α=0.880,95%CI(0.854,0.903)).
WefinallyassessedsuicidalideationwiththeScaleforSuicideIdeation(SSI)[39],aself-administered
questionnaire comprised of 19 items rated on a 3-point Likert scale (score range: 0–38). The SSI has
beenvalidatedinFrench[40,41].Inoursample,internalconsistencywasgood(α=0.844,95%CI(0.811,
0.874)).IntheabsenceofversionsvalidatedinFrench,theGCSQandtheINQ-Rweretranslatedand
adapted according to the WHO guidelines [42] (http://www.who.int/substance-abuse/research_
tools/translation/en/).
2.3. StatisticalAnalyses
Afullsetofdescriptivestatisticalanalysesincludingmeans,standarddeviations,skewness,kurtosis,
andbivariatePearson’scorrelationswereinitiallycomputedforallvariablesofinterest.Wethenexplored
thepredictivevalueoftheFFMpersonalitytraits(N,E,A,O,andC)inthevarianceofeachIPTSvariable
(PB,TB,andAC)aswellasofameasureofcurrentsuicidalideation(SI)withmultiplelinearregression
models. Missingdataappearedinthreecasesoutof201,whichconstitutesaratioof0.015,andwere
handledwithmultipleimputation.InlinewiththerecommendationsofWhite,RoystonandWood[43],
who suggested that the imputation number should be equal or superior to the percentage of cases
containingmissingvalues,weconductedstatisticalanalysesaftergeneratingfivedatasets. Wedecidedto
relyonoriginaldata—whichincludedmissingdata—toassessparametersthatcouldnotbecomputed
throughcombinedestimationfrommultipleimputations(e.g.,standarddeviations,Fisher’stestin
multiple regression models). These analyses were performed with SPSS version 24.0 (IBM SPSS,
Chicago, IL, USA). Exploratory and control analyses (e.g., residuals, outliers, heteroscedasticity,
multicolinearity)wereconductedforeveryanalysis. Noassumptionviolationwasobserved.
Inasecondstep, weinvestigatedthemediatingfunctionofIPTSvariablesintherelationship
between FFM personality traits and current SI using a path analysis framework. To this end, we
Int.J.Environ.Res.PublicHealth2018,15,636 5of14
specifiedsixmodels. ThefirstonedrewontheliteratureandhypothesizedthatPBmediatesapositive
relationshipbetweenNandSIandanegativerelationshipbetweenEandSI,whereasTBmediates
negativerelationshipsbetweenEandSIandbetweenAandSI.Wethencomparedthismodelwith
anestedmodeladdingdirectpathwaysbetweenFFMvariables(i.e.,N,E,andA)andSI.Asimilar
procedurewasreplicatedforthenexttwomodels. BothexploredthemediatingfunctionofACinthe
relationshipbetweentheFFMandSI,thesecondoneincludingdirecteffects;yet,inlightofthesparse,
inconsistentfindingsaddressingpersonalityandAC(orratherAC’spossibleproxies)wedecidedto
includethefiveFFMvariablesaspredictorswithoutfurtherspecifyingresearchhypotheses.
Finally,wetestedtwoadditionalmodelsthatincludedPB/TB,respectivelyAC,asmediating
variables. ThesemodelsweredesignedinincludingFFM,IPTSandSIvariablesshowingsignificant
relationshipsinregressionanalyses. Weexpectedthesemodelstodemonstrateabetterfittothedata
thantheabove-mentionedmodelsbasedontheliterature;however,directstatisticalcomparisonswere
notpossibleasthelasttwomodelswerenotnestedwithintheinitialfour.
PathanalyseswereconductedusingAmosversion23.0(IBMSPSS,Chicago,IL,USA),which
estimatesmissingdatawithafullinformationmaximumlikelihoodestimator. Amosneverthelessdoes
notcomputeconfidenceintervalsforparametersestimatedfromdatabasescontainingmissingdata,as
wellasstandarderrorsandp-valuesforsomeanalyses(e.g.,indirecteffects). WeusedStatisticaversion
12.7(StatsSoftInc.,Tulsa,OK,USA)toestimatetheχ2 differencebetweennestedmodels(i.e.,model1
vs. model2,andmodel3vs. model4).
3. Results
3.1. DescriptiveStatistics
Means, standard deviations, skewness, kurtosis, and bivariate Pearson’s correlations for the
variablesofinterestofthisstudyaredisplayedinTable1. Takingsamplesizeintoaccount, Z-test
values for skewness (between |0.052| and |2.488|) and kurtosis (between |0.064| and |2.825|)
suggest that normality was preserved in the sample distribution of each variable [44]. The four
personalitytraitsofE,O,A,andCwereallpositivelyassociatedwithoneanother,exceptforOand
C whose correlation was not significant. On the other hand, N was negatively related to E and C,
and positively to O. The IPTS variables of PB and TB were positively related to each other and to
SI.ThevariablesassociatedwithACwereSI,PBandC,allpositively. OtherIPTS-FFMsignificant
relationshipswereNwithPBandTB(bothpositively),EwithPBandTB(bothnegatively),andAwith
TB(negatively). Finally,significantcorrelationsbetweenFFMtraitsandSIconcernedN(positively),
EandA(bothnegatively). Rangingfrom|0.143|to|0.453|,significantcorrelationsrepresentedsmall
tomediumeffectsizes(r)accordingtothecriteriaofCohen[45].
3.2. RegressionAnalyses
Wefirstbuiltthreemultiplelinearregressionequationsincludingthefivepersonalitydimensions
aspredictorsand,respectively,PB,TB,andACasthedependentvariable. Theinitialmodelaccounted
for a significant proportion of the variance in PB, F(5, 193) = 5.115, Adjusted R2 = 0.094, p < 0.001.
ThecriterionwaspredictedbyincreasedlevelsofN(b=0.422,t=4.186,p<0.001)anddecreasedlevels
ofO(b=−0.251,t=−2.248,p=0.025). However,noregressioneffectwasfoundforE(b=−0.160,
t = −1.391, p = 0.164), A (b = −0.024, t = −0.212, p = 0.832), and C (b = 0.048, t = 0.563, p = 0.573).
ThesecondmodelexplainedasignificantproportionofthevarianceinTB,F(5,193) =7.175,Adjusted
R2=0.135,p<0.001,withTBbeingpredictedbyN(b=0.338,t=3.265,p=0.001)andE(b=−0.364,
t = −3.092, p = 0.002). While the predicting value of O (b = −0.223, t = −1.932, p = 0.053) and
A(b= −0.213, t= −1.868, p=0.062)approachedthesignificancethreshold, C(b=0.084, t=0.954,
p=0.340)wasnotrelatedtothecriterion. Finally,thesetofFFMvariablespredictedvarianceinAC,
F(5,193)=3.203,AdjustedR2=0.053,p=0.008,withspecificeffectsfoundinE(b=−0.185,t=−2.372,
p = 0.018) and C (b = 0.191, t = 3.327, p = 0.001) but not in N (b = −0.070, t = −1.032, p = 0.302),
O(b=0.026,t=0.347,p=0.729),andA(b=−0.100,t=−1.332,p=0.183).
Int.J.Environ.Res.PublicHealth2018,15,636 6of14
Table1.Means,standarddeviationsandbivariatePearson’scorrelationmatrixforvariablesofinterest(n=201withfiveimputations).
Variables Mean SD† Skewness† Kurtosis† 1 2 3 4 5 6 7 8 9
Neuroticism 33.703 7.470 −0.428 −0.357 1
Extraversion 25.185 6.434 −0.340 0.047 −0.176* 1
Opennesstoexperience 28.678 6.699 0.097 −0.235 0.269** 0.187** 1
Agreeableness 30.735 6.656 −0.413 −0.022 −0.086 0.191** 0.175* 1
Conscientiousness 26.967 8.990 −0.185 −0.306 −0.217** 0.296** 0.118 0.342** 1
Perceivedburdensomeness 23.375 10.120 0.009 −0.966 0.283** −0.179* −0.083 −0.081 −0.088 1
Thwartedbelongingness 40.020 10.661 −0.164 −0.304 0.233** −0.292** −0.132 −0.198** −0.109 0.380** 1
Acquiredcapabilityforsuicide 29.898 6.696 0.149 −0.100 −0.089 −0.103 −0.019 −0.034 0.189** 0.143* 0.100 1
Suicidalideation 14.707 8.134 −0.268 −0.779 0.322** −0.358** −0.038 −0.167* −0.099 0.453** 0.369** 0.251** 1
Note: *p-valuessignificantatthe0.05threshold;**p-valuessignificantatthe0.01threshold;†computedwiththeoriginaldata,i.e.,n=200(exceptforsuicidalideation,n=201).
Thestandarderrorwas0.172forskewnessand0.342forkurtosis.
Int.J.Environ.Res.PublicHealth2018,15,636 7of14
WethenconstructedafourthmodelwithSImeasuredbytheSSIasthedependentvariableandthe
similarfivepersonalityvariablesaspredictors. Additionally,wecontrolledfortwosocio-demographic
variablessusceptibletohaveaconfoundinginfluence,namelygenderandage. Suicidalideationhas
indeedbeenidentifiedasmorefrequentinfemalethanmalegender[46,47], andtendstobemore
prevalentinolderthanyoungeragegroups[48,49]. Wetransformedgenderintoadummyvariable,
coding0forfemalesand1formales. ThemodelpredictedSI,F(7,192)=7.877,AdjustedR2=0.195,
p<0.001, as well as higher N (b = 0.304, t = 3.743, p<0.001) and lower E (b=−0.416, t=−4.598,
p<0.001). However, no effect was found for O (b=−0.062, t = −734, p = 0.463), A (b=−0.141,
t=−1.662, p = 0.097) and C (b = 0.100, t = 1.535, p = 0.125). Similarly, neither age (b = −0.017,
t=−0.423,p=0.672)norgender(b=−1.232,t=−1.096,p=0.273)hadasignificantinfluenceonSI.
3.3. PathAnalyses
AsdetailedinTable2,pathanalyseswithPBandTBasmediatorswerepoorlyadjustedtothe
data. Thiswasthecaseformodelswithout(χ2=65.241,df=9,p<0.001,CFI=0.609,RMSEA=0.177
90%CI(0.138–0.219))andwith(χ2 =42.551, df=6, p<0.001, CFI=0.747, RMSEA=0.17590%CI
(0.127–0.226))directestimatesofFFMtraitsonSI.Althoughmodel2yieldedthreedegreesoffreedom
tomodel1,theformerwassignificantlyimproved(∆χ2=22.69,∆df=3,p<0.001). Alldirecteffects
weresignificantexceptAtoSIinmodel2. Despitep-valueswereunavailableforPBandTBmediation
effectsofFFMtraitsonSI,standardizedeffectsrangedfrom|0.021|to|0.115|suggestingthatthese
effectsareweak.
Nextwetestedamodel(model5)basedonsignificantpredictiverelationshipshighlightedin
regressionanalyses. Thus,weincludedN,E,andOinthemodelandtestedthefollowingdirecteffects:
NtoPB,NtoTB,NtoSI;EtoTB,EtoSI;OtoPB.AsdisplayedinTable4, alldirecteffectswere
significantbutthemodeldidnotadjustwelltothedata(χ2 =60.073,df=7,p<0.001,CFI=0.681,
RMSEA=0.19590%CI(0.151–0.242)).
Table 2. Estimated unstandardized (with standard errors) and standardized parameters for the
regression weights, as well as fit indices, for the two models with PB and TB as mediators in the
relationshipbetweenFFMvariablesandSI.
Model1,fitindices:χ2=65.241,df=9,p<0.001,CFI=0.609,RMSEA=0.17790%CI(0.138–0.219)
Parameterestimates Unstandardizedβ Standardizedβ p-value
NtoPB 0.345(0.091) 0.256 <0.001
EtoPB −0.221(0.106) −0.141 0.038
EtoTB −0.437(0.111) −0.266 <0.001
AtoTB −0.237(0.108) −0.149 0.027
PBtoSI 0.294(0.050) 0.375 <0.001
TBtoSI 0.175(0.047) 0.235 <0.001
NtoSI(indirect) 0.101 0.096 N/A
EtoSI(indirect) −0.141 −0.115 N/A
AtoSI(indirect) −0.041 −0.035 N/A
Model2,fitindices:χ2=42.551,df=6,p<0.001,CFI=0.747,RMSEA=0.17590%CI(0.127–0.226)
Parameterestimates Unstandardizedβ Standardizedβ p-value
NtoPB 0.342(0.092) 0.254 <0.001
EtoPB −0.215(0.106) −0.137 0.043
EtoTB −0.437(0.111) −0.266 <0.001
AtoTB −0.238(0.108) −0.150 0.027
PBtoSI 0.250(0.049) 0.321 <0.001
TBtoSI 0.104(0.047) 0.140 0.027
NtoSI 0.181(0.065) 0.172 0.006
EtoSI −0.278(0.077) −0.229 <0.001
AtoSI −0.068(0.072) −0.058 0.342
NtoSI(indirect) 0.085 0.081 N/A
EtoSI(indirect) −0.099 −0.081 N/A
AtoSI(indirect) −0.025 −0.021 N/A
Note:Standarderrorsandp-valuesnotavailableforindirectpathwayestimatesduetomissingvalues.Inbold,
effectssignificantatthe0.05threshold.
Int.J.Environ.Res.PublicHealth2018,15,636 8of14
Inthesamevein,pathanalysisframeworksusingACasmediatoroftheeffectofFFMtraitsonSI
didnotdemonstrateadequatefittothedatainbothmodels(model3: χ2=142.311,df=15,p<0.001,
CFI=0.111,RMSEA=0.20690%CI(0.176–0.238);model4: χ2=92.562,df=10,p<0.001,CFI=0.423,
RMSEA = 0.203 90% CI (0.166–0.242)) as described in Table 3. Despite the loss of five degrees of
freedom,model4hadasignificantlybetteradjustmentthanmodel3asassessedbytheChi-squared
analysis(∆χ2 =49.749,∆df=5,p<0.001). Inbothmodels,E(negatively)andC(positively)hadan
influenceonAC,andACpositivelypredictedSI.Additionally,directeffectsofN(positively)andE
(negatively)onSIwerefoundinmodel4. StandardizedindirecteffectsofFFMtraitsonSIshowedlow
valuesrangingfrom|0.006|to|0.064|.
Next,wetestedamodel(model6)withACplayingamediatingrolebetweenFFMvariablesand
SIbasedontheregressionanalyses. WeincludedN,EandC,andestimatedthefollowingrelations:
NtoSI;EtoAC,EtoSI;CtoAC.AsdisplayedinTable4,alldirecteffectsweresignificantbutthe
modeldidnotadjustwelltothedata(χ2=31.452,df=5,p<0.001,CFI=0.705,RMSEA=0.16390%CI
(0.111–0.219)).
Table 3. Estimated unstandardized (with standard errors) and standardized parameters for the
regressionweights,aswellasfitindices,forthetwomodelswithACasmediatorintherelationship
betweenFFMvariablesandSI.
Model3,fitindices:χ2=142.311,df=15,p<0.001,CFI=0.111,RMSEA=0.20690%CI(0.176–0.238)
Parameterestimates Unstandardizedβ Standardizedβ p-value
NtoAC −0.069(0.061) −0.076 0.257
EtoAC −0.185(0.071) −0.174 0.009
OtoAC 0.024(0.068) 0.024 0.725
AtoAC −0.098(0.069) −0.095 0.155
CtoAC 0.191(0.051) 0.251 <0.001
ACtoSI 0.306(0.082) 0.257 <0.001
NtoSI(indirect) −0.021 −0.019 N/A
EtoSI(indirect) −0.057 −0.045 N/A
OtoSI(indirect) 0.007 0.006 N/A
AtoSI(indirect) −0.030 −0.024 N/A
CtoSI(indirect) 0.058 0.064 N/A
Model4,fitindices:χ2=92.562,df=10,p<0.001,CFI=0.423,RMSEA=0.20390%CI(0.166–0.242)
Parameterestimates Unstandardizedβ Standardizedβ p-value
NtoAC −0.068(0.061) −0.075 0.266
EtoAC −0.185(0.071) −0.175 0.009
OtoAC 0.023(0.068) 0.023 0.731
AtoAC −0.098(0.069) −0.096 0.153
CtoAC 0.191(0.051) 0.252 <0.001
ACtoSI 0.293(0.077) 0.247 <0.001
NtoSI 0.349(0.066) 0.322 <0.001
EtoSI −0.327(0.078) −0.260 <0.001
OtoSI −0.071(0.074) −0.059 0.336
AtoSI −0.100(0.074) −0.082 0.179
CtoSI 0.034(0.057) 0.038 0.548
NtoSI(indirect) −0.020 −0.018 N/A
EtoSI(indirect) −0.054 −0.043 N/A
OtoSI(indirect) 0.007 0.006 N/A
AtoSI(indirect) −0.029 −0.024 N/A
CtoSI(indirect) 0.056 0.062 N/A
Note:Standarderrorsandp-valuesnotavailableforindirectpathwayestimatesduetomissingvalues.Inbold,
effectssignificantatthe0.05threshold.
Int.J.Environ.Res.PublicHealth2018,15,636 9of14
Table 4. Estimated unstandardized (with standard errors) and standardized parameters for the
regressionweights,aswellasfitindices,foramodelwithPBandTB(model5)andamodelwithAC
(model6)asmediatorsintherelationshipbetweenFFMvariablesandSI.
Model5,fitindices:χ2=60.073,df=7,p<0.001,CFI=0.681,RMSEA=0.19590%CI(0.151–0.242)
Parameterestimates Unstandardizedβ Standardizedβ p-value
NtoPB 0.442(0.091) 0.321 <0.001
OtoPB −0.272(0.101) −0.177 0.007
EtoTB −0.428(0.110) −0.261 <0.001
NtoTB 0.268(0.095) 0.189 0.005
PBtoSI 0.250(0.048) 0.327 <0.001
TBtoSI 0.110(0.047) 0.148 0.020
NtoSI 0.181(0.068) 0.172 0.008
EtoSI −0.291(0.068) −0.238 <0.001
NtoSI(indirect) 0.140 0.133 N/A
EtoSI(indirect) −0.047 −0.039 N/A
OtoSI(indirect) −0.068 −0.058 N/A
Model6,fitindices:χ2=31.452,df=5,p<0.001,CFI=0.705,RMSEA=0.16390%CI(0.111–0.219)
Parameterestimates Unstandardizedβ Standardizedβ p-value
EtoAC −0.181(0.071) −0.172 0.011
CtoAC 0.180(0.051) 0.238 <0.001
ACtoSI 0.302(0.074) 0.253 <0.001
NtoSI 0.328(0.066) −0.150 <0.001
EtoSI −0.350(0.078) −0.278 <0.001
EtoSI(indirect) −0.055 −0.043 N/A
CtoSI(indirect) 0.054 0.060 N/A
Note:Standarderrorsandp-valuesnotavailableforindirectpathwayestimatesduetomissingvalues.Inbold,
effectssignificantatthe0.05threshold.
4. Discussion
Thisstudyisthefirsttoinvestigatetherelationshipbetweenpersonalitydimensionsasdefined
by the FFM model and the main constructs of the IPTS in a large clinical population with current
SI.Ourresultsfromcorrelationandregressionanalysesaregloballyinlinewiththeliteratureand
suggest that certain dispositions of personality traits may foster the occurrence of PB, TB and AC.
Whensignificantfindingsoccurredtheireffectsizeremainedgloballymodest,asillustratedbythe
percentagesofvarianceinPBandTB(9.4%and13.5%,respectively)explainedbyFFMtraitsthrough
regressionequations. However,wedidnotreplicatethefindingsofCrameretal.[11,13,14],notably
withregardtothetrait-interpersonalhypothesis. Overall,thissuggeststhatpersonalitydoesplaya
roleinthevarianceofSIorvariablesapproachingSI(thatis,inthiscase,PBandTBfromtheIPTS)but
thattheimportanceofthisroleshouldbeconsideredwithcautionandnotyieldtoover-interpretation.
Interestingly,personalitywasfoundtobepredictiveofACaswell. Thisis,toourknowledge,thefirst
timethatanassociationbetweenpersonalityandACisstatisticallyassessed.
Inlinewiththeliteratureandinbothcorrelationandregressionanalyses,wefoundarelationship
betweenNandPB,andbetweenN,EandTB.Additionally,OwasanegativepredictorofPB,and
EandAnegativelycorrelatedwithPB,respectivelyTB.AnassociationbetweenN,E,andvariables
relatedtosuicidedesirehasalreadybeenidentified[9,50]. Additionally,aconjointpresenceofhigher
levelsofNandlowerlevelsofE,AandCcharacterizedindividualswithpsychiatricdisorders[51]
and more specifically SI [19]. Whereas C was associated with neither PB nor TB in our study, the
correlationbetweenAandTBwasnotsurprisingasagreeablenessisstronglyrelatedtointerpersonal
behavior. Besides, low levels of A might complicate the formation of interpersonal relationships,
indirectly fostering feelings of not belonging to a social group [36]. Similarly, the finding that PB
wasnegativelypredictedbyOshouldnotbeasurprisewhenoneconsidersthat“opennessaffects
socialperceptionsandtheformationofsocialattitudes,thechoiceoffriendsandspouses,political
Int.J.Environ.Res.PublicHealth2018,15,636 10of14
activity,andculturalinnovation”([52],p. 257). Thus,ourresultssuggestthatthewayoneconsiders
interpersonalrelationshipsispivotalintheformationofSI,providinginthiswaysupporttotheIPTS.
HighlevelsofCandlowlevelsofEwerepredictiveofagreatercapabilitytoactonone’ssuicidal
desire,whichimpliesthatpersonalitycandirectlyorindirectlycontributetothewayone’smayacquire
thecapabilitytocommitsuicideassuggestedbyothers[7,8]. Thesestudiesyetaddresseddimensions
notconsideredbytheFFM(e.g.,psychotism,stoicism,sensationseeking),whichlimitscomparisons
withours. ThefactthatEplayedaroleintheacquisitionofaconstructsuchasAC,whichiscomprised
of a habituation component, is in line with other studies that found a relationship between E and
habituation.However,incontrastwithourresults,thesestudiesoftenconcludedthathabituationtakes
placefasterinextrovertsthanintroverts[30,53,54]. Fromanotherstandpoint,ithasbeenunderscored
thatmedicallyserioussuicideattempters—thatis,individualssupposedtohaveacquiredthecapability
toseriouslyharmthemselves—thoughtmoreaboutsuicidepreparationandweremoreprecautious
against the discovery of their intentions than individuals whose suicide attempts were medically
lessserious[55,56]. Thisimpliesthatthosewithmedicallyserioussuicideattemptshowedfeatures
correspondingtohighscoresontheCdimensionsuchasachievementstriving,self-disciplineand
deliberation. AsparticipantsinourstudywereincludedwithamixedcriterionofpresenceofSIand
presenceofasuicideattempt,aswellasdifferenthistoriesofsuicideattempts,wewerenotableto
investigatethisaspectmorespecifically. Thus,furtherresearchshouldbeconducted,ontheonehand,
to confirm the role of E in the development of habituation and other components from which AC
theoreticallyderives(e.g.,opponentprocesses),and,ontheotherhand,tospecifytheinterplayof
personalityandACintakingdifferentlevelsoflethalityintoconsideration.
We could not replicate the findings of Cramer and colleagues [11,13,14] in that none of the
trait-interpersonal model tested, either with PB/TB or with AC, fit the data in an adequate way.
This was the case for theory-driven models (i.e., models 1 to 4) and models based on regression
findings(i.e.,models5and6). ThisissomewhatsurprisingasdirecteffectsbetweenFFM,IPTSandSI
variableswereconsistentlysignificant;yetthelimitedsamplesizeandtherelativesmallproportionof
varianceinIPTScomponentsandSIexplainedbyregressionmodelsmightimplyalackofstatistical
power that could account for the poor fit of global models. Similarly, these results might be due
tothetypeofpopulationstudied(i.e.,patientsvisitingemergencywithsuicidalideation)asithas
been proposed that relationships between personality and suicidality could depend on this [4,11].
Anotherpossiblereason,ourmeasureofoutcomewasSIwhilethatofCramerandcolleagueswas
suicideproneness. Yet,suicidepronenesshasbeenassociatedwithSI[25],makingthisassumption
unlikely. Finally,contrarytoCramerandcollaborators[11,13],wedidnotallowmeasurementerrors
ofsomeofourvariablestocorrelate. Whilesuchaproceduremayleadtosignificantimprovementof
apathanalysismodel,itappearstobetheoreticallyunjustifiedinalmostallcases[57]. Cramerand
colleaguesdidnotprovideadetailedrationaleforallowingmeasurementerrorstocorrelateand,as
faraswewereconcerned,wedidnothaveany. Itappearsthenpivotalthattrait-interpersonalmodels
continuetobeinvestigatedundertheory-andevidence-basedhypotheses,withvariouspopulation
typesandimportantsamplesizes. Thisconstitutesaprerequisitebeforeanydefinitiveconclusionson
theviabilityofthisapproachcanbedrawn.
Thisstudyhadseverallimitations,whichshouldbetakenintoaccount. First,ourexperimental
designdidnotallowustocollectlongitudinaldata,whichlimitstheinterpretationofthefindingsin
apredictiveway. Second,wedefinedourresearchhypothesesonthebasisoftheexistingliterature.
However,studiesexploringtherelationshipsbetweensuicidebehaviors,theFFMandtheIPTSdidnot
considerthesamemeasurementoutcomes. Thus,whileweusedameasurementofSI,othersutilized
historyofSI[9],suicideproneness[11,13,14],andself-injuriousbehavior[4]. Thesemajordifferences
mayhavecontributedtosomeoftheinconsistentresultsobserved. Third,lowinternalconsistency
valuesintwosubscalesoftheNEOFFI-R,namelyOandA,mightimplythatitemswerenotsufficient
innumbertosatisfactorilyaccountforthesedimensions. Thismayhaveimpactedourresults. Inline
withthat,theuseofthe60-itemNEOFFI-Rratherthanthelonger,240-itemNEOPI-Rdidnotofferthe