Journal of Affective Disorders 166 (2014) 1–5
Contents lists available at ScienceDirect
Journal of Affective Disorders journal homepage: www.elsevier.com/locate/jad
Research report
Posttraumatic stress disorder and suicide in 5.9 million individuals receiving care in the veterans health administration health system Kenneth R. Conner a,b,n, Robert M. Bossarte a,b, Hua He a,b, Jyoti Arora a,b, Naiji Lu a,b, Xin M. Tu a,b, Ira R. Katz c a
US Department of Veterans Affairs (VA) VISN 2 Center of Excellence for Suicide Prevention, USA University of Rochester Medical Center, USA c VA Office of Mental Health Operations, USA b
art ic l e i nf o
a b s t r a c t
Article history: Received 24 April 2014 Accepted 25 April 2014 Available online 5 May 2014
Background: Post-traumatic stress disorder (PTSD) confers risk for suicidal ideation and suicide attempts but a link with suicide is not yet established. Prior analyses of users of the Veterans health administration (VHA) Health System suggest that other mental disorders strongly influence the association between PTSD and suicide in this population. We examined the association between PTSD and suicide in VHA users, with a focus on the influence of other mental disorders. Methods: Data were based on linkage of VA National Patient Care Database records and the Centers for Disease Control and Prevention's National Death Index, with data from fiscal year 2007–2008. Analyses were based on multivariate logistic regression and structural equation models. Results: Among users of VHA services studied (N¼ 5,913,648), 0.6% (N¼3620) died by suicide, including 423 who had had been diagnosed with PTSD. In unadjusted analysis, PTSD was associated with increased risk for suicide, with odds ratio, OR (95% confidence interval, 95% CI) ¼1.34 (1.21, 1.48). Similar results were obtained after adjustment for demographic variables and veteran characteristics. After adjustment for multiple other mental disorder diagnoses, PTSD was associated with decreased risk for suicide, OR (95% CI)¼ 0.77 (0.69, 0.86). Major depressive disorder (MDD) had the largest influence on the association between PTSD and suicide. Limitations: The analyses were cross-sectional. VHA users were studied, with unclear relevance to other populations. Conclusion: The findings suggest the importance of identifying and treating comorbid MDD and other mental disorders in VHA users diagnosed with PTSD in suicide prevention efforts. & 2014 Published by Elsevier B.V.
Keywords: Suicide Veteran PTSD Comorbid Risk factor
1. Introduction Suicide is the 10th leading cause of death in the United States, accounting for approximately 38,000 deaths annually (Hoyert and Xu, 2012). There is solid evidence that posttraumatic stress disorder (PTSD) is associated with suicidal thoughts and nonlethal suicide attempts (Krysinska and Lester, 2010; Panagioti et al., 2012). However, far more limited are studies of PTSD and suicide deaths (heretofore referred to as suicide). Along these lines, metaanalyses have not demonstrated that PTSD is associated with suicide (Krysinska and Lester, 2010; Panagioti et al., 2012). n Correspondence to: VISN 2 Center of Excellence for Suicide Prevention, Department of Veterans Affairs, 400 Fort Hill Avenue, Canandaigua, NY 14424, USA. Tel.: þ1 585 393 7548; fax: þ1 585 393 7985. E-mail address: [email protected] (K.R. Conner).
http://dx.doi.org/10.1016/j.jad.2014.04.067 0165-0327/& 2014 Published by Elsevier B.V.
Although the role of PTSD in suicide is unclear and the study of PTSD and suicide overall is at a nascent stage, there is a growing database on PTSD and suicide in users of the Veterans health administration (VHA) healthcare system (Bullman and Kang, 1994; Desai et al., 2005; Ilgen et al., 2010; Zivin et al., 2007). As a result, there is an opportunity to take stock of what has been learned about PTSD and suicide in VHA users in order to inform hypothesisdriven research moving forward. Such work is critical because VHA is the largest integrated health care system in the United States and individuals who use VHA services, the large majority of whom are Veterans, are at increased risk for suicide compared to the US general population (McCarthy et al., 2009). Taken as a whole, studies of VHA users suggest the central importance of considering mental disorders that co-occur with PTSD in assessing the link between PTSD and suicide. In particular, studies of general samples of VHA users (i.e., those not selected based on clinical diagnosis or
2
K.R. Conner et al. / Journal of Affective Disorders 166 (2014) 1–5
treatment setting) have shown that PTSD is associated with suicide including analyses of a cohort who had been deployed to Viet Nam (Bullman and Kang, 1994) and analyses of a more recent study of the total population of VHA users that adjusted for age (Ilgen et al., 2010). In the study of Viet Nam Veterans the association between PTSD and suicide was reduced (though not eliminated) after adjustment for mental disorders (Bullman and Kang, 1994). Studies of clinical populations of VHA users with high rates of mental disorders have not concluded that PTSD is associated with suicide including analyses of cohort of patients discharged from inpatient psychiatric treatment that adjusted for comorbid mental disorders (Desai et al., 2005) and analyses of a cohort of patients, all with depressive disorders, which showed that a diagnosis of comorbid PTSD is associated with lowered risk for suicide (Zivin et al., 2007). In summary, the results suggest that PTSD is associated with increased risk for suicide in VHA users generally (Ilgen et al., 2010) but the association may be attenuated (Bullman and Kang, 1994), not observed at a statistically significant level (Desai et al., 2005), or PTSD may be associated with lower risk (Zivin et al., 2007) in studies that adjust for other mental disorders and/or in analyses of clinical populations with high rates of mental disorders. There are several potential explanations for attenuation of the relationship between PTSD and suicide after adjustment for mental disorders among VHA users that include: there are many common contributors to PTSD and other mental disorders (e.g., trauma history) which may lower the independent contribution to suicide risk of different mental disorders when they are modeled simultaneously; the difficulty of accurate differential diagnosis in complex patients, for example the potential for severe PTSD to be mislabeled as other condition(s); the ability to tolerate PTSD without developing other mental health or substance use disorders as complications may indicate less severe cases and/or increased resilience and stress tolerance; and that the identification of mental disorders comorbid with PTSD may lead to more intensive clinical care and monitoring which may serve to lower risk for suicidal behavior among Veterans treated in VHA (Conner et al., 2013; Desai et al., 2005; Zivin et al., 2007). Another potential explanation is that PTSD may lead to or exacerbate the symptoms of other mental disorders, for example major depressive disorder (MDD), which may in turn confer risk for suicide (Panagioti et al., 2009). In this scenario PTSD plays a causal role in suicide risk albeit an indirect one (i.e., by promoting other mental disorders), consistent with the concept of mediation (Kaplan, 2000; Xia et al., 2012). With few exceptions (Gradus et al., 2010), reports of PTSD and suicide are based on users of VHA services, per above. However, a recent report of current and former US military service members is topical insofar as it examined the roles of PTSD, comorbid conditions, and deployment history in suicide (LeardMann et al., 2013). The investigators reported that PTSD is not associated with suicide in unadjusted analyses or in analyses that adjusted for age and sex and therefore conducted no further analyses of PTSD. Military deployment history was associated with risk for suicide in models with minimal adjustment but not after additional adjustment for depression and other correlates. The study had a moderate number of suicides (N ¼83), limiting statistical power. The purpose of the current study was to examine the association of PTSD and suicide in a well powered, national analysis of VHA service users. We hypothesized that PTSD is associated with suicide in unadjusted analyses and those that adjust for demographic variables and military service during the wars in Afghanistan (i.e., Operation Enduring Freedom, OEF) or Iraq (i.e., Operation Iraqi Freedom, OIF). We further hypothesized that the association between PTSD and suicide is attenuated, if not eliminated, after statistical adjustment for mental disorders. If our hypothesis is confirmed, particularly if an association between PTSD and suicide is not observed after adjustment for mental
disorders, it would suggest the importance of identifying and treating co-occurring mental disorders in efforts to reduce suicide risk in VHA users with PTSD. We also explored the potential mediating role of other mental disorders on the association between PTSD and suicide.
2. Method Data sources: Data are based on linkage of the VA National Patient Care Database (NPCD) and the Centers for Disease Control and Prevention's National Death Index (NDI). The NPCD included demographic and diagnostic information for all treatment contacts of patients seen anywhere within the VHA treatment system. Diagnoses in the NPCD are based on clinical assessments and correspond to the International Classification of Diseases, Ninth Revision, Clinical Modification (Medicode, 1995). NDI provided information about vital status and cause of death for all US residents from state vital statistics offices. Suicide decedents were identified using International Classification of Diseases, 10th Revision, Clinical Modification Codes X60-X84 and Y87.0 (World Health Organization, 2004). Individuals who died by other causes were censored in analyses. For more information on the use of these sources in suicide research see prior reports (McCarthy et al., 2009; Bohnert et al., 2014). Sample: The population analyzed consisted of all users of VHA health care services in FY 2007–2008, (N ¼5,913,648). Measures: PTSD (vs. no. PTSD) was the primary variable of interest. Other measures that served as covariates in analyses included demographic variables, Veteran-specific variables, and other mental disorders. Demographic characteristics include age categorized into four groups (ages 18–29, 30–59, 60–74, and 75 þ), sex, married (vs. unmarried), urban (vs. rural), and region of the country in four categories (northeast, south, Midwest, west). Veteran characteristics included OEF/OIF Veteran (vs. non-OEF/ OIF Veteran), multiple OEF/OIF deployments (vs. non-multiple OEF/OIF deployments), and being new to VHA (vs. experienced users). Along with PTSD, mental disorders included a diagnosis (vs. no such diagnosis) of MDD, non-MDD depression, drug use disorder, alcohol use disorder, bipolar disorder, anxiety disorder (non-PTSD), and schizophrenia. Analyses: Descriptive analytic methods were used to examine characteristics of the total cohort and by PTSD status. The number and percentage of deaths by suicide of the total sample as well as those with and without a diagnosis of PTSD were also calculated. A series of logistic regressions, generalized linear models with the logit link (Tang et al., 2012), were used to estimate the association between PTSD and the risk of suicide, with analyses yielding Odds Ratios (ORs) and corresponding 95% confidence intervals (CIs). These analyses began with an unadjusted model to examine the bivariate association of PTSD and suicide (model 1) and built on this model by adding covariates in blocks including demographic variables (model 2), Veteran-specific variables (model 3), and other mental disorders (model 4). The potential problem of multi-colinearity was examined in models 2, 3 and 4 using the variance inflation factor (Hair et al., 2006). Logistic regression analyses were conducted using SAS 9.3 and these results are presented based on the template used in a recent study of tobacco use disorder and suicide (Bohnert et al., 2014). We ran further analyses using structural equation models (SEM) in order to explore the potential mediating influence of mental disorders on the association between PTSD and suicide (Kaplan, 2000; Xia et al., 2012). In these models the estimated parameters for the various associations or “pathways” are standardized regression coefficients for the probit model for binary responses. A statistically significant positive sign of an estimate
K.R. Conner et al. / Journal of Affective Disorders 166 (2014) 1–5
3
Table 1 Users of VHA health care services in FY 2007–2008 (N ¼5,913,648), total, and by posttraumatic stress disorder (PTSD) status. Characteristic
Total n
Total Age 18–29 Age 30–59 Age 60–74 Age 75þ Male Married Urban Northeast region South region Midwest region West region OEF/OIF Veteran Multiple OEF/OIF deployments New to VHA Posttraumatic stress disorder (PTSD) Major depressive disorder (MDD) Non-MDD depression Drug use disorder Alcohol use disorder Bipolar disorder Anxiety disorder Schizophrenia Absence of non-PTSD mental health disorders Number of suicide deaths
5,913,648 320,693 2,156,693 1,901,189 1,523,172 5,354,429 3,224,100 3,801,339 1,079,320 2,272,851 1,318,305 1,243,172 302,644 106,276 816,737 532,009 314,641 832,040 281,030 434,495 134,774 451,208 110,715 4,297,082 3620
PTSD diagnosis %
Yes
100.00 5.42 36.47 32.15 25.76 90.54 54.52 64.33 18.25 38.43 22.29 21.02 5.12 1.80 13.81 9.00 5.32 14.07 4.75 7.35 2.28 7.63 1.87 72.66 0.06
No
n
%
n
%
532,009 49,720 260,766 178,798 42,701 493,725 290,706 333,924 93,838 212,146 94,832 131,193 78,904 29,542 60,979 532,009 123,836 207,562 78,019 114,905 38,836 119,967 18,412 146,129 423
100.00 9.35 49.02 33.61 8.03 92.80 54.64 62.79 17.64 39.88 17.83 24.66 14.83 5.55 11.46 100.00 23.28 39.01 14.66 21.60 7.30 22.55 3.46 27.46 0.08
5,381,639 270,973 1,895,927 1,722,391 1,480,471 4,860,704 2,933,394 3,467,415 985,482 2,060,705 1,223,473 1,111,979 223,740 76,734 755,758 0 190,805 624,478 203,011 319,590 95,938 331,241 92,303 4,150,953 3197
100.00 5.04 35.23 32.01 27.51 90.32 54.51 64.48 18.31 38.29 22.73 20.66 4.16 1.43 14.04 0.00 3.55 11.60 3.77 5.94 1.78 6.16 1.72 77.13 0.06
Note:The percentages provided are column percentages. Tests of differences of proportions between groups with PTSD and those without PTSD were conducted for all rows, for example the proportion of men (vs. women) with a PTSD diagnosis vs. without a PTSD diagnosis, and in all cases are statistically significant (po 0.001).
indicates that the probability of an event (e.g., suicide) increases when the value of the predictor (e.g., PTSD) increases; a statistically significant negative sign indicates that the probability of an event decreases when the value of the predictor increases. In the current analyses the direct effect is the pathway from PTSD to suicide, while controlling for mental disorder(s). The indirect, or mediation effect, describes the pathway from PTSD to suicide through the mental disorder(s). The total effect is the sum of the direct and indirect effects of the predictor on suicide. Model fits were tested using the Chi-square test and Root Mean Square Error of Approximation (MacCallum et al., 1996). Mplus was used for all SEM analyses (Muthen and Muthen, 2006). 3. Results Descriptive data on the population of VHA users are provided in Table 1 including overall results and those stratified by PTSD diagnosis. Of the total population of nearly 6 million, about nine in ten were male (90.5%); about two-thirds were ages 30–59 (36.5%) or ages 60–74 (32.2%); about two-thirds lived in urban as opposed to rural areas (64.3%); and each of the four regions of the United States were well represented, with the south region being the largest (38.4%). A minority of individuals were new to VHA services (13.8%), were an OEF/OIF Veteran (5.1%), or had a history of multiple OEF/OIF deployments (1.8%). Nine percent of the population had been diagnosed with PTSD. Among those with a PTSD diagnosis, about one quarter (27.5%) had no other mental disorder. A total of 3620 (0.06%) VHA Veterans died by suicide including 423 who had been diagnosed with PTSD. Results of the multiple logistic regression analyses are provided in Table 2. In the unadjusted analysis (model 1), PTSD is associated with suicide at a statistically significant level, with odds ratio, OR (95% confidence interval, 95% CI) ¼1.34 (1.21, 1.48), p o0.001. This result suggests that, prior to adjustment for other variables, VHA
Veterans with a PTSD diagnosis are at about 34% (21–48%) increased risk for suicide compared to those without a PTSD diagnosis. Results are similar after adjustment for demographic variables (model 2) and further adjustment for Veteran-specific variables including OEF/OIF Veteran status and multiple OEF/OIF deployments (model 3). Finally, the result pertaining to PTSD after additional adjustment for other mental disorder diagnoses are dramatically different, OR (95% CI)¼ 0.77 (0.69, 0.86), p o0.001. The result suggests that, after adjustment for mental disorders, VHA Veterans with a PTSD diagnosis are at about 23% (31%, 14%) lower risk for suicide compared to Veterans without a PTSD diagnosis. There was no evidence of multi-colinearity among the explanatory variables across the four models. We ran further analyses in order to investigate the effect of adjustment for other mental disorders on the association between PTSD and suicide. First, we ran a series of multivariate logistic models in which we added, in turn, each of the mental disorders (MDD, etc.) to model 3. These results are provided in Table 3. Covarying MDD stands out for having the largest impact on the association between PTSD and suicide; after adjustment for MDD, PTSD shows no association with suicide, OR (95% CI)¼ 1.00 (0.90, 1.12), p¼ 0.977. To further explore the impact of MDD on the association between PTSD and suicide, we ran an unadjusted SEM model that shows a statistically significant association between PTSD and suicide (standardized regression coefficient¼0.043, po0.001), along with a second SEM model to examine the potential mediating role of MDD on the association between PTSD and suicide. The latter model includes paths from PTSD to suicide, PTSD to MDD, and MDD to suicide, and indicates no direct association between PTSD and suicide (standardized regression coefficient¼ 0.001, p¼0.769), that PTSD is associated with MDD (standardized regression coefficient¼0.294, po0.001), and MDD is associated with suicide (standardized regression coefficient¼0.152, po0.001). Furthermore, the results yield
4
K.R. Conner et al. / Journal of Affective Disorders 166 (2014) 1–5
Table 2 The association between PTSD and risk of death by suicide during FY 2007–2008 among users of VHA healthcare services. n
%
Model 1 OR
PTSD Yes No
532,009 5,381,639
9.00 91.00
Model 2 95% CI
1.340 [1.210, 1.483]
OR
Model 3 95% CI
1.298 [1.171, 1.439]
OR
Model 4 95% CI
1.312 [1.183, 1.456]
OR
95% CI
0.774 [0.693, 0.864]
1. Model 1 is unadjusted. 2. Model 2 is adjusted for demographic variables of age, sex, marital status, urban-rural residence, region of country. 3. Model 3 is adjusted for demographics and Veteran characteristics of OEF/OIF service era, multiple OEF/OIF deployments, and new to VHA. 4. Model 4 is adjusted for demographics, Veteran characteristics, and diagnostic variables of major depressive disorder (MDD), non-MDD depression, alcohol use disorder, drug use disorder, anxiety disorder, bipolar disorder, and schizophrenia.
Table 3 Association of PTSD with suicide after adjustment for model 3 variables and each mental disorder examined separately. Analytic model
OR (PTSD)
95% CI
p-value
Model Model Model Model Model Model Model Model
1.312 1.002 1.099 1.138 1.170 1.153 1.169 1.299
[1.183, 1.456] [0.898, 1.117] [0.987, 1.272] [1.023, 1.265] [1.053, 1.301] [1.037, 1.282] [1.053, 1.299] [1.171, 1.441]
o 0.001 0.977 0.085 0.017 0.004 0.009 0.004 o 0.001
3 3 þmajor depressive disorder (MDD) 3 þnon-MDD depression 3 þalcohol use disorder 3 þdrug use disorder 3 þother anxiety disorder 3 þbipolar disorder 3 þschizophrenia
Model 3 is adjusted for demographics (age, sex, marital status, urban–rural residence, region of country) and Veteran characteristics of OEF/OIF service era, multiple deployments, and new to VA.
Fig. 1. Structural equation model results of the potential mediating role of depressive disorders and substance use disorders on the association between PTSD and suicide in users of VHA health care services in FY 2007–2008, Notes: Path coefficients shown are standardized. Asterisk indicates po0.05. PTSD¼ posttraumatic stress disorder, MDD¼major depressive disorder, non-MDD depression¼depressive disorders other than MDD, drug¼drug use disorder, alcohol¼ alcohol use disorder.
a statistically significant total effect estimate (standardized regression coefficient¼0.043, po0.001) and a statistically significant indirect effect estimate (standardized regression coefficient¼ 0.045,
po0.001). Overall, the models suggest that association between PTSD and suicide is indirect and fully accounted for by MDD, consistent with mediation. The result for multivariate logistic regression model 4 indicates that, when multiple mental disorders are accounted for, PTSD shows an inverse association with suicide. Although an inverse association is not produced using SEM when MDD is considered alone (per above), various combinations of mental disorders that may be conceptualized as mediators of the association between PTSD and suicide produce such an effect. For example we ran a SEM model to explore the mediating effects of depressive disorders (MDD, non-MDD depression) and substance use disorders (AUD, DUD) on the association between PTSD and suicide. The results are shown in Fig. 1 and yield a statistically significant inverse association between PTSD and suicide (standardized regression coefficient¼ 0.116, po0.001), along with a statistically significant total effect (standardized regression coefficient¼0.043, po0.001) and a statistically significant and large indirect effect (standardized regression coefficient¼0.159, po0.001). Note that the indirect effect 0.159 is the difference between the total and direct effects and, because the direct effect is negative, in this instance the indirect effect is larger than the total effect. The results further show that PTSD is positively associated with each mental disorder at a statistically significant level and that each disorder in turn is positively associated with suicide. Overall, these results suggest that the inverse association between PTSD and suicide is attributable to the mediating influence of the other mental disorders, all of which are positively associated with both PTSD and suicide. In the aforementioned SEM models the model fit results indicate that the analyses fit the data well.
4. Discussion The analyses reported here expand upon and clarify previous findings for the total population of VHA users on the impact of PTSD on suicide (Ilgen et al., 2010) by adjusting for the effects of
K.R. Conner et al. / Journal of Affective Disorders 166 (2014) 1–5
other mental disorders. As hypothesized, we identified an association of PTSD and increased risk for suicide in the unadjusted model. Results were similar after adjustment for demographic variables and Veteran-specific variables including OEF/OIF service history. However, the association between PTSD and suicide is significantly affected by adjustment for other mental disorders. Indeed, the change in the association between PTSD and suicide is dramatic, with results suggesting that VHA users with a PTSD diagnosis are at lower risk for suicide than those without a PTSD diagnosis. Potential explanations were discussed earlier and include the fact that there are many common contributors to PTSD and other mental disorders, the difficulty of accurate differential diagnosis in complex patients, the ability to tolerate PTSD without developing other mental health or substance use disorders as complications may indicate less severe cases and/or increased resilience and stress tolerance, and that the identification of mental disorders comorbid with PTSD may lead to more intensive clinical care and monitoring, and that PTSD may lead to other mental disorders (e.g., MDD) which may in turn confer risk for suicide, a mediation scenario that we explored using SEM. There were limitations of the study. The data on mental disorder diagnoses were cross-sectional and, therefore, the mediation analyses are considered exploratory because the temporal ordering of PTSD and other diagnoses was not established. Mental disorder diagnoses were based on clinical records with associated limitations of reliability and validity. Some variables that may also influence the association between PTSD and suicidal behavior, for example social support (DeBeer et al., 2014), are not routinely available in clinical records. Nearly three quarters (72.5%) of VHA users with a PTSD diagnosis had one or more other mental disorders, creating a challenge to the development of policies or practices based on results that a PTSD diagnosis without comorbidity poses no added suicide risk (or is protective). We examined VHA users, with unclear relevance to other populations. Because a small percentage of VHA users are not Veterans, it was not strictly a Veteran sample. There were also many strengths of the study. VHA is the largest healthcare system in the United States. The analysis was of a whole population (VHA users), eliminating concerns about sampling biases within this population. The study was well-powered to examine the association of PTSD and suicide and confounding influences. The analyses examined the potential impact service during the wars in Iraq and Afghanistan on the results. In conclusion the results of the current report show that, among VHA users, clinical diagnoses of PTSD are associated with lower risk for suicide after accounting for other mental disorders. Further research to explain this phenomenon (mediation, etc.) is needed. The findings suggest the importance of identifying and treating comorbid mental disorders (e.g., MDD) that confer risk for suicide in VHA users diagnosed with PTSD in suicide prevention efforts. Role of funding source This research was supported by the Department of Veterans Affairs (VA) Office of Mental Health Services and the VA Office of Mental Health Operations. The views expressed in this report are those of the authors and do not necessarily represent those of the Department of Veterans Affairs.
Conflict of interest The authors have no conflicts of interest to report.
5
Acknowledgments This research was supported by the U.S. Department of Veterans Affairs (VA) Office of Mental Health Services and the VA Office of Mental Health Operations. The views expressed in this report are those of the authors and do not necessarily represent those of the Department of Veterans Affairs. The authors thank William Voss, Ph.D., for a literature review and contributions to an early draft.
References Bohnert, K.M., Ilgen, M.A., McCarthy, J.F., Ignacio, R.V., Blow, F.C., Katz, I.R., 2014. Tobacco use disorder and the risk of suicide mortality. Addiction 109, 155–162. Bullman, T.A., Kang, H.K., 1994. Posttraumatic stress disorder and the risk of traumatic deaths among Vietnam veterans. J. Nerv. Ment. Dis. 182, 604–610. Conner, K.R., Bohnert, A.S., McCarthy, J.F., Valenstein, M., Bossarte, R., Ignacio, R., Lu, N., Ilgen, M.A., 2013. Mental disorder comorbidity and suicide among 2.96 million men receiving care in the Veterans Health Administration health system. J. Abnorm. Psychol. 122, 256–263. DeBeer, B.B., Kimbrel, N.A., Meyer, E.C., Gulliver, S.B., Morissette, S.B., 2014. Combined PTSD and depressive symptoms interact with post-deployment social support to predict suicidal ideation in Operation Enduring Freedom and Operation Iraqi Freedom veterans. Psychiatry Res. 216, 357–362. Desai, R.A., Dausey, D.J., Rosenheck, R.A., 2005. Mental health service delivery and suicide risk: the role of individual patient and facility factors. Am. J. Psychiatry 162, 311–318. Gradus, J.L., Qin, P., Lincoln, A.K., Miller, M., Lawler, E., Sorensen, H.T., Lash, T.L., 2010. Posttraumatic stress disorder and completed suicide. Am. J. Epidemiol. 171, 721–727. Hair, J.F., Anderson, R., Tatham, R.L., Black, W.C., 2006. Multivariate Data Analysis, 6th ed. Prentice Hall, Union Saddle River, N.J. Hoyert, D.L., Xu, J., 2012. Deaths: preliminary data for 2011. National Vital Statistics Reports, Vol. 61 no. 6. Ilgen, M.A., Bohnert, A.S., Ignacio, R.V., McCarthy, J.F., Valenstein, M.M., Kim, H.M., Blow, F.C., 2010. Psychiatric diagnoses and risk of suicide in Veterans. Arch. Gen. Psychiatry 67, 1152–1158. Kaplan, D., 2000. Structural Equation Modeling: Foundations and Extensions, 10th ed. SAGE Publications, Inc., Thousand Oaks, CA. Krysinska, K., Lester, D., 2010. Post-traumatic stress disorder and suicide risk: a systematic review. Arch. Suicide Res. 14, 1–23. LeardMann, C.A., Powell, T.M., Smith, T.C., Bell, M.R., Smith, B., Boyko, E.J., Tomoko, I.H., Gackstetter, G.D., Ghamsary, M., Hoge, C.W., 2013. Risk factors associated with suicide in current and former US military personnel. J. Am. Med. Assoc. 310, 496–506. MacCallum, R.C., Browne, M.W., Sugawara, H.M., 1996. Power analysis and determination of sample size for covariance structure modeling. Psychol. Methods 1, 130–149. McCarthy, J., Valenstein, M., Kim, H.M., Ilgen, M., Zivin, K., Blow, F., 2009. Suicide mortality among patients receiving care in the Veterans Health Administration health system. Am. J. Epidemiol. 169, 1033–1038. Medicode., 1995. ICD-9-CM: International Classification of Diseases; 9th revision, clinical modification. Medicode, Salt Lake City, UT. Muthen, B.O., Muthen, L.K., 2006. M þ user's guide. Muthen and Muthen, Los Angeles, CA. Panagioti, M., Gooding, P.A., Tarrier, N., 2012. A meta-analysis of the association between posttraumatic stress disorder and suicidality: the role of comorbid depression. Compr Psychiatry 53, 915–930. Panagioti, M., Gooding, P., Tarrier, N., 2009. Post-traumatic stress disorder and suicidal behavior: a narrative review. Clin. Psychol. Rev. 29, 471–482. Tang, W., He, H., Tu, X.M., 2012. Applied Categorical and Count Data Analysis. Chapman & Hall/CRC, FL. Xia, Y., Lu, N., Zhang, H., Gunzler, D., Zubenko, G.S., Tu, X.M., 2012. Statistical Methods and Issues in the Study of Suicide. In: Lavigne, J., Kemp, J. (Eds.), Frontiers in Suicide Prevention and Research Hauppauge. Nova Science, NY, pp. 139–158. World Health Organization, 2004. International Statistical Classification of Diseases, 10th Revision (ICD-10) World Health Organization, Geneva, Switzerland. Zivin, K., Kim, H.M., McCarthy, J.F., Austin, K.L., Hoggart, K.J., Walters, H., Valenstein, M., 2007. Suicide mortality among individuals receiving treatment for depression in the Veterans Affairs health system: associations with patient and treatment setting characteristics. Am. J. Public Health 97, 2193–2198.
Contents lists available at ScienceDirect
Journal of Affective Disorders journal homepage: www.elsevier.com/locate/jad
Research report
Posttraumatic stress disorder and suicide in 5.9 million individuals receiving care in the veterans health administration health system Kenneth R. Conner a,b,n, Robert M. Bossarte a,b, Hua He a,b, Jyoti Arora a,b, Naiji Lu a,b, Xin M. Tu a,b, Ira R. Katz c a
US Department of Veterans Affairs (VA) VISN 2 Center of Excellence for Suicide Prevention, USA University of Rochester Medical Center, USA c VA Office of Mental Health Operations, USA b
art ic l e i nf o
a b s t r a c t
Article history: Received 24 April 2014 Accepted 25 April 2014 Available online 5 May 2014
Background: Post-traumatic stress disorder (PTSD) confers risk for suicidal ideation and suicide attempts but a link with suicide is not yet established. Prior analyses of users of the Veterans health administration (VHA) Health System suggest that other mental disorders strongly influence the association between PTSD and suicide in this population. We examined the association between PTSD and suicide in VHA users, with a focus on the influence of other mental disorders. Methods: Data were based on linkage of VA National Patient Care Database records and the Centers for Disease Control and Prevention's National Death Index, with data from fiscal year 2007–2008. Analyses were based on multivariate logistic regression and structural equation models. Results: Among users of VHA services studied (N¼ 5,913,648), 0.6% (N¼3620) died by suicide, including 423 who had had been diagnosed with PTSD. In unadjusted analysis, PTSD was associated with increased risk for suicide, with odds ratio, OR (95% confidence interval, 95% CI) ¼1.34 (1.21, 1.48). Similar results were obtained after adjustment for demographic variables and veteran characteristics. After adjustment for multiple other mental disorder diagnoses, PTSD was associated with decreased risk for suicide, OR (95% CI)¼ 0.77 (0.69, 0.86). Major depressive disorder (MDD) had the largest influence on the association between PTSD and suicide. Limitations: The analyses were cross-sectional. VHA users were studied, with unclear relevance to other populations. Conclusion: The findings suggest the importance of identifying and treating comorbid MDD and other mental disorders in VHA users diagnosed with PTSD in suicide prevention efforts. & 2014 Published by Elsevier B.V.
Keywords: Suicide Veteran PTSD Comorbid Risk factor
1. Introduction Suicide is the 10th leading cause of death in the United States, accounting for approximately 38,000 deaths annually (Hoyert and Xu, 2012). There is solid evidence that posttraumatic stress disorder (PTSD) is associated with suicidal thoughts and nonlethal suicide attempts (Krysinska and Lester, 2010; Panagioti et al., 2012). However, far more limited are studies of PTSD and suicide deaths (heretofore referred to as suicide). Along these lines, metaanalyses have not demonstrated that PTSD is associated with suicide (Krysinska and Lester, 2010; Panagioti et al., 2012). n Correspondence to: VISN 2 Center of Excellence for Suicide Prevention, Department of Veterans Affairs, 400 Fort Hill Avenue, Canandaigua, NY 14424, USA. Tel.: þ1 585 393 7548; fax: þ1 585 393 7985. E-mail address: [email protected] (K.R. Conner).
http://dx.doi.org/10.1016/j.jad.2014.04.067 0165-0327/& 2014 Published by Elsevier B.V.
Although the role of PTSD in suicide is unclear and the study of PTSD and suicide overall is at a nascent stage, there is a growing database on PTSD and suicide in users of the Veterans health administration (VHA) healthcare system (Bullman and Kang, 1994; Desai et al., 2005; Ilgen et al., 2010; Zivin et al., 2007). As a result, there is an opportunity to take stock of what has been learned about PTSD and suicide in VHA users in order to inform hypothesisdriven research moving forward. Such work is critical because VHA is the largest integrated health care system in the United States and individuals who use VHA services, the large majority of whom are Veterans, are at increased risk for suicide compared to the US general population (McCarthy et al., 2009). Taken as a whole, studies of VHA users suggest the central importance of considering mental disorders that co-occur with PTSD in assessing the link between PTSD and suicide. In particular, studies of general samples of VHA users (i.e., those not selected based on clinical diagnosis or
2
K.R. Conner et al. / Journal of Affective Disorders 166 (2014) 1–5
treatment setting) have shown that PTSD is associated with suicide including analyses of a cohort who had been deployed to Viet Nam (Bullman and Kang, 1994) and analyses of a more recent study of the total population of VHA users that adjusted for age (Ilgen et al., 2010). In the study of Viet Nam Veterans the association between PTSD and suicide was reduced (though not eliminated) after adjustment for mental disorders (Bullman and Kang, 1994). Studies of clinical populations of VHA users with high rates of mental disorders have not concluded that PTSD is associated with suicide including analyses of cohort of patients discharged from inpatient psychiatric treatment that adjusted for comorbid mental disorders (Desai et al., 2005) and analyses of a cohort of patients, all with depressive disorders, which showed that a diagnosis of comorbid PTSD is associated with lowered risk for suicide (Zivin et al., 2007). In summary, the results suggest that PTSD is associated with increased risk for suicide in VHA users generally (Ilgen et al., 2010) but the association may be attenuated (Bullman and Kang, 1994), not observed at a statistically significant level (Desai et al., 2005), or PTSD may be associated with lower risk (Zivin et al., 2007) in studies that adjust for other mental disorders and/or in analyses of clinical populations with high rates of mental disorders. There are several potential explanations for attenuation of the relationship between PTSD and suicide after adjustment for mental disorders among VHA users that include: there are many common contributors to PTSD and other mental disorders (e.g., trauma history) which may lower the independent contribution to suicide risk of different mental disorders when they are modeled simultaneously; the difficulty of accurate differential diagnosis in complex patients, for example the potential for severe PTSD to be mislabeled as other condition(s); the ability to tolerate PTSD without developing other mental health or substance use disorders as complications may indicate less severe cases and/or increased resilience and stress tolerance; and that the identification of mental disorders comorbid with PTSD may lead to more intensive clinical care and monitoring which may serve to lower risk for suicidal behavior among Veterans treated in VHA (Conner et al., 2013; Desai et al., 2005; Zivin et al., 2007). Another potential explanation is that PTSD may lead to or exacerbate the symptoms of other mental disorders, for example major depressive disorder (MDD), which may in turn confer risk for suicide (Panagioti et al., 2009). In this scenario PTSD plays a causal role in suicide risk albeit an indirect one (i.e., by promoting other mental disorders), consistent with the concept of mediation (Kaplan, 2000; Xia et al., 2012). With few exceptions (Gradus et al., 2010), reports of PTSD and suicide are based on users of VHA services, per above. However, a recent report of current and former US military service members is topical insofar as it examined the roles of PTSD, comorbid conditions, and deployment history in suicide (LeardMann et al., 2013). The investigators reported that PTSD is not associated with suicide in unadjusted analyses or in analyses that adjusted for age and sex and therefore conducted no further analyses of PTSD. Military deployment history was associated with risk for suicide in models with minimal adjustment but not after additional adjustment for depression and other correlates. The study had a moderate number of suicides (N ¼83), limiting statistical power. The purpose of the current study was to examine the association of PTSD and suicide in a well powered, national analysis of VHA service users. We hypothesized that PTSD is associated with suicide in unadjusted analyses and those that adjust for demographic variables and military service during the wars in Afghanistan (i.e., Operation Enduring Freedom, OEF) or Iraq (i.e., Operation Iraqi Freedom, OIF). We further hypothesized that the association between PTSD and suicide is attenuated, if not eliminated, after statistical adjustment for mental disorders. If our hypothesis is confirmed, particularly if an association between PTSD and suicide is not observed after adjustment for mental
disorders, it would suggest the importance of identifying and treating co-occurring mental disorders in efforts to reduce suicide risk in VHA users with PTSD. We also explored the potential mediating role of other mental disorders on the association between PTSD and suicide.
2. Method Data sources: Data are based on linkage of the VA National Patient Care Database (NPCD) and the Centers for Disease Control and Prevention's National Death Index (NDI). The NPCD included demographic and diagnostic information for all treatment contacts of patients seen anywhere within the VHA treatment system. Diagnoses in the NPCD are based on clinical assessments and correspond to the International Classification of Diseases, Ninth Revision, Clinical Modification (Medicode, 1995). NDI provided information about vital status and cause of death for all US residents from state vital statistics offices. Suicide decedents were identified using International Classification of Diseases, 10th Revision, Clinical Modification Codes X60-X84 and Y87.0 (World Health Organization, 2004). Individuals who died by other causes were censored in analyses. For more information on the use of these sources in suicide research see prior reports (McCarthy et al., 2009; Bohnert et al., 2014). Sample: The population analyzed consisted of all users of VHA health care services in FY 2007–2008, (N ¼5,913,648). Measures: PTSD (vs. no. PTSD) was the primary variable of interest. Other measures that served as covariates in analyses included demographic variables, Veteran-specific variables, and other mental disorders. Demographic characteristics include age categorized into four groups (ages 18–29, 30–59, 60–74, and 75 þ), sex, married (vs. unmarried), urban (vs. rural), and region of the country in four categories (northeast, south, Midwest, west). Veteran characteristics included OEF/OIF Veteran (vs. non-OEF/ OIF Veteran), multiple OEF/OIF deployments (vs. non-multiple OEF/OIF deployments), and being new to VHA (vs. experienced users). Along with PTSD, mental disorders included a diagnosis (vs. no such diagnosis) of MDD, non-MDD depression, drug use disorder, alcohol use disorder, bipolar disorder, anxiety disorder (non-PTSD), and schizophrenia. Analyses: Descriptive analytic methods were used to examine characteristics of the total cohort and by PTSD status. The number and percentage of deaths by suicide of the total sample as well as those with and without a diagnosis of PTSD were also calculated. A series of logistic regressions, generalized linear models with the logit link (Tang et al., 2012), were used to estimate the association between PTSD and the risk of suicide, with analyses yielding Odds Ratios (ORs) and corresponding 95% confidence intervals (CIs). These analyses began with an unadjusted model to examine the bivariate association of PTSD and suicide (model 1) and built on this model by adding covariates in blocks including demographic variables (model 2), Veteran-specific variables (model 3), and other mental disorders (model 4). The potential problem of multi-colinearity was examined in models 2, 3 and 4 using the variance inflation factor (Hair et al., 2006). Logistic regression analyses were conducted using SAS 9.3 and these results are presented based on the template used in a recent study of tobacco use disorder and suicide (Bohnert et al., 2014). We ran further analyses using structural equation models (SEM) in order to explore the potential mediating influence of mental disorders on the association between PTSD and suicide (Kaplan, 2000; Xia et al., 2012). In these models the estimated parameters for the various associations or “pathways” are standardized regression coefficients for the probit model for binary responses. A statistically significant positive sign of an estimate
K.R. Conner et al. / Journal of Affective Disorders 166 (2014) 1–5
3
Table 1 Users of VHA health care services in FY 2007–2008 (N ¼5,913,648), total, and by posttraumatic stress disorder (PTSD) status. Characteristic
Total n
Total Age 18–29 Age 30–59 Age 60–74 Age 75þ Male Married Urban Northeast region South region Midwest region West region OEF/OIF Veteran Multiple OEF/OIF deployments New to VHA Posttraumatic stress disorder (PTSD) Major depressive disorder (MDD) Non-MDD depression Drug use disorder Alcohol use disorder Bipolar disorder Anxiety disorder Schizophrenia Absence of non-PTSD mental health disorders Number of suicide deaths
5,913,648 320,693 2,156,693 1,901,189 1,523,172 5,354,429 3,224,100 3,801,339 1,079,320 2,272,851 1,318,305 1,243,172 302,644 106,276 816,737 532,009 314,641 832,040 281,030 434,495 134,774 451,208 110,715 4,297,082 3620
PTSD diagnosis %
Yes
100.00 5.42 36.47 32.15 25.76 90.54 54.52 64.33 18.25 38.43 22.29 21.02 5.12 1.80 13.81 9.00 5.32 14.07 4.75 7.35 2.28 7.63 1.87 72.66 0.06
No
n
%
n
%
532,009 49,720 260,766 178,798 42,701 493,725 290,706 333,924 93,838 212,146 94,832 131,193 78,904 29,542 60,979 532,009 123,836 207,562 78,019 114,905 38,836 119,967 18,412 146,129 423
100.00 9.35 49.02 33.61 8.03 92.80 54.64 62.79 17.64 39.88 17.83 24.66 14.83 5.55 11.46 100.00 23.28 39.01 14.66 21.60 7.30 22.55 3.46 27.46 0.08
5,381,639 270,973 1,895,927 1,722,391 1,480,471 4,860,704 2,933,394 3,467,415 985,482 2,060,705 1,223,473 1,111,979 223,740 76,734 755,758 0 190,805 624,478 203,011 319,590 95,938 331,241 92,303 4,150,953 3197
100.00 5.04 35.23 32.01 27.51 90.32 54.51 64.48 18.31 38.29 22.73 20.66 4.16 1.43 14.04 0.00 3.55 11.60 3.77 5.94 1.78 6.16 1.72 77.13 0.06
Note:The percentages provided are column percentages. Tests of differences of proportions between groups with PTSD and those without PTSD were conducted for all rows, for example the proportion of men (vs. women) with a PTSD diagnosis vs. without a PTSD diagnosis, and in all cases are statistically significant (po 0.001).
indicates that the probability of an event (e.g., suicide) increases when the value of the predictor (e.g., PTSD) increases; a statistically significant negative sign indicates that the probability of an event decreases when the value of the predictor increases. In the current analyses the direct effect is the pathway from PTSD to suicide, while controlling for mental disorder(s). The indirect, or mediation effect, describes the pathway from PTSD to suicide through the mental disorder(s). The total effect is the sum of the direct and indirect effects of the predictor on suicide. Model fits were tested using the Chi-square test and Root Mean Square Error of Approximation (MacCallum et al., 1996). Mplus was used for all SEM analyses (Muthen and Muthen, 2006). 3. Results Descriptive data on the population of VHA users are provided in Table 1 including overall results and those stratified by PTSD diagnosis. Of the total population of nearly 6 million, about nine in ten were male (90.5%); about two-thirds were ages 30–59 (36.5%) or ages 60–74 (32.2%); about two-thirds lived in urban as opposed to rural areas (64.3%); and each of the four regions of the United States were well represented, with the south region being the largest (38.4%). A minority of individuals were new to VHA services (13.8%), were an OEF/OIF Veteran (5.1%), or had a history of multiple OEF/OIF deployments (1.8%). Nine percent of the population had been diagnosed with PTSD. Among those with a PTSD diagnosis, about one quarter (27.5%) had no other mental disorder. A total of 3620 (0.06%) VHA Veterans died by suicide including 423 who had been diagnosed with PTSD. Results of the multiple logistic regression analyses are provided in Table 2. In the unadjusted analysis (model 1), PTSD is associated with suicide at a statistically significant level, with odds ratio, OR (95% confidence interval, 95% CI) ¼1.34 (1.21, 1.48), p o0.001. This result suggests that, prior to adjustment for other variables, VHA
Veterans with a PTSD diagnosis are at about 34% (21–48%) increased risk for suicide compared to those without a PTSD diagnosis. Results are similar after adjustment for demographic variables (model 2) and further adjustment for Veteran-specific variables including OEF/OIF Veteran status and multiple OEF/OIF deployments (model 3). Finally, the result pertaining to PTSD after additional adjustment for other mental disorder diagnoses are dramatically different, OR (95% CI)¼ 0.77 (0.69, 0.86), p o0.001. The result suggests that, after adjustment for mental disorders, VHA Veterans with a PTSD diagnosis are at about 23% (31%, 14%) lower risk for suicide compared to Veterans without a PTSD diagnosis. There was no evidence of multi-colinearity among the explanatory variables across the four models. We ran further analyses in order to investigate the effect of adjustment for other mental disorders on the association between PTSD and suicide. First, we ran a series of multivariate logistic models in which we added, in turn, each of the mental disorders (MDD, etc.) to model 3. These results are provided in Table 3. Covarying MDD stands out for having the largest impact on the association between PTSD and suicide; after adjustment for MDD, PTSD shows no association with suicide, OR (95% CI)¼ 1.00 (0.90, 1.12), p¼ 0.977. To further explore the impact of MDD on the association between PTSD and suicide, we ran an unadjusted SEM model that shows a statistically significant association between PTSD and suicide (standardized regression coefficient¼0.043, po0.001), along with a second SEM model to examine the potential mediating role of MDD on the association between PTSD and suicide. The latter model includes paths from PTSD to suicide, PTSD to MDD, and MDD to suicide, and indicates no direct association between PTSD and suicide (standardized regression coefficient¼ 0.001, p¼0.769), that PTSD is associated with MDD (standardized regression coefficient¼0.294, po0.001), and MDD is associated with suicide (standardized regression coefficient¼0.152, po0.001). Furthermore, the results yield
4
K.R. Conner et al. / Journal of Affective Disorders 166 (2014) 1–5
Table 2 The association between PTSD and risk of death by suicide during FY 2007–2008 among users of VHA healthcare services. n
%
Model 1 OR
PTSD Yes No
532,009 5,381,639
9.00 91.00
Model 2 95% CI
1.340 [1.210, 1.483]
OR
Model 3 95% CI
1.298 [1.171, 1.439]
OR
Model 4 95% CI
1.312 [1.183, 1.456]
OR
95% CI
0.774 [0.693, 0.864]
1. Model 1 is unadjusted. 2. Model 2 is adjusted for demographic variables of age, sex, marital status, urban-rural residence, region of country. 3. Model 3 is adjusted for demographics and Veteran characteristics of OEF/OIF service era, multiple OEF/OIF deployments, and new to VHA. 4. Model 4 is adjusted for demographics, Veteran characteristics, and diagnostic variables of major depressive disorder (MDD), non-MDD depression, alcohol use disorder, drug use disorder, anxiety disorder, bipolar disorder, and schizophrenia.
Table 3 Association of PTSD with suicide after adjustment for model 3 variables and each mental disorder examined separately. Analytic model
OR (PTSD)
95% CI
p-value
Model Model Model Model Model Model Model Model
1.312 1.002 1.099 1.138 1.170 1.153 1.169 1.299
[1.183, 1.456] [0.898, 1.117] [0.987, 1.272] [1.023, 1.265] [1.053, 1.301] [1.037, 1.282] [1.053, 1.299] [1.171, 1.441]
o 0.001 0.977 0.085 0.017 0.004 0.009 0.004 o 0.001
3 3 þmajor depressive disorder (MDD) 3 þnon-MDD depression 3 þalcohol use disorder 3 þdrug use disorder 3 þother anxiety disorder 3 þbipolar disorder 3 þschizophrenia
Model 3 is adjusted for demographics (age, sex, marital status, urban–rural residence, region of country) and Veteran characteristics of OEF/OIF service era, multiple deployments, and new to VA.
Fig. 1. Structural equation model results of the potential mediating role of depressive disorders and substance use disorders on the association between PTSD and suicide in users of VHA health care services in FY 2007–2008, Notes: Path coefficients shown are standardized. Asterisk indicates po0.05. PTSD¼ posttraumatic stress disorder, MDD¼major depressive disorder, non-MDD depression¼depressive disorders other than MDD, drug¼drug use disorder, alcohol¼ alcohol use disorder.
a statistically significant total effect estimate (standardized regression coefficient¼0.043, po0.001) and a statistically significant indirect effect estimate (standardized regression coefficient¼ 0.045,
po0.001). Overall, the models suggest that association between PTSD and suicide is indirect and fully accounted for by MDD, consistent with mediation. The result for multivariate logistic regression model 4 indicates that, when multiple mental disorders are accounted for, PTSD shows an inverse association with suicide. Although an inverse association is not produced using SEM when MDD is considered alone (per above), various combinations of mental disorders that may be conceptualized as mediators of the association between PTSD and suicide produce such an effect. For example we ran a SEM model to explore the mediating effects of depressive disorders (MDD, non-MDD depression) and substance use disorders (AUD, DUD) on the association between PTSD and suicide. The results are shown in Fig. 1 and yield a statistically significant inverse association between PTSD and suicide (standardized regression coefficient¼ 0.116, po0.001), along with a statistically significant total effect (standardized regression coefficient¼0.043, po0.001) and a statistically significant and large indirect effect (standardized regression coefficient¼0.159, po0.001). Note that the indirect effect 0.159 is the difference between the total and direct effects and, because the direct effect is negative, in this instance the indirect effect is larger than the total effect. The results further show that PTSD is positively associated with each mental disorder at a statistically significant level and that each disorder in turn is positively associated with suicide. Overall, these results suggest that the inverse association between PTSD and suicide is attributable to the mediating influence of the other mental disorders, all of which are positively associated with both PTSD and suicide. In the aforementioned SEM models the model fit results indicate that the analyses fit the data well.
4. Discussion The analyses reported here expand upon and clarify previous findings for the total population of VHA users on the impact of PTSD on suicide (Ilgen et al., 2010) by adjusting for the effects of
K.R. Conner et al. / Journal of Affective Disorders 166 (2014) 1–5
other mental disorders. As hypothesized, we identified an association of PTSD and increased risk for suicide in the unadjusted model. Results were similar after adjustment for demographic variables and Veteran-specific variables including OEF/OIF service history. However, the association between PTSD and suicide is significantly affected by adjustment for other mental disorders. Indeed, the change in the association between PTSD and suicide is dramatic, with results suggesting that VHA users with a PTSD diagnosis are at lower risk for suicide than those without a PTSD diagnosis. Potential explanations were discussed earlier and include the fact that there are many common contributors to PTSD and other mental disorders, the difficulty of accurate differential diagnosis in complex patients, the ability to tolerate PTSD without developing other mental health or substance use disorders as complications may indicate less severe cases and/or increased resilience and stress tolerance, and that the identification of mental disorders comorbid with PTSD may lead to more intensive clinical care and monitoring, and that PTSD may lead to other mental disorders (e.g., MDD) which may in turn confer risk for suicide, a mediation scenario that we explored using SEM. There were limitations of the study. The data on mental disorder diagnoses were cross-sectional and, therefore, the mediation analyses are considered exploratory because the temporal ordering of PTSD and other diagnoses was not established. Mental disorder diagnoses were based on clinical records with associated limitations of reliability and validity. Some variables that may also influence the association between PTSD and suicidal behavior, for example social support (DeBeer et al., 2014), are not routinely available in clinical records. Nearly three quarters (72.5%) of VHA users with a PTSD diagnosis had one or more other mental disorders, creating a challenge to the development of policies or practices based on results that a PTSD diagnosis without comorbidity poses no added suicide risk (or is protective). We examined VHA users, with unclear relevance to other populations. Because a small percentage of VHA users are not Veterans, it was not strictly a Veteran sample. There were also many strengths of the study. VHA is the largest healthcare system in the United States. The analysis was of a whole population (VHA users), eliminating concerns about sampling biases within this population. The study was well-powered to examine the association of PTSD and suicide and confounding influences. The analyses examined the potential impact service during the wars in Iraq and Afghanistan on the results. In conclusion the results of the current report show that, among VHA users, clinical diagnoses of PTSD are associated with lower risk for suicide after accounting for other mental disorders. Further research to explain this phenomenon (mediation, etc.) is needed. The findings suggest the importance of identifying and treating comorbid mental disorders (e.g., MDD) that confer risk for suicide in VHA users diagnosed with PTSD in suicide prevention efforts. Role of funding source This research was supported by the Department of Veterans Affairs (VA) Office of Mental Health Services and the VA Office of Mental Health Operations. The views expressed in this report are those of the authors and do not necessarily represent those of the Department of Veterans Affairs.
Conflict of interest The authors have no conflicts of interest to report.
5
Acknowledgments This research was supported by the U.S. Department of Veterans Affairs (VA) Office of Mental Health Services and the VA Office of Mental Health Operations. The views expressed in this report are those of the authors and do not necessarily represent those of the Department of Veterans Affairs. The authors thank William Voss, Ph.D., for a literature review and contributions to an early draft.
References Bohnert, K.M., Ilgen, M.A., McCarthy, J.F., Ignacio, R.V., Blow, F.C., Katz, I.R., 2014. Tobacco use disorder and the risk of suicide mortality. Addiction 109, 155–162. Bullman, T.A., Kang, H.K., 1994. Posttraumatic stress disorder and the risk of traumatic deaths among Vietnam veterans. J. Nerv. Ment. Dis. 182, 604–610. Conner, K.R., Bohnert, A.S., McCarthy, J.F., Valenstein, M., Bossarte, R., Ignacio, R., Lu, N., Ilgen, M.A., 2013. Mental disorder comorbidity and suicide among 2.96 million men receiving care in the Veterans Health Administration health system. J. Abnorm. Psychol. 122, 256–263. DeBeer, B.B., Kimbrel, N.A., Meyer, E.C., Gulliver, S.B., Morissette, S.B., 2014. Combined PTSD and depressive symptoms interact with post-deployment social support to predict suicidal ideation in Operation Enduring Freedom and Operation Iraqi Freedom veterans. Psychiatry Res. 216, 357–362. Desai, R.A., Dausey, D.J., Rosenheck, R.A., 2005. Mental health service delivery and suicide risk: the role of individual patient and facility factors. Am. J. Psychiatry 162, 311–318. Gradus, J.L., Qin, P., Lincoln, A.K., Miller, M., Lawler, E., Sorensen, H.T., Lash, T.L., 2010. Posttraumatic stress disorder and completed suicide. Am. J. Epidemiol. 171, 721–727. Hair, J.F., Anderson, R., Tatham, R.L., Black, W.C., 2006. Multivariate Data Analysis, 6th ed. Prentice Hall, Union Saddle River, N.J. Hoyert, D.L., Xu, J., 2012. Deaths: preliminary data for 2011. National Vital Statistics Reports, Vol. 61 no. 6. Ilgen, M.A., Bohnert, A.S., Ignacio, R.V., McCarthy, J.F., Valenstein, M.M., Kim, H.M., Blow, F.C., 2010. Psychiatric diagnoses and risk of suicide in Veterans. Arch. Gen. Psychiatry 67, 1152–1158. Kaplan, D., 2000. Structural Equation Modeling: Foundations and Extensions, 10th ed. SAGE Publications, Inc., Thousand Oaks, CA. Krysinska, K., Lester, D., 2010. Post-traumatic stress disorder and suicide risk: a systematic review. Arch. Suicide Res. 14, 1–23. LeardMann, C.A., Powell, T.M., Smith, T.C., Bell, M.R., Smith, B., Boyko, E.J., Tomoko, I.H., Gackstetter, G.D., Ghamsary, M., Hoge, C.W., 2013. Risk factors associated with suicide in current and former US military personnel. J. Am. Med. Assoc. 310, 496–506. MacCallum, R.C., Browne, M.W., Sugawara, H.M., 1996. Power analysis and determination of sample size for covariance structure modeling. Psychol. Methods 1, 130–149. McCarthy, J., Valenstein, M., Kim, H.M., Ilgen, M., Zivin, K., Blow, F., 2009. Suicide mortality among patients receiving care in the Veterans Health Administration health system. Am. J. Epidemiol. 169, 1033–1038. Medicode., 1995. ICD-9-CM: International Classification of Diseases; 9th revision, clinical modification. Medicode, Salt Lake City, UT. Muthen, B.O., Muthen, L.K., 2006. M þ user's guide. Muthen and Muthen, Los Angeles, CA. Panagioti, M., Gooding, P.A., Tarrier, N., 2012. A meta-analysis of the association between posttraumatic stress disorder and suicidality: the role of comorbid depression. Compr Psychiatry 53, 915–930. Panagioti, M., Gooding, P., Tarrier, N., 2009. Post-traumatic stress disorder and suicidal behavior: a narrative review. Clin. Psychol. Rev. 29, 471–482. Tang, W., He, H., Tu, X.M., 2012. Applied Categorical and Count Data Analysis. Chapman & Hall/CRC, FL. Xia, Y., Lu, N., Zhang, H., Gunzler, D., Zubenko, G.S., Tu, X.M., 2012. Statistical Methods and Issues in the Study of Suicide. In: Lavigne, J., Kemp, J. (Eds.), Frontiers in Suicide Prevention and Research Hauppauge. Nova Science, NY, pp. 139–158. World Health Organization, 2004. International Statistical Classification of Diseases, 10th Revision (ICD-10) World Health Organization, Geneva, Switzerland. Zivin, K., Kim, H.M., McCarthy, J.F., Austin, K.L., Hoggart, K.J., Walters, H., Valenstein, M., 2007. Suicide mortality among individuals receiving treatment for depression in the Veterans Affairs health system: associations with patient and treatment setting characteristics. Am. J. Public Health 97, 2193–2198.
