UBC Faculty Research and Publications

Longitudinal Associations Between Types of Childhood Trauma and Suicidal Behavior Among Substance Users… Marshall, Brandon David Lewis; Galea, Sandro; Wood, Evan; Kerr, Thomas Sep 30, 2013

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-Marshall_B_et_al_Longitudinal_associations_between.pdf [ 491.08kB ]
JSON: 52383-1.0340662.json
JSON-LD: 52383-1.0340662-ld.json
RDF/XML (Pretty): 52383-1.0340662-rdf.xml
RDF/JSON: 52383-1.0340662-rdf.json
Turtle: 52383-1.0340662-turtle.txt
N-Triples: 52383-1.0340662-rdf-ntriples.txt
Original Record: 52383-1.0340662-source.json
Full Text

Full Text

Longitudinal Associations Between Types of ChildhoodTrauma and Suicidal Behavior Among Substance Users:A Cohort StudyBrandon D. L. Marshall, PhD, Sandro Galea, MD, DrPH, Evan Wood, MD, PhD, and Thomas Kerr, PhDThe global burden of suicide is considerableand is the tenth leading cause of death world-wide, with annual mortality estimated at 14.5deaths per 100 000.1 Suicide is a major andpreventable public health problem amongyoung people aged 15 to 24 years in Canada(second leading cause of death2) and theUnited States (third leading cause of death3).Each year, approximately1million adults in theUnited States attempt suicide, resulting in35 000 deaths and more than 320 000emergency department visits.4 The societal,financial, and public health burdens associatedwith suicide are therefore substantial.The epidemiology of suicide is multifactorialand complex.1 The 2012 National Strategy forSuicide Prevention identifies several groups atparticularly high risk of suicide in the UnitedStates, including individuals with a past historyof suicidal behavior, members of the armedforces and veterans, American Indians/AlaskaNatives, men in midlife, and individuals injustice and child welfare settings.5 Of publichealth concern in Canada, suicide rates amongAboriginal Peoples are 2 to 3 times thatobserved in the nonaboriginal population.6,7 Alarge body of literature has also demonstratedhigh rates of suicidal behavior among lesbian,gay, bisexual, and transgender populations.8---10People who use illicit drugs are particularlyvulnerable to suicidal ideation and behavior,and suicide is a leading cause of death indrug-using populations.11,12 Furthermore, therelationship between substance abuse and in-creased suicide risk has been well estab-lished.13,14 A growing body of research hasexamined various correlates of suicide attemptsamong drug users. In treatment-seeking sam-ples of drug and alcohol abusers, major de-pressive disorder and other psychiatric condi-tions (e.g., borderline personality disorder,anxiety, agoraphobia) have been associatedwith a history of suicide attempts.15---20Furthermore, markers of social disadvantageand marginalization, such as unemploymentand homelessness, are associated with a height-ened risk of suicide and are common amongdrug users.11Specific typologies of drug use havealso been linked to a greater likelihood ofattempting suicide, including longer durations ofsubstance use,18,21 polysubstance use,20,22 andinjection methamphetamine use.23In recent years, childhood maltreatmenthas emerged as a consistent correlate of sui-cidal ideation and behavior among drugusers.19,24,25 These studies provide preliminaryevidence that childhood abuse and neglect areimportant determinants of suicide in drug-using populations. However, it is unclearwhether certain types of childhood maltreat-ment are more strongly associated with suicidalbehavior than others. The majority of studies todate have examined suicide and early trau-matic experiences among clinical samples,which may be subject to selection bias if beingin treatment is a common effect of bothexposure (i.e., childhood maltreatment) andother, unmeasured factors that may causesuicidal behavior (e.g., genetic or familial sus-ceptibility to psychiatric disorders such as de-pression). To our knowledge no studies haveprospectively examined suicidal behavior ina community-recruited cohort of drug users.Drug-using cohorts are particularly well suitedto examining the relationship between child-hood maltreatment and recurrent suicidal be-havior because of their high rates of suicide.We used recurrent event survival models todetermine the longitudinal associations be-tween exposure to different types and sever-ities of childhood trauma and suicide attempts,measured prospectively in a cohort of drugusers in Vancouver, British Columbia.METHODSThe objectives, design, and methods of theVancouver Injection Drug Users Study (enroll-ing HIV-negative persons) and the AIDS CareObjectives. We examined the longitudinal associations between differenttypes and severities of childhood trauma and suicide attempts among illicitdrug users.Methods. Data came from 2 prospective cohort studies of illicit drug users inVancouver, Canada, in 2005 to 2010. We used recurrent event proportionalmeans models to estimate adjusted and weighted associations between typesand severities of childhood maltreatment and suicide attempts.Results. Of 1634 participants, 411 (25.2%) reported a history of suicidalbehavior at baseline. Over 5 years, 80 (4.9%) participants reported 97 suicideattempts, a rate of 2.6 per 100 person-years. Severe to extreme levels of sexualabuse (adjusted hazard ratio [AHR] = 2.5; 95% confidence interval [CI] = 1.4, 4.4),physical abuse (AHR = 2.0; 95% CI = 1.1, 3.8), and emotional abuse (AHR = 3.5;95% CI = 1.4, 8.7) predicted suicide attempts. Severe forms of physical andemotional neglect were not significantly associated with an increased risk ofsuicidal behavior.Conclusions. Severe sexual, physical, and emotional childhood abuse confersubstantial risk of repeated suicidal behavior in adulthood. Illicit drug usersrequire intensive secondary suicide prevention efforts, particularly among thosewith a history of childhood trauma. (Am J Public Health. 2013;103:e69–e75. doi:10.2105/AJPH.2013.301257)RESEARCH AND PRACTICESeptember 2013, Vol 103, No. 9 | American Journal of Public Health Marshall et al. | Peer Reviewed | Research and Practice | e69Cohort to Evaluate Exposure to Survival Ser-vices (enrolling HIV-positive drug users) aredescribed in detail elsewhere.26,27 Briefly,since May 1996, 2707 drug users have beenrecruited into the studies. The instruments andall other follow-up procedures for each studyare identical to allow for combined analyses.Recruitment proceeds by word of mouth,street-based outreach, and referrals from localsocial services. Participants are eligible to par-ticipate if they are aged at least 14 years, residein the greater Vancouver region, and reportinjection drug use within the past 6 months.Commencing in 2005, an additional waveof recruitment was conducted, with similarcriteria. However, eligibility for the AIDSCare Cohort to Evaluate Exposure to SurvivalServices study was expanded to include per-sons reporting heavy noninjection drug usein the past 6 months (e.g., crack cocaine,methamphetamine).Measures ascertaining exposure to child-hood trauma were added in late 2005; there-fore, we restricted our analyses to the 1804individuals who were enrolled and had com-pleted at least 1 follow-up visit betweenDecember 1, 2005, and December 1, 2010.We further excluded 12 (0.7%) individualswho were missing sociodemographic data and158 (8.8%) who refused to provide informa-tion regarding childhood traumatic experi-ences. The final analytic sample was 1634.Every 6 months, participants attend studyvisits and complete an interviewer-administeredquestionnaire. Nurses also assess participantsfor various health conditions and obtainblood specimens for HIV and hepatitis Cserology. Participants receive Can$20 foreach visit.MeasuresThe studies collected information on a widearray of participant characteristics, such associodemographic information, drug use pat-terns, health service utilization, HIV risk be-haviors, and health conditions, including sui-cidal ideation and behavior. Our primaryoutcome was response to the question, “In thelast 6 months, have you attempted suicide?”Participants who responded affirmatively werecounseled by on-site nurses and referred toadditional services, if appropriate. At their firstvisit during the study period, participants werealso asked if they had ever attempted suicide intheir lifetime.The primary exposure of interest was child-hood trauma, as measured by responses to theChildhood Trauma Questionnaire (CTQ).28All participants completed this instrument attheir first study visit. The CTQ is a 25-itemvalidated instrument used to assess, retrospec-tively, 3 forms of abuse (sexual, physical, andemotional) and 2 forms of neglect (physical andemotional) occurring in childhood. The CTQprovides a separate score for 5 subscales, eachwith 5 items, that correspond to each type ofabuse or neglect. On a 5-point Likert scale from1 (never true) to 5 (very often true), partici-pants respond to statements such as “When Iwas growing up I had to wear dirty clothes”(physical neglect) and “Someone tried to makeme do sexual things” (sexual abuse). Eachsubscale produces scores ranging from 5 to 25.We used recommended and a priori cutoffscores to translate each subscale score into 1 of4 levels of childhood trauma: none or minimal(5---8), low to moderate (9---12), moderate tosevere (13---15), and severe to extreme (> 15).28The reliability and validity of the CTQ has beendemonstrated,28---30 and the instrument hasbeen used successfully in several studies of illicitdrug users.31,32To examine the extent to which childhoodmaltreatment predicts adult suicidal behavioroutside of pathways involving established sui-cide risk factors, we included in all modelsindividual characteristics known to increase therisk of attempting suicide. Variables consideredas confounders were year of birth (age inyears); gender (female vs male); sexual orien-tation (lesbian, gay, bisexual, or transgender vsheterosexual); aboriginal ancestry (Inuit, Métis,or First Nations vs other); HIV status, deter-mined by serological testing (positive vs nega-tive); and lifetime history of suicidal behavior atbaseline (any attempts vs none). We also in-cluded the following as time-updated covari-ates in all hazards models (all, yes vs no): activeinjection drug use in the past 6 months,homelessness in the past 6 months, physical orsexual victimization in the past 6 months,recent nonfatal overdose experience, and cur-rent enrollment in a drug or alcohol treatmentprogram. Finally, in light of the establishedrelationship between depressive symptomol-ogy and suicidal behavior, we included asa time-updated covariate self-reported depres-sion, measured at each follow-up with thepreviously validated Center for EpidemiologicStudies Depression Scale and defined by anestablished cutoff (‡ 22) indicating high levelsof depressive symptoms.33Statistical AnalysesAs a first step, we computed mean scoresand Cronbach a for each CTQ subscale. Weused the Pearson v2 test and the Wilcoxon testto compare sociodemographic characteristicsand CTQ responses among those who reportedattempting suicide during follow-up with thosewho did not.Second, we used Kaplan---Meier methods34to determine the cumulative incidence ofattempting suicide during follow-up amongstudy participants, stratified by CTQ severitycategory for each type of trauma. The studyquestionnaires assessed suicidal behavior oc-curring in the past 6 months, so we estimatedthe date of suicide attempt as occurring 3months prior to the interview date. In time tofirst event analyses, we right-censored all par-ticipants who reported suicidal behavior as ofthe date of their first suicide attempt reportedduring follow-up. We right-censored personswho never reported a suicide attempt as oftheir last visit. We used the person---timemethod to calculate the cumulative incidenceof suicidal behavior over the study period. Weused the log-rank test to compare the survivaldistributions of the 4 CTQ severity categories.We used Cox proportional hazards regressionto determine the crude and adjusted associa-tions between each type of childhood traumaand time to first report of attempting suicideduring follow-up.Third, because some participants reportedmore than 1 suicide attempt during the studyperiod, we constructed 5 recurrent event sur-vival models to examine the relationship be-tween the outcome of interest and exposure toeach type of childhood trauma. These modelsincorporated information on all suicide at-tempts recorded over the entire study periodand improved precision of the estimates ofinterest, because of more suicide events in eachanalysis. We used a proportional rates---meansmodel described by Lin et al.35 to accountfor correlation among the length of individuals’repeated time at risk for a suicide attempt. InRESEARCH AND PRACTICEe70 | Research and Practice | Peer Reviewed | Marshall et al. American Journal of Public Health | September 2013, Vol 103, No. 9these models, we specified a counting processframework to define time to repeated events,such that individuals were considered to be atrisk from time zero to the first event, from thefirst event to the second event, and so forth.As in the case of Cox proportional hazardsregression, the model assumed proportionalmeans.36 We assessed this assumption for eachvariable of interest by visual inspection of theSchoenfeld residuals plots37 and by examiningtime-by-covariate interactions.38We first computed hazard ratios represent-ing the bivariable associations between severityof childhood traumatic experiences and re-peated suicide attempts. We then constructedmultivariable models that incorporated allvariables hypothesized a priori as confounders.To account for possible informative censoring,we constructed a final set of models withstabilized inverse probability of censoringweights (IPCW).39 In IPCW analyses, the sam-ple is reweighted such that the contributionof participants who drop out is effectivelyinflated. If model specification is correct, IPCWthus permits the estimation of effect estimatesthat would have been observed if all partici-pants had stayed in the study. We modeled theprobability of remaining uncensored, condi-tional on the study exposure and other cova-riates associated with the outcome, with pooledlogistic regression, as described elsewhere.40,41We conducted all statistical analyses with SASversion 9.3 (SAS Institute Inc, Cary, NC). Wereported 2-sided P values.RESULTSThe 1634 participants eligible for analysisattended 6217 study visits (median = 4; inter-quartile range [IQR] = 1---6). The median age atfirst study visit was 42 years (IQR = 36---48;range = 19---71); 1109 (67.9%) were male,534 (32.7%) were of aboriginal ancestry, and203 (12.4%) self-identified as lesbian, gay,bisexual, or transgender. At their first studyvisit, 411 (25.2%) participants reporteda lifetime history of suicidal behavior. Overthe study period, 195 (11.9%) were lost tofollow-up; dropouts were younger thanretained respondents (P< .001) and morelikely to have reported a history of suicidalbehavior at their first study visit (35.3% vs24.9%; P= .002). Persons lost to follow-upalso reported higher scores on each CTQ sub-scale (P< .05). We accounted for these differ-ences in the IPCW analysis.Over the 5-year study period, 65 partici-pants reported 1 suicide attempt, 13 reported 2attempts, and 2 reported 3 attempts, for a totalof 97 events among 80 participants. Theresulting incidence density was 2.6 per 100person-years (95% confidence interval [CI] =2.1, 3.2). As shown in Table 1, persons whoreported attempting suicide during follow-upwere more likely than those who did not tobe female (51.3% vs 31.2%; P< .001); beof Aboriginal ancestry (45.0% vs 32.0%;P= .016); score higher on the depressionscale, indicating a greater prevalence ofdepressive symptomology at baseline (medianscore = 35 vs 24; P < .001); and report ahistory of attempting suicide at baseline(68.7% vs 22.9%; P < .001).We observed a high prevalence of childhoodtraumatic experiences among study partici-pants. The mean scores on the CTQ subscaleswere sexual abuse, 10 (SD = 7); physical abuse,11 (SD = 6); emotional abuse, 13 (SD = 6);physical neglect, 10 (SD = 5); and emotionalneglect, 13 (SD = 6). As shown in Table 2, theproportion of participants reporting severe toextreme levels of maltreatment ranged from15.4% (physical neglect) to 34.2% (emotionalneglect). Notably, one quarter of participantsreported severe to extreme forms of sexual andphysical abuse. For each subscale, Cronbach awas greater than 0.75, indicating very goodinternal consistency. As shown in Table 2, forevery CTQ subscale, we observed a significantassociation between the severity of childhoodtrauma and suicidal behavior during follow-up(all, P< .01).We observed statistically significant rela-tionships between trauma severity and in-creased cumulative incidence of suicidal be-havior for all types of abuse and neglect (alllog-rank P values < .01; Figures A---E, availableas a supplement to the online version of thisarticle at http://www.ajph.org). The cumulativeTABLE 1—Sociodemographic Characteristics of Participants Who Did and Did Not AttemptSuicide: Vancouver Injection Drug Users Study and AIDS Care Cohort to Evaluate Exposureto Survival Services, Vancouver, British Columbia, 2005–2010CharacteristicAttempted Suicide(n = 80), Median(IQR) or No. (%)Did Not Attempt Suicide(n = 1554), Median(IQR) or No. (%) PAge,a y 41 (33–47) 43 (36–48) .111Gender <.001Female 41 (51.3) 484 (31.2)Male 39 (48.7) 1070 (68.8)Aboriginal ancestryb .016Yes 36 (45.0) 498 (32.0)No 44 (55.0) 1056 (68.0)Sexual orientation .474LGBT 12 (15.0) 191 (12.3)Heterosexual 68 (85.0) 1363 (87.7)HIV status .749Positive 30 (37.5) 611 (39.3)Negative 50 (62.5) 944 (60.7)History of suicidal behavior at baseline <.001Yes 55 (68.7) 356 (22.9)No 24 (31.3) 1137 (77.1)Baseline depression scorec 35 (26–45) 24 (14–33) <.001Note. IQR = interquartile range; LGBT = lesbian, gay, bisexual, or transgender. Columns do not add to 100% because ofmissing or unavailable data.aAt first interview during follow-up.bSelf-identified First Nation, Inuit, or Me´tis.cScore on Center for Epidemiologic Studies Depression scale ‡ 22 indicated major depressive symptomology.RESEARCH AND PRACTICESeptember 2013, Vol 103, No. 9 | American Journal of Public Health Marshall et al. | Peer Reviewed | Research and Practice | e71incidence of attempting suicide was particularlyelevated among persons with severe to extremelevels of childhood traumatic experiences.Table 3 depicts the crude, adjusted, andadjusted with IPCW recurrent event models foreach type of childhood trauma. Plots ofSchoenfeld residuals and inclusion of productterms between variables of interest and logtime did not suggest evidence of nonpropor-tionality in any of the 5 models. In adjustedand weighted proportional means models, se-vere to extreme sexual (adjusted hazard ratio[AHR] = 2.46; 95% CI = 1.37, 4.42), physical(AHR = 2.00; 95% CI = 1.06, 3.78), andemotional (AHR = 3.52; 95% CI = 1.42, 8.71)abuse were all significantly associated with anincreased risk of suicidal behavior. By contrast,severe to extreme childhood physical (AHR =1.43; 95% CI = 0.71, 2.91) and emotional(AHR= 1.37; 95% CI = 0.70, 2.69) neglectdid not predict suicide attempts during follow-up. The adjusted HRs corresponding to the lowto moderate and moderate to severe categoriesfor all types of childhood maltreatment werenot significant.In a subanalysis entering all consideredcovariates and each type of abuse into 1 model,severe to extreme childhood sexual abuseremained positively associated with suicidalbehavior (AHR = 2.77; 95% CI = 1.46, 5.23),but severe to extreme physical (AHR = 0.73;95% CI = 0.35, 1.51) and emotional (AHR =2.10; 95% CI = 0.75, 5.90) abuse lost statisti-cal significance. The results of the standardCox proportional hazards regression weresimilar to those obtained in the recurrent eventmodels (Table A, available as a supplementto the online version of this article at http://www.ajph.org).DISCUSSIONIn this prospective cohort study of more than1600 illicit drug users, a lifetime history ofsuicidal behavior was highly prevalent; fur-thermore, suicide attempts were common overthe 5-year study period. We observed anincreased hazard of suicidal behavior amongpersons who reported experiencing severe toextreme levels of childhood physical, emo-tional, and sexual abuse in weighted andadjusted recurrent event analyses.To our knowledge, ours is the first studyto examine prospectively the relationshipbetween suicidal behavior and exposure tochildhood trauma among a community-recruited cohort of drug users. The observedlifetime prevalence of attempting suicide (25%)in our cohort was consistent with other samplesof injection and noninjection drug users,which have reported prevalence estimatesbetween 17% and 45%.11,42---44 Suicidal be-havior was significantly more common thanin studies of the general population; forexample, our observed incidence of attempt-ing suicide was at least 5 times the rate inrepresentative samples of adult populationsin the Netherlands,45 Australia,46 and theUnited States.47Our finding that emotional abuse wasstrongly associated with an increased hazard ofsuicide attempts supports a small but growingbody of cross-sectional studies suggestingthat this type of maltreatment is, in addition tosexual and physical abuse, an important riskfactor for suicidal behavior.48,49 For example,a study of low-income African Americanwomen demonstrated that persons who expe-rienced any type of childhood abuse—physical,sexual, or emotional—were significantly morelikely than women with no history of abuse toreport suicidal behavior.50 The strong associa-tions between emotional abuse and suicidalbehavior observed in our study and othersmerit further investigation but nonethelessTABLE 2—Associations of Childhood Trauma With Suicide Attempts Reported DuringFollow-Up: Vancouver Injection Drug Users Study and AIDS Care Cohort to EvaluateExposure to Survival Services, Vancouver, British Columbia, 2005–2010Type of MaltreatmentaTotal (n = 1634),No. (%)Attempted Suicide(n = 80), No. (%)Did Not AttemptSuicide (n = 1554),No. (%) PbSexual abuse <.001None or minimal 960 (60.2) 25 (32.0) 935 (61.7)Low to moderate 127 (8.0) 4 (5.1) 123 (8.1)Moderate to severe 135 (8.5) 6 (7.7) 129 (8.5)Severe to extreme 372 (23.3) 43 (55.1) 329 (21.7)Physical abuse .004None or minimal 720 (44.9) 24 (31.2) 696 (45.6)Low to moderate 301 (18.7) 10 (13.0) 291 (19.0)Moderate to severe 182 (11.3) 12 (15.6) 170 (11.1)Severe to extreme 402 (25.1) 31 (40.2) 371 (24.3)Emotional abuse <.001None or minimal 497 (31.3) 8 (10.2) 489 (32.4)Low to moderate 323 (20.4) 14 (18.0) 309 (20.5)Moderate to severe 248 (15.6) 15 (19.2) 233 (15.5)Severe to extreme 518 (32.7) 41 (52.6) 477 (31.6)Physical neglect <.001None or minimal 652 (40.7) 23 (35.9) 629 (41.3)Low to moderate 429 (26.8) 16 (20.5) 413 (27.2)Moderate to severe 273 (17.1) 11 (14.1) 262 (17.2)Severe to extreme 246 (15.4) 28 (29.5) 218 (14.3)Emotional neglect .007None or minimal 416 (26.2) 13 (16.7) 403 (26.7)Low to moderate 328 (20.6) 10 (12.8) 318 (21.1)Moderate to severe 301 (19.0) 15 (19.2) 286 (18.9)Severe to extreme 543 (34.2) 40 (51.3) 503 (33.3)Note. Not all columns add to 1634 because of missing values.aDerived from responses to the Childhood Trauma Questionnaire.30bTwo-sided, from the v2 test with 3 df.RESEARCH AND PRACTICEe72 | Research and Practice | Peer Reviewed | Marshall et al. American Journal of Public Health | September 2013, Vol 103, No. 9suggest that emotional maltreatment should beaddressed in therapeutic and clinical contexts.The etiologic mechanisms through whichchildhood trauma heightens vulnerability tosuicidal behavior have been the focus ofclinical, epidemiological, and genetic studies.Evidence shows that early exposure to stressand trauma adversely affects brain develop-ment, which in turn results in an increased riskof psychopathological symptoms.51 Many me-diating factors have also been identified, forexample, alcoholism,52 depression,53,54aggression,55 posttraumatic stress disorder,53and impulsivity.56 We were unable to identifymediational mechanisms because of limitedpower; however, it is likely that some of thesepathways explain the relationships we observed.Clinicians and community health profes-sionals working with illicit drug users should beaware that persons with a history of severetraumatic childhood experiences are at highrisk for suicidal behavior. Therefore identifyingand treating persons with severe childhoodtrauma may have a significant impact on thispublic health problem. In addition, clinicianswho identify suicide attempts should rou-tinely assess patients for childhood abuse toaddress these problems and reduce the risk ofrecurrent suicidal behavior.57 Finally, theintegration of suicide prevention interven-tions within programs frequently accessed bydrug users, such as community health clinicsand needle exchange programs,58 requiresincreased attention, consideration, andevaluation.LimitationsAlthough we were able to account forseveral important risk factors associated withsuicidal behavior, there exists the possibility ofunmeasured confounding. Specifically, wewere unable to examine early childhood envi-ronmental and familial factors that have beenshown to frequently co-occur with childhoodabuse and neglect.49,59 We were also unableto account for recently identified geneticfactors that have been shown to moderate therelationship between childhood traumatic ex-periences and adulthood psychopathol-ogies.60,61 The lengthy period between expo-sure to childhood trauma and participants’responses likely led to some under- or mis-reporting of some traumatic experiences as aresult of response biases. We attempted toreduce the degree of exposure misclassifi-cation by relying on a previously validatedinstrument that showed a high level ofinternal consistency among study participants.It is also likely that some suicide attemptswere either not reported because of recall erroror were considered unintentional overdoses bystudy participants. Whenever possible, nursesand trained interviewers attempted to distin-guish suicide events from overdose experi-ences, and they assured participants of confi-dentiality throughout the interviewing processto minimize underreporting of suicidality. Weexcluded 158 participants who refused toprovide information regarding childhood trau-matic experiences, which may have introducedselection bias. Only a small proportion of re-spondents did not complete the CTQ, so weexpect that the magnitude of this bias, ifpresent, was small. Finally, in light of the non-random sampling procedures, the results maynot be generalizable to drug users in otherurban areas.TABLE 3—Crude and Adjusted Hazard Ratios for Attempting Suicide During Follow-Up From5 Recurrent Event Models for Types of Childhood Trauma: Vancouver Injection Drug UsersStudy and AIDS Care Cohort to Evaluate Exposure to Survival Services, Vancouver, BritishColumbia, 2005–2010Type of MaltreatmentaParticipants,No.SuicideAttempts, No.Crude HR(95% CI)AHRb(95% CI)AHR With IPCWb,c(95% CI)Sexual abuseNone to minimal 960 30 1.00 (Ref) 1.00 (Ref) 1.00 (Ref)Low to moderate 127 4 1.20 (0.41, 3.49) 0.92 (0.29, 2.93) 0.85 (0.28, 2.56)Moderate to severe 135 7 1.65 (0.66, 4.18) 1.05 (0.31, 3.58) 0.52 (0.13, 2.13)Severe to extreme 372 54 5.05 (3.00, 8.52) 2.81 (1.54, 5.14) 2.46 (1.37, 4.42)Physical abuseNone to minimal 720 29 1.00 (Ref) 1.00 (Ref) 1.00 (Ref)Low to moderate 301 11 1.05 (0.49, 2.25) 0.81 (0.34, 1.94) 1.16 (0.45, 2.96)Moderate to severe 182 13 1.86 (0.91, 3.82) 1.10 (0.44, 2.75) 1.11 (0.37, 3.39)Severe to extreme 402 41 2.79 (1.59, 4.89) 1.57 (0.83, 2.99) 2.00 (1.06, 3.78)Emotional abuseNone to minimal 497 8 1.00 (Ref) 1.00 (Ref) 1.00 (Ref)Low to moderate 323 20 3.45 (1.42, 8.38) 2.32 (0.88, 6.14) 2.43 (0.94, 6.29)Moderate to severe 248 16 3.85 (1.63, 9.08) 2.01 (0.77, 5.24) 1.41 (0.50, 3.95)Severe to extreme 518 51 6.53 (3.02, 14.12) 2.85 (1.20, 6.76) 3.52 (1.42, 8.71)Physical neglectNone to minimal 652 26 1.00 (Ref) 1.00 (Ref) 1.00 (Ref)Low to moderate 429 21 1.22 (0.62, 2.41) 0.79 (0.37, 1.70) 1.02 (0.45, 2.30)Moderate to severe 273 13 1.09 (0.52, 2.29) 0.57 (0.22, 1.49) 0.57 (0.22, 1.50)Severe to extreme 246 34 3.72 (2.09, 6.61) 1.60 (0.81, 3.18) 1.43 (0.71, 2.91)Emotional neglectNone to minimal 416 16 1.00 (Ref) 1.00 (Ref) 1.00 (Ref)Low to moderate 328 12 1.09 (0.46, 2.58) 0.92 (0.39, 2.15) 0.92 (0.38, 2.19)Moderate to severe 301 19 1.83 (0.83, 4.02) 1.17 (0.50, 2.72) 1.22 (0.49, 3.04)Severe to extreme 543 48 2.59 (1.35, 4.95) 1.48 (0.76, 2.88) 1.37 (0.70, 2.69)Note. AHR = adjusted hazard ratio; CI = confidence interval; IPCW = inverse probability of censoring weights.aDerived from responses to the Childhood Trauma Questionnaire.30bAll models adjusted for age, gender, aboriginal ancestry, sexual orientation, HIV status, lifetime history of suicidal behavior atbaseline, depression, active injection drug use, homelessness, victimization, recent overdose experience, and currentenrollment in drug or alcohol treatment.cFive sets of censoring weights were constructed from pooled logistic models with the following covariates: age, gender,aboriginal ancestry, lifetime history of suicidal behavior at baseline, depression, victimization, recent overdose experience,and each type of childhood trauma (for each of 5 models).RESEARCH AND PRACTICESeptember 2013, Vol 103, No. 9 | American Journal of Public Health Marshall et al. | Peer Reviewed | Research and Practice | e73ConclusionsOur study adds to a growing body of re-search suggesting that severe childhood mal-treatment confers significant detrimental im-pacts, including an elevated risk of suicidalbehavior in adulthood. Additional research isrequired to elucidate the etiological mecha-nisms and mediational pathways that explainthese relationships. Continued developmentand evaluation of effective interventions toprevent suicide in high-risk populations ofdrug users are required. jAbout the AuthorsAt the time of writing, Brandon D. L. Marshall and SandroGalea were with the Department of Epidemiology, MailmanSchool of Public Health, Columbia University, New York,NY. Brandon D. L. Marshall was also with and Evan Woodand Thomas Kerr are with the Urban Health ResearchInitiative, British Columbia Centre for Excellence in HIV/AIDS, Vancouver. Evan Wood and Thomas Kerr are alsowith the Department of Medicine, University of BritishColumbia, Vancouver.Correspondence should be sent to Brandon D. L.Marshall, PhD, Dept of Epidemiology, Brown University,Box G-S-121-2, 121 South Main St, Providence, RI02912 (e-mail: brandon_marshall@brown.edu). Reprintscan be ordered at http://www.ajph.org by clicking the“Reprints” link.This article was accepted January 23, 2013.ContributorsE. Wood and T. Kerr originated the Vancouver In-jection Drug Users Study and the AIDS Care Cohort toEvaluate Exposure to Survival Services study andcontributed to their development and procurement offunding. B. D. L. Marshall conducted the literature re-view, designed and conducted the analysis, and directedits implementation, with input from S. Galea and T.Kerr. B. D. L. Marshall wrote the article, and all authorscritically revised it, provided important intellectualcontent, and approved the final version.AcknowledgmentsThis study was funded by the National Institutes ofHealth (grants R01 DA011591 and R01 DA021525).T. Kerr is also supported by a Senior Scholar Award fromthe Michael Smith Foundation for Health Research (CI-SCH-085 05-1) and a New Investigator Award from theCanadian Institutes of Health Research (MSH-80J36). Atthe time of writing, B. D. L. Marshall was supported bya Fellowship Award from the Canadian Institutes ofHealth Research. E. Wood is supported through a CanadaResearch Chair.We thank the study participants for their contributionto the research, as well as current and past investigatorsand staff. We thank Deborah Graham, Peter Vann,Caitlin Johnston, and Steve Kain for their research andadministrative assistance.Human Participant ProtectionThe Vancouver Injection Drug Users Study and AIDSCare Cohort to Evaluate Exposure to Survival Servicesstudy were approved by the University of British Co-lumbia and Providence Health Care Research Instituteresearch ethics boards. All participants provided in-formed consent.References1. Hawton K, van Heeringen K. Suicide. Lancet.2009;373(9672):1372---1381.2. Statistics Canada. Leading Causes of Death in Canada,2008. Ottawa, Ontario: Statistics Canada; 2011.3. Murphy SL, Xu J, Kochanek KD. Deaths: preliminarydata for 2010. Natl Vital Stat Rep. 2012;60(4):1---68.4. Crosby AE, Han B, Ortega LA, Parks SE, Gfroerer J.Suicidal thoughts and behaviors among adults aged‡18 years—United States, 2008---2009. MMWR Sur-veill Summ. 2011;60(13):1---22.5. Office of the Surgeon General and National ActionAlliance for Suicide Prevention. 2012 National Strategyfor Suicide Prevention: Goals and Objectives for Action.Washington, DC: US Department of Health and HumanServices; 2012.6. Malchy B, Enns MW, Young TK, Cox BJ. Suicideamong Manitoba’s Aboriginal People, 1988 to 1994.CMAJ. 1997;156(8):1133---1138.7. Cutcliffe JR. Toward an understanding of suicide inFirst-Nation Canadians. Crisis. 2005;26(3):141---145.8. Garofalo R, Wolf RC, Wissow LS, Woods ER,Goodman E. Sexual orientation and risk of suicideattempts among a representative sample of youth. ArchPediatr Adolesc Med. 1999;153(5):487---493.9. Silenzio VM, Pena JB, Duberstein PR, Cerel J,Knox KL. Sexual orientation and risk factors forsuicidal ideation and suicide attempts among adoles-cents and young adults. Am J Public Health. 2007;97(11):2017---2019.10. Remafedi G, French S, Story M, Resnick MD, BlumR. The relationship between suicide risk and sexualorientation: results of a population-based study. Am JPublic Health. 1998;88(1):57---60.11. Darke S, Ross J. Suicide among heroin users: rates,risk factors and methods. Addiction. 2002;97(11):1383---1394.12. Bohnert AS, Roeder K, Ilgen MA. Unintentionaloverdose and suicide among substance users: a review ofoverlap and risk factors. Drug Alcohol Depend. 2010;110(3):183---192.13. Borges G, Walters EE, Kessler RC. Associations ofsubstance use, abuse, and dependence with subsequentsuicidal behavior. Am J Epidemiol. 2000;151(8):781---789.14. Wilcox HC, Conner KR, Caine ED. Association ofalcohol and drug use disorders and completed suicide: anempirical review of cohort studies. Drug Alcohol Depend.2004;76(suppl):S11---S19.15. Malbergier A, de Andrade AG. Depressive disordersand suicide attempts in injecting drug users with andwithout HIV infection. AIDS Care. 2001;13(1):141---150.16. Johnsson E, Fridell M. Suicide attempts in a cohort ofdrug abusers: a 5-year follow-up study. Acta PsychiatrScand. 1997;96(5):362---366.17. Aharonovich E, Liu X, Nunes E, Hasin DS. Suicideattempts in substance abusers: effects of major depressionin relation to substance use disorders. Am J Psychiatry.2002;159(9):1600---1602.18. Landheim AS, Bakken K, Vaglum P. What charac-terizes substance abusers who commit suicide attempts?Factors related to Axis I disorders and patterns ofsubstance use disorders. A study of treatment-seekingsubstance abusers in Norway. Eur Addict Res. 2006;12(2):102---108.19. Bohnert AS, Roeder KM, Ilgen MA. Suicide attemptsand overdoses among adults entering addictions treat-ment: comparing correlates in a U.S. National Study. DrugAlcohol Depend. 2011;119(1---2):106---112.20. Maloney E, Degenhardt L, Darke S, Mattick RP,Nelson E. Suicidal behaviour and associated risk factorsamong opioid-dependent individuals: a case-controlstudy. Addiction. 2007;102(12):1933---1941.21. Ilgen MA, Harris AH, Moos RH, Tiet QQ. Predictorsof a suicide attempt one year after entry into substanceuse disorder treatment. Alcohol Clin Exp Res. 2007;31(4):635---642.22. Darke S, Ross J, Williamson A, Mills KL, Havard A,Teesson M. Patterns and correlates of attempted suicideby heroin users over a 3-year period: findings from theAustralian treatment outcome study. Drug Alcohol De-pend. 2007;87(2---3):146---152.23. Marshall BDL, Galea S, Wood E, Kerr T. Injectionmethamphetamine use is associated with an increasedrisk of attempted suicide: a prospective cohort study.Drug Alcohol Depend. 2011;119(1---2):134---137.24. Lloyd JJ, Ricketts EP, Havens JR, et al. The re-lationship between lifetime abuse and suicidal ideation ina sample of injection drug users. J Psychoactive Drugs.2007;39(2):159---166.25. Roy A. Risk factors for attempting suicide in heroinaddicts. Suicide Life Threat Behav. 2010;40(4):416---420.26. Tyndall MW, Currie S, Spittal P, et al. Intensiveinjection cocaine use as the primary risk factor in theVancouver HIV-1 epidemic. AIDS. 2003;17(6):887---893.27. Strathdee SA, Patrick DM, Currie SL, et al. Needleexchange is not enough: lessons from the Vancouverinjecting drug use study. AIDS. 1997;11(8):F59---F65.28. Bernstein DP, Fink L. Childhood Trauma Question-naire: A Retrospective Self-Report Manual. San Antonio,TX: Psychological Corp; 1998.29. Scher CD, Stein MB, Asmundson GJ, McCreary DR,Forde DR. The Childhood Trauma Questionnaire ina community sample: psychometric properties and nor-mative data. J Trauma Stress. 2001;14(4):843---857.30. Bernstein DP, Stein JA, Newcomb MD, et al. De-velopment and validation of a brief screening versionof the Childhood Trauma Questionnaire. Child AbuseNegl. 2003;27(2):169---190.31. Ducci F, Roy A, Shen PH, et al. Association ofsubstance use disorders with childhood trauma but notAfrican genetic heritage in an African American cohort.Am J Psychiatry. 2009;166(9):1031---1040.32. Walton G, Co SJ, Milloy MJ, Qi J, Kerr T, Wood E.High prevalence of childhood emotional, physical andsexual trauma among a Canadian cohort of HIV-seropositive illicit drug users. AIDS Care. 2011;23(6):714---721.33. Radloff LS. The CES-D scale: a self-report depres-sion scale for research in the general population. ApplPsychol Meas. 1977;1(3):385---401.RESEARCH AND PRACTICEe74 | Research and Practice | Peer Reviewed | Marshall et al. American Journal of Public Health | September 2013, Vol 103, No. 934. Kaplan EL, Meier P. Nonparametric estimation fromincomplete observations. J Am Stat Assoc. 1958;53(282):457---481.35. Lin DY, Wei LJ, Yang I, Ying Z. Semiparametricregression for the mean and rate functions of recurrentevents. J R Stat Soc Series B Stat Methodol. 2000;62(4):711---730.36. Lin DY. Proportional means regression for censoredmedical costs. Biometrics. 2000;56(3):775---778.37. Grambsch PM, Therneau TM. Proportional hazardstests and diagnostics based on weighted residuals. Bio-metrika. 1994;81(3):515---526.38. Hess KR. Assessing time-by-covariate interactions inproportional hazards regression models using cubicspline functions. Stat Med. 1994;13(10):1045---1062.39. Robins JM, Finkelstein DM. Correcting for non-compliance and dependent censoring in an AIDS clinicaltrial with inverse probability of censoring weighted(IPCW) log-rank tests. Biometrics. 2000;56(3):779---788.40. Cole SR, Hernan MA, Margolick JB, Cohen MH,Robins JM. Marginal structural models for estimating theeffect of highly active antiretroviral therapy initiationon CD4 cell count. Am J Epidemiol. 2005;162(5):471---478.41. Cain LE, Cole SR. Inverse probability-of-censoringweights for the correction of time-varying noncompliancein the effect of randomized highly active antiretroviraltherapy on incident AIDS or death. Stat Med. 2009;28(12):1725---1738.42. Darke S, Kaye S. Attempted suicide among injectingand noninjecting cocaine users in Sydney, Australia. JUrban Health. 2004;81(3):505---515.43. Glasner-Edwards S, Mooney LJ, Marinelli-Casey P,Hillhouse M, Ang A, Rawson R. Risk factors for suicideattempts in methamphetamine-dependent patients. AmJ Addict. 2008;17(1):24---27.44. Rossow I, Lauritzen G. Shattered childhood: a keyissue in suicidal behavior among drug addicts? Addiction.2001;96(2):227---240.45. ten Have M, de Graaf R, van Dorsselaer S, et al.Incidence and course of suicidal ideation and suicideattempts in the general population. Can J Psychiatry.2009;54(12):824---833.46. Pirkis J, Burgess P, Dunt D. Suicidal ideation andsuicide attempts among Australian adults. Crisis.2000;21(1):16---25.47. Kessler RC, Berglund P, Borges G, Nock M, WangPS. Trends in suicide ideation, plans, gestures, andattempts in the United States, 1990---1992 to 2001---2003. JAMA. 2005;293(20):2487---2495.48. Twomey HB, Kaslow NJ, Croft S. Childhood mal-treatment, object relations, and suicidal behavior inwomen. Psychoanal Psychol. 2000;17(2):313---335.49. Brown J, Cohen P, Johnson JG, Smailes EM. Child-hood abuse and neglect: specificity of effects onadolescent and young adult depression and suicidal-ity. J Am Acad Child Adolesc Psychiatry. 1999;38(12):1490---1496.50. Anderson PL, Tiro JA, Price AW, Bender MA,Kaslow NJ. Additive impact of childhood emotional,physical, and sexual abuse on suicide attempts amonglow-income African American women. Suicide Life ThreatBehav. 2002;32(2):131---138.51. Teicher MH, Andersen SL, Polcari A, Anderson CM,Navalta CP. Developmental neurobiology of childhoodstress and trauma. Psychiatr Clin North Am. 2002;25(2):397---426.52. Dube SR, Anda RF, Felitti VJ, Chapman DP,Williamson DF, Giles WH. Childhood abuse, householddysfunction, and the risk of attempted suicide throughoutthe life span: findings from the Adverse ChildhoodExperiences Study. JAMA. 2001;286(24):3089---3096.53. Bedi S, Nelson EC, Lynskey MT, et al. Risk forsuicidal thoughts and behavior after childhood sexualabuse in women and men. Suicide Life Threat Behav.2011;41(4):406---415.54. Sigfusdottir ID, Asgeirsdottir BB, Gudjonsson GH,Sigurdsson JF. A model of sexual abuse’s effects onsuicidal behavior and delinquency: the role of emotionsas mediating factors. J Youth Adolesc. 2008;37(6):699---712.55. Swogger MT, You S, Cashman-Brown S, Conner KR.Childhood physical abuse, aggression, and suicide at-tempts among criminal offenders. Psychiatry Res. 2011;185(3):363---367.56. Braquehais MD, Oquendo MA, Baca-Garcia E, SherL. Is impulsivity a link between childhood abuse andsuicide? Compr Psychiatry. 2010;51(2):121---129.57. Tiet QQ, Finney JW, Moos RH. Recent sexual abuse,physical abuse, and suicide attempts among male vet-erans seeking psychiatric treatment. Psychiatr Serv.2006;57(1):107---113.58. Havens JR, Strathdee SA, Fuller CM, et al. Correlatesof attempted suicide among young injection drug users ina multi-site cohort. Drug Alcohol Depend. 2004;75(3):261---269.59. Drake B, Pandey S. Understanding the relationshipbetween neighborhood poverty and specific types ofchild maltreatment. Child Abuse Negl. 1996;20(11):1003---1018.60. Roy A, Hu XZ, Janal MN, Goldman D. Interactionbetween childhood trauma and serotonin transportergene variation in suicide. Neuropsychopharmacology.2007;32(9):2046---2052.61. Caspi A, McClay J, Moffitt TE, et al. Role of genotypein the cycle of violence in maltreated children. Science.2002;297(5582):851---854.RESEARCH AND PRACTICESeptember 2013, Vol 103, No. 9 | American Journal of Public Health Marshall et al. | Peer Reviewed | Research and Practice | e75


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items