Prior mental disorders and lifetime suicidal behaviors among U.S. Army soldiers in the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS)

Abstract

Here we report on associations of retrospectively-reported temporally prior mental disorders and Army career characteristics with subsequent first onset of suicidal behaviors in a large, representative sample of U.S. Army soldiers who participated in the Consolidated All Army Survey (AAS) of the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS; n=29,982). Results reveal that among men and women, all self-reported lifetime disorders measured (some assessed with screening scales) are associated with subsequent onset of suicide ideation. Among men, three disorders characterized by agitation and impulsiveness (intermittent explosive disorder, panic disorder, and substance disorders) predict the transition from suicide ideation to attempt. For both men and women, being in the Regular Army (vs. National Guard or Army Reserve) and being junior-rank enlisted (vs. being an officer) predict suicide attempts in the total sample. For men, a history of deployment and junior rank are predictors of suicide attempts after adjusting for pre-enlistment disorders but not accounting for pre- and post-enlistment disorders, suggesting that post-enlistment disorders account for some of the increased suicide risk among these career characteristics. Overall, these results highlight associations between mental disorders and suicidal behaviors, but underscore limitations predicting which people with ideation attempt suicide.

Over the past decade, there has been a marked increase in the rate of suicide in the US Army (Kuehn, 2009). A number of studies have examined potential risk factors for suicide among Army soldiers using electronic records maintained by the Army (Bachynski et al., 2012; Black, Gallaway, Bell, & Ritchie, 2011; Bush et al., 2013; Hyman, Ireland, Frost, & Cottrell, 2012; Logan, Skopp, Karch, Reger, & Gahm, 2012). In an effort to identify a broader array of potential risk and protective factors for suicide and suicidal behavior, the Army initiated a large-scale, multi-component epidemiological and neurobiological study known as the Army Study to Assess Risk and Resilience in Servicemembers, or Army STARRS (http://starrs-ls.org). One of the major components of Army STARRS is the All Army Survey (AAS), a large, representative survey of active duty Army soldiers. A previous paper using early replicates of the AAS reported on the prevalence of non-lethal suicidal behaviors (i.e., suicide ideation, plans, and attempts) among Army soldiers, and on demographic, Army career, and mental disorders as risk factors for these outcomes (Nock et al., 2014). However, those results were limited by the use of a relatively small sample size (n=5,428) that excluded deployed soldiers and Army National Guard and Army Reserve (G/R) troops. To address these limitations, we combined the original AAS with two other Army STARRS surveys that included deployed soldiers and those in the G/R, which we refer to as the Consolidated AAS (n=29,982).

In an initial paper describing this sample, we reported on the pre- and post-enlistment prevalence of suicidal behaviors as well as their associations with Army career characteristics (e.g. Regular Army versus G/R, rank, occupation) (Millner et al., in press). Here we expand on that work by examining the role of both pre- and post-enlistment mental disorders in predicting the subsequent onset of suicidal behaviors as of the time of the AAS survey. Consistent with the earlier report, we examine men and women separately due to the marked sex differences in non-lethal suicidal behavior rates (Nock et al., 2014; Ursano, Heeringa, et al., 2015; Ursano, Kessler, Stein et al., et al, 2015). We also examine which predictors of attempts also predict suicide ideation and the transition from suicide ideation to suicide attempts.

METHOD

Sample

Data are from the Consolidated All-Army Survey (AAS), a series of three Army STARRS surveys merged together. This report uses data from self-assessment questionnaires (SAQs) completed by active duty soldiers excluding servicemembers in Basic Training (BT), who, instead, were included in the New Soldier Survey, a separate component of Army STARRS (Ursano et al., 2014). The largest of the three surveys included in the Consolidated AAS was the All-Army Survey (AAS), a de-identified, cross-sectional survey administered quarterly in 2011–2012, and among G/R units in 2013. The AAS was completed by active duty soldiers, excluding soldiers in BT or deployed to a combat theatre, drawn from probability samples, stratified by Army Command-location, of units or sub-units selected with probabilities proportional to authorized unit strength exclusive of units with less than 30 soldiers, which make up less than 2% of Army personnel. Except for those with conflicting duty assignments (20.2%), all personnel in selected units attended an informed consent session describing study purpose, confidentiality, and voluntary participation. They then were requested to provide written informed consent for a group administered SAQ, to link SAQ responses to their administrative records, and to provide permission to be re-contacted for future data collections. Respondents were given a 90-minute session to complete the survey and most incomplete surveys were due to units arriving late or leaving the session early, although some respondents did not finish in the allotted time. Forms containing respondents’ identifying information was kept in a separate secure file. See Table 1 for sample size, percentages of those that gave consent, completed the survey, and provided linkage as well as response rate and completion-successful-linkage cooperation rate.

Table 1.

Percentage of soldiers that provided informed consent, completed the survey, and provided record linkage as well as response rate and completion-successful-linkage cooperation rates among the surveys making up the Consolidated All-Army Survey.

Survey Name	Sample characteristics	Attendees that provided consent (%)	Consenters that completed the survey (%)	Completers that provided linkage (%)	Sample size	Completion- successful- linkage cooperation rate (%)1	Response Rate (%)1
All-Army Survey (AAS)	Active duty (excluding those is basic training or deployed)	95.0	97.3	63.1	17,462	58.3	46.5
Supplemental AAS	soldiers deployed in Afghanistan	80.9	86.5	55.6	3,987	38.9	N/A2
Prospective pre-post deployment survey (PPDS)	just before soldiers deployed to Afghanistan	98.7	99.2	90.9	8,558	89.0	86.1

¹Based on the American Association of Public Opinion Research COOP1 and RR1 calculation methods (American Association for Public Opinion Research, 2009).

²Response rate could not be calculated because the denominator for this rate (the number of soldiers invited to the sessions) was not recorded.

The second of the three merged surveys making up the Consolidated AAS was the AAS administered to a special, supplemental sample of soldiers deployed in Afghanistan because the main AAS sample did not contain soldiers currently deployed to a combat zone. Constraints on our ability to collect data in Afghanistan caused us to administer surveys to soldiers who were in Kuwait, waiting to be processed for transit to and from mid-deployment leave. Aside from having to use an alternative location for survey administration, recruitment, consent, and data collection procedures were the same as those for the main AAS (Table 1).

In addition to data from the main AAS and Kuwait AAS samples, the Consolidated AAS also contains data from a prospective pre-post deployment survey (PPDS) of personnel in three Brigade Combat Teams. The PPDS was administered first just before soldiers deployed to Afghanistan and then again three times after returning from deployment. The Consolidated AAS only contains the baseline PPDS data, which is a valuable addition because soon-to-deploy units were underrepresented in the main AAS sample due to logistical complications. The recruitment, consent, and data collection procedures for the PPDS were identical to those in the main AAS (Table 1).

All procedures regarding recruitment, informed consent, and data protection for the surveys described above were approved by the Human Subjects Committees of the Uniformed Services University of the Health Sciences for the Henry M. Jackson Foundation (the primary grantee), the Institute for Social Research at the University of Michigan (the organization collecting the data), and all other collaborating organizations. SAQ responses from soldiers who provided consent for administrative data linkage were doubly-weighted before combining to adjust for inconsistencies between the sample and population. The first weight (W1) adjusted for differences in survey responses between the respondents who did and did not provide record linkage. The second weight (W2) adjusted for differences in multivariate administrative record profiles between weighted (W1) survey completers that provided record linkage and the target population. Specifically, this latter weight adjusted the sample to be representative of all active duty soldiers during the years 2011–2012 on the cross-classification of socio-demographics (age, sex, race-ethnicity, education, marital status), command (e.g., Forces Command, Training and Doctrine Command, Reserve Command [Army Reserve, Army National Guard], Component Commands), occupation (Combat Arms, Combat Support, Combat Service Support), rank (E1-E4, E5-E9, W1-W4, O1-O10), and deployment status-history (never-deployed, currently-deployed [the Kuwait supplemental sample], previously deployed). The doubly-weighted (W1xW2) data were combined to create the Consolidated All-Army Survey. A more detailed description of AAS weighting is presented elsewhere (Kessler, Heeringa, et al., 2013). Finally, all participants that had unknown survey dates (n = 25) were dropped from the final analytic sample.

Measures

Suicidal behaviors.

Suicidal behaviors were assessed using a modified version of the Columbia Suicidal Severity Rating Scale (C-SSRS; Posner et al., 2011) that assessed lifetime occurrence of suicide ideation (“Did you ever in your life have thoughts of killing yourself” or “Did you ever wish you were dead or would go to sleep and never wake up?”) and, among respondents who reported lifetime ideation, suicide plans (“Did you ever have any intention to act [on these thoughts/on that wish]?” and, if so, “Did you ever think about how you might kill yourself [e.g., taking pills, shooting yourself] or work out a plan of how to kill yourself?”) and attempts (“Did you ever make a suicide attempt [i.e., purposefully hurt yourself with at least some intention to die]?”). Age-of-onset (AOO) was also assessed for those that endorsed suicidal ideation, plans or attempts. In the current analysis, we focus on lifetime attempts as the primary outcome of interest and decompose this outcome into four intermediate outcomes: lifetime ideation, lifetime plans among ideators, lifetime attempts among ideators with a plan, and lifetime attempts among ideators without a plan.

Mental disorders.

Lifetime prevalence of eight common lifetime mental disorders was estimated in the surveys using diagnostic screening scales. We screened for five internalizing disorders (major depressive episode [MDE], bipolar disorder [BPD], panic disorder [PD], generalized anxiety disorder [GAD], post-traumatic stress disorder [PTSD]) and three externalizing disorders (attention-deficit/hyperactivity disorder [ADHD], intermittent explosive disorder [IED], and substance use disorders [SUD] including alcohol and drug abuse and dependence). The BPD assessment included BP-I, BP-II, and sub-threshold BPD (hypomania without history of major depression or sub-threshold hypomania) (Merikangas et al., 2011). To estimate prevalence of lifetime BPD, PD, and IED along with ADHD in the 6 months before SAQ completion, we assessed these disorders with screening scales from the Composite International Diagnostic Interview (CIDI) (Kessler & Üstün, 2004). To estimate the prevalence of the remaining disorders we used a revised self-report version of the Family History Screen (FHS) (Weissman et al., 2000), which was modified to assess personal rather than family history of these disorders. Respondents were asked to estimate AOO of each lifetime disorder. Recent ADHD was assumed to have been present since childhood based on prior validity studies (Kessler, Adler, et al., 2005; Kessler et al., 2007). The CIDI screening scales have good concordance with independent clinical diagnoses in the main AAS (area under the ROC curve of .69-.79 across diagnoses) (Kessler et al., 2013). The FHS has been shown to have acceptable concordance with best-estimate clinical diagnoses (Weissman et al., 2000), although the items used in the AAS yielded implausibly high prevalence estimates. Therefore, diagnoses based on the FHS should consequently be considered combinations of threshold and sub-threshold disorders. The implausibly high prevalence estimates also suggests that screening scales that have acceptable-to-good concordance with full diagnostic criteria may not provide accurate diagnoses or prevalence estimates and therefore rates of diagnoses assessed with screening scales should be interpreted in the context of this limitation. Unfortunately, it was not feasible to conduct full diagnostic interviews for all 29,982 participants.

Army career variables:

We examine four main Army career variables: component (Regular Army versus Reserve Component [i.e., activated G/R]), Military Occupational Specialty (MOS), rank (junior enlisted E1-E4, senior enlisted E5-E9, and officers [combining Warrant officers and Commissioned officers]), and deployment status-history (never deployed to a combat theatre, currently deployed, previously deployed). Consistent with previous research examining variations in soldier health across different occupations,(Gubata, Piccirillo, Packnett, & Cowan, 2013; Lindstrom et al., 2006; Niebuhr et al., 2011), we differentiated along three broad classes of occupations: combat arms occupations, which are involved directly in ground combat; combat support occupations, which provide operational assistance to combat arms; and all other occupations, which are referred to collectively as combat service support occupations (Kirin & Winkler, 1992; Layne, Naftel, Thie, & Kawata, 2001). For more details regarding MOS coding in STARRS, see Kessler et al., 2015.

Analysis Methods

Discrete-time survival analysis with person-year the unit of analysis, time-varying predictors, and a logistic link function (Efron, 1988) was used to examine associations of predictors with first lifetime onset of suicidal behavior. Both an additive model and a number of different non-additive models were examined to characterize the joint associations of the eight temporally primary mental disorders with the subsequent onset of a suicide attempt. Comparative model fit was evaluated with the Akaike Information Criterion (Burnham & Anderson, 2002). The final specification was then used to estimate parallel models decomposing the associations of the disorders with suicide attempts through the prior onset of ideation, of plans among ideators, of attempts among planners, and of attempts among ideators without a plan.

We then examined the associations of Army career characteristics with suicide attempts. This was done initially in a model that did not control for mental disorders. We then estimated two additional models that controlled either all the prior lifetime mental disorders considered here or only the disorders with pre-enlistment onsets. Comparisons of the coefficients associated with Army career characteristics across these models are used to assess the roles of selection into these characteristics based on pre-enlistment disorders and the effects of post-enlistment disorders.

Survival coefficients in all models were exponentiated to create odds ratios (ORs) with 95% confidence intervals. As the survey data are both clustered and weighted, the design-based Taylor series linearization method was used to produce standard errors (Wolter, 1985). Multivariate significance was examined using design-based Wald F tests based on 210 strata and 420 sampling-error calculation units in the data.

RESULTS

Lifetime prevalence of suicidal behaviors and mental disorders

As previously reported (Millner et al., in press), lifetime prevalence of attempted suicide in the total sample is 2.5% among men and 5.1% among women, with 57.3–64.3% (men-women) of these respondents reporting that their first suicide attempt occurred before they enlisted in the Army (Table 2). Decomposition shows that 41.8–46.1% of men-women with suicide ideation develop a plan, 33.9–38.3% of men-women with a plan, make an attempt, and 9.2–14.8% of men-women with suicide ideation but no plan make a suicide attempt. The majority of all five suicidal outcomes examined first occur prior to enlistment (55.1–70.7%). Both the prevalence and the proportion with pre-enlistment onsets are higher among women than men across all five outcomes.

Table 2.

Lifetime Prevalence of Suicidal Behaviors and Mental Disorders in the Consolidated AAS¹

	Prevalence				Pre-enlistment proportion			Pre-enlistment proportion
	Men1		Women		Men			Women
	%	SE	%	SE	%	SE	ndenominator	%	SE	ndenominator
I. Suicidal Behaviors2
Lifetime ideation among total sample	12.7	0.4	20.1	0.9	62.1	1.3	(3,344)	70.7	2.6	(575)
Lifetime attempt among total sample	2.5	0.2	5.1	0.5	57.3	3.6	(635)	64.3	6.1	(170)
Lifetime plan among ideators	41.8	1.6	46.1	3.2	57.0	2.0	(1,194)	69.3	3.9	(263)
Lifetime attempt among ideators with a plan	33.9	1.9	38.3	4.4	58.1	4.2	(416)	65.7	7.8	(119)
Lifetime attempt among ideators without a plan	9.2	0.9	14.8	2.4	55.1	5.7	(219)	61.4	10.6	(51)
Mental disorders
II. Internalizing Disorders
Major depressive episode (SC)	22.4	0.5	34.9	1.2	50.3	1.4	(5306)	54.1	2.8	(963)
Generalized anxiety disorder (SC)	28.8	0.6	40.4	1.5	39.8	1.5	(6431)	45.4	2.3	(1107)
Post-traumatic stress disorder (SC)	31.7	0.6	41.6	1.3	31.1	2.0	(7258)	42.4	2.0	(1192)
Bipolar disorder3	3.7	0.2	3.7	0.5	45.7	2.3	(1129)	39.7	6.5	(121)
Panic disorder	4.3	0.3	6.8	0.5	34.1	2.3	(1170)	50.8	4.3	(203)
III. Externalizing Disorders
Intermittent explosive disorder	19.8	0.4	14.8	1.0	78.7	1.3	(5456)	71.6	3.1	(479)
Attention-deficit/hyperactivity disorder	5.6	0.3	5.8	0.5	49.6	2.6	(1657)	49.5	4.7	(185)
Substance use disorders (SC)	14.4	0.5	11.0	0.9	44.3	1.8	(4503)	43.6	3.8	(351)

¹n=26,927.men; n=3,055 women. The disorders assessed with screening questions are designated by “(SC),” while the other disorders were assessed with the Composite International Diagnostic Interview (Kessler & Üstün, 2004). See the text for more detail.

²The pre- and post-enlistment proportions shown here differ slightly from those presented in a prior paper using these data (Millner et al., in press) because age of enlistment was amended in this paper to reflect administrative data when the age at survey did not coincide with the birth date in Army records. The resulting changes are minimal as they involved only: n=14 attempts (1.7% of total attempts), n=84 ideators (2.1% of total ideators), n=38 planners (2.6% of total planners), n=11 planned attempts (2.1% of total planned attempts), n=3 unplanned attempts (1.1% of total unplanned attempts).

³The assessment of bipolar disorder included bipolar disorder I, II, and sub-threshold BPD. See the text for more detail.

Lifetime prevalence estimates of mental disorders in the total sample range from highs of 31.7 (men) and 41.6% (women) for the PTSD screener to a low of 3.7% (identical for men and women) for broadly-defined DSM-IV/CIDI BPD (Table 2). Prevalence estimates are significantly higher among women than men for four outcomes (screens for MDE, GAD, PTSD, and DSM-IV/CIDI PD) and higher among men than women for two (the screen for substance disorder and DSM-IV/CIDI IED). A clear majority of both men and women with IED reported onsets prior to enlistment (78.7–71.6%), whereas small majorities reported pre-enlistment onsets of MDE (50.3–54.1%) and, among women, PD (50.8%). Substantial minorities (31.1–49.6%) reported pre-enlistment onsets of all other disorders.

Associations of temporally primary mental disorders with subsequent onset of suicidal behaviors

As noted above in the section on analysis methods, a number of multivariate models were estimated for the joint associations of temporally primary mental disorders with the subsequent first occurrence of suicide attempts over the lifetime, including both pre-enlistment and post-enlistment onsets (detailed results are available on request). The best-fitting specification was one that included main effects of each of the eight disorders considered here along with a single composite interaction term for number of comorbid disorders. The latter variable was coded zero for respondents with either none or exactly one disorder and between two and eight for respondents with between two and eight disorders. The OR for this comorbidity variable can be interpreted as the proportional difference in the odds of the outcome associated with the multivariate combination of comorbid disorders versus the odds predicted from the main effects.

ORs for individual disorders predicting suicide attempts in this total lifetime assessment are consistently elevated and significant for both men and women (Table 3). The highest OR for both men and women is associated with the screen for MDE (OR = 36.3 and 36.2, respectively), with the other disorder-specific ORs in the range 1.9–5.4 among men and 3.2–11.3 among women. The OR associated with the comorbidity variable, in comparison, is significantly less than 1.0 among both men and women (OR = 0.5–0.3), indicating the existence of sub-additive interactions among comorbid conditions in predicting attempted suicide. For example, given ORs of 36.3 and 4.3 for MDE and PTSD, respectively among men, a respondent with both of these disorders would have a predicted odds of subsequent attempted suicide under the model of 78.1 (i.e., 36.3 × 4.3 × 0.5) compared to a respondent with no lifetime disorder.

Table 3.

Associations of Prior Mental Disorders with Suicidal Behaviors in the Consolidated AAS¹

	In the total sample				Among ideators		Among ideators with a plan		Among ideators without a plan
	Lifetime ideators		Lifetime attempt		Lifetime plan		Lifetime attempt		Lifetime attempt
	OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI
Men
Internalizing disorders
MDE (SC)	26.9*	(21.2-34.2)	36.3*	(25.9-50.9)	1.5*	(1.0-2.1)	1.6	(0.9-2.9)	1.6	(0.7-3.3)
GAD (SC)	2.9*	(1.9-4.3)	2.3*	(1.3-3.9)	0.9	(0.6-1.5)	1.2	(0.5-2.5)	1.2	(0.4-3.2)
PTSD (SC)	4.3*	(3.4-5.6)	4.3*	(2.4-7.6)	1.4	(0.8-2.5)	1.6	(0.7-3.7)	1.5	(0.6-3.7)
Bipolar disorder2	2.9*	(2.0-4.4)	2.9*	(1.8-4.8)	1.2	(0.6-2.4)	1.0	(0.5-2.2)	2.4	(0.9-6.5)
Panic disorder	2.3*	(1.5-3.7)	3.1*	(1.7-5.8)	1.1	(0.6-1.9)	1.8	(0.8-4.1)	2.5*	(1.0-6.0)
Externalizing disorders
IED	5.6*	(4.4-7.0)	5.4*	(3.4-8.5)	0.9	(0.7-1.2)	2.0*	(1.0-3.8)	0.8	(0.3-1.9)
ADHD3	3.0*	(1.8-5.0)	1.9*	(1.1-3.0)	0.9	(0.5-1.4)	1.1	(0.5-2.4)	1.2	(0.5-3.2)
SUD (SC)	2.4*	(1.6-3.7)	3.7*	(2.2-6.4)	1.3	(0.9-1.8)	2.2*	(1.0-4.6)	1.2	(0.5-2.7)
F8,204	133.7*	<.001	58.6*	<.001	3.0*	0.003	1.2	0.31	1.5	0.16
Total number of mental disorders (values=0,0,2-8)	0.5*	(0.4-0.6)	0.5*	(0.4-0.7)	1.0	(0.7-1.4)	1.0	(0.6-1.6)	0.9	(0.5-1.8)
Women
Internalizing disorders
MDE (SC)	33.3*	(23.2-47.7)	36.2*	(21.7-60.3)	2.1	(1.0-4.4)	0.7	(0.2-2.2)	1.7	(0.5-5.6)
GAD (SC)	4.5*	(2.8-7.2)	3.4*	(1.2-9.9)	1.2	(0.5-2.6)	0.4	(0.1-1.9)	2.9	(0.5-15.4)
PTSD (SC)	7.2*	(4.4-11.7)	7.9*	(3.3-18.6)	0.5	(0.2-1.0)	1.0	(0.3-4.1)	1.0	(0.2-4.7)
Bipolar disorder2	2.7*	(1.2-6.2)	3.2*	(1.1-9.6)	1.4	(0.4-4.9)	1.2	(0.4-4.2)	1.9	(0.1-35.1)
Panic disorder	5.0*	(3.1-8.1)	7.7*	(2.8-21.4)	0.6	(0.3-1.3)	2.0	(0.4-9.5)	1.3	(0.1-17.8)
Externalizing disorders
IED	8.0*	(5.0-12.8)	11.3*	(3.5-36.2)	1.4	(0.70-2.71)	1.0	(0.3-3.9)	6.2	(0.9-42.5)
ADHD3	6.4*	(3.6-11.4)	3.9*	(1.4-11.4)	0.9	(0.38-1.97)	0.4	(0.1-2.4)	0.2	(0.0-4.0)
SUD (SC)	4.4*	(2.3-8.5)	11.3*	(4.4-29.4)	0.9	(0.35-2.42)	1.8	(0.5-6.0)	5.4	(0.5-61.5)
F8,204	42.5*	<.001	23.4*	<.001	2.8*	0.006	1.0	0.41	1.9	0.07
Total number of mental disorders (values=0,0,2-8)	0.3*	(0.2-0.4)	0.3*	(0.2-0.6)	1.3	(0.8-2.3)	1.6	(0.6-4.2)	0.6	(0.2-1.6)

¹Major depressive episode (MDE); Generalized anxiety disorder (GAD); Post-traumatic stress disorder (PTSD); Intermittent explosive disorder (IED); Substance abuse disorder (SUD). The disorders assessed with screening questions are designated by “(SC),” while the other disorders were assessed with the Composite International Diagnostic Interview (Kessler & Üstün, 2004). See the text for more detail. Models controlled for time-varying demographic characteristics (person-year interval, race, education, marital status) and time-varying army characteristics (pre-/post-enlistment, component (Regular Army vs Army Reserves/National Guard), military occupational specialty (MOS), rank, and deployment status). Race did not vary across person-years.

²The assessment of bipolar disorder included bipolar disorder I, II, and sub-threshold BPD. See the text for more detail.

³Attention-deficit/hyperactivity disorder (ADHD) was assessed at 6 months instead of at the lifetime level.

^*p < .05.

The decomposition models using the same specification as the model to predict attempted suicides show that the significant associations of mental disorders with subsequent suicide attempts are due largely to intervening associations with ideation. That is, temporally primary mental disorders predict the subsequent onset of suicide ideation in the total sample, with all eight disorders having significantly elevated ORs among both men (OR = 2.3–26.9) and women (OR = 2.7–33.3), but only one or two of the ORs are significant among men and none among women in predicting any of the later outcomes (i.e., plans among ideators, attempts among planners, attempts among ideators without a plan). The sub-additive interaction term for comorbid disorders is also significant in predicting ideation in the total sample (OR = 0.5–0.3) but not in any of the models for subsequent outcomes.

As noted above, the results in Table 3 include both suicide attempts that occurred in the years prior to enlistment and those that occurred in the years subsequent to enlistment. As shown in Table 2, 57.3% of men and 64.3% of women who made a suicide attempt engaged in their first attempt prior to enlistment. When the models are estimated separately in each of these two time periods, results are quite similar for attempts that occurred prior to enlistment as in the total sample. That is, ORs for specific disorders predicting attempted suicides are consistently elevated and statistically significant for both men and women, with the highest ORs associated with the screen for MDE, and the OR associated with comorbid disorders consistently negative and significant. Disaggregation through ideation and plans show that the significant predictors of attempts are largely mediated by ideation, with far fewer significant predictors of plans among ideators, attempts among planners, or attempts among ideators without a plan (detailed results are available on request).

Associations are more complex in predicting suicide attempts that occurred after enlistment due to the fact that we consider the possibility of different associations involving mental disorders with pre-enlistment and post-enlistment onsets. Among men, six of the eight pre-enlistment onset disorders (all other than GAD and ADHD) and five of the post-enlistment onset disorders (all other than GAD, panic disorder, and ADHD) have significantly elevated ORs predicting post-enlistment suicide attempts. The screen for MDE has the highest OR in both models (OR = 25.0–58.1), with the other significant ORs in the range 2.1–4.3. Only two of the eight pairs of pre- versus post-enlistment ORs are significantly different from each other, one involving a higher OR for the post-enlistment than pre-enlistment MDE screen (OR = 58.1 vs. 25.0, F_1,209=6.7, p=.01) and the other a higher OR for pre- than post-enlistment panic disorder (OR = 3.6 vs. 1.4; F_1,209=4.8, p=.029). The ORs for pre-enlistment and post-enlistment comorbidity are both negative and significant (OR = 0.6–0.8). Disaggregation again shows that the associations of temporally primary mental disorders with post-enlistment suicide attempts are mediated largely through ideation (detailed results are available on request).

The situation is different among women, where only three pre-enlistment disorders (the MDE and substance screens, and IED) and three mostly similar post-enlistment disorders (the MDE, PTSD, and substance screens) are significant predictors of post-enlistment suicide attempts. As with men, the ORs for the MDE screen are highest (OR = 12.5–29.5, with other significant ORs in the range 2.7–4.9). The ORs for pre-enlistment comorbidity and post-enlistment comorbidity, in comparison, are nonsignificant, with the exception of odds lower than 1.0 for pre-enlistment suicidal ideation. Disaggregation showed that, as with men, the associations of temporally primary mental disorders with post-enlistment suicide attempts are mediated largely through ideation (detailed results are available on request).

Associations of demographic variables with post-enlistment attempts

Differences in race and marital status emerged as the two demographic variables associated with post-enlistment suicide attempts. More specifically, odds of post-enlistment suicide attempt are significantly lower for black male soldiers (OR = 0.4) relative to white male soldiers (Table 4, Model 1). Compared with their currently married counterparts, previously-married men have higher odds (OR = 2.1), whereas never-married women have lower odds of post-enlistment suicide attempts (OR = 0.4).

Table 4.

Time-Varying Demographic and Career Predictors of Lifetime Suicide Attempt in the Consolidated AAS

	Men								Women
	Variable distribution		Multivariate 1		Multivariate 22		Multivariate 33		Variable distribution		Multivariate 1		Multivariate 22		Multivariate 33
	%	SE	OR	95% CI	OR	95% CI	OR	95% CI	%	SE	OR	95% CI	OR	95% CI	OR	95% CI
Demographics1
Race: Black	18.5	0.9	0.4*	(0.3-0.7)	0.4*	(0.3-0.8)	0.6*	(0.3-1.0)	34.6	2.0	-	-	-	-	-	-
Race: Hispanic	11.2	0.6	0.7	(0.4-1.3)	0.7	(0.4-1.4)	0.7	(0.4-1.4)	13.5	1.4	-	-	-	-	-	-
Race: Other	6.9	0.4	1.0	(0.6-1.9)	1.0	(0.5-2.0)	1.1	(0.6-2.0)	9.9	1.0	-	-	-	-	-	-
Race: White (reference)	63.4	1.3	-	-	-	-	-	-	42.0	2.0	2.1	(0.8-5.3)	2.0	(0.8-4.9)	1.7	(0.7-4.0)
F3,208			3.5*	0.017	3.0*	0.033	1.4	0.23
Previously married	4.9	0.2	2.1*	(1.0-4.3)	2.1*	(1.0-4.2)	1.4	(0.7-2.9)	9.1	0.8	0.9	(0.2-3.7)	0.9	(0.2-4.2)	0.9	(0.2-3.6)
Never married	36.4	0.7	0.7	(0.4-1.1)	0.7	(0.4-1.2)	0.8	(0.5-1.3)	50.5	1.5	0.4*	(0.2-0.8)	0.4*	(0.2-0.8)	0.4*	(0.2-1.0)
Currently married (reference)	58.7	0.6	-	-	-	-	-	-	40.4	1.3	-	-	-	-	-	-
F2,209			2.8	0.06	2.6	0.08	0.7	0.50			3.3*	0.038	2.9	0.06	1.8	0.16
Post-enlist., before 2000	83.1	0.6	0.6	(0.2-1.5)	0.6	(0.2-1.5)	0.5	(0.2-1.5)	86.6	0.8	5.1	(0.6-44.8)	4.7	(0.5-40.9)	4.4*	(0.5-41.4)
Post-enlist., after 2000 (reference)	16.9	0.6	-	-	-	-	-	-	13.4	0.8	-	-	-	-	-	-
Army characteristics
Regular army	83.6	0.9	3.4*	(1.9-6.3)	3.3*	(1.8-6.1)	2.6*	(1.3-5.0)	83.4	1.4	3.8*	(1.1-13.1)	3.3	(1.0-11.4)	2.4	(0.7-8.2)
Guard-reserve (reference)	16.4	0.7	-	-	-	-	-	-	16.6	1.4	-	-	-	-	-	-
MOS: combat arms4	38.0	1.5	0.8	(0.5-1.2)	0.8	(0.5-1.2)	0.8	(0.5-1.2)	5.3	0.6	1.2	(0.3-5.1)	1.4	(0.3-6.1)	0.8	(0.1-4.6)
MOS: combat support	26.3	1.1	0.7	(0.4-1.1)	0.6	(0.4-1.0)	0.7	(0.4-1.1)	22.0	1.6	1.3	(0.5-3.5)	1.3	(0.6-3.3)	1.4	(0.5-3.3)
MOS: combat service support (reference)	35.7	1.2	-	-	-	-	-	-	72.7	1.6	-	-	-	-	-	-
F2,209			2.0	0.14	2.1	0.12	1.6	0.21			0.1	0.88	0.2	0.80	0.3	0.77
Rank: junior	41.3	0.9	3.9*	(1.6-9.4)	3.6*	(1.5-8.7)	2.0	(0.8-5.0)	45.7	1.4	2.2	(0.8-6.0)	1.8	(0.7-5.1)	1.1	(0.4-3.4)
Rank: senior	38.2	0.8	1.7	(0.7-4.4)	1.6	(0.6-4.2)	1.1	(0.4-2.9)	32.1	1.5	2.0	(0.5-7.2)	1.5	(0.4-5.5)	1.0	(0.3-3.8)
Rank: officer (reference)	20.5	1.0	-	-	-	-	-	-	22.2	1.5	-	-	-	-	-	-
F2,209			9.2*	<.001	7.4*	0.001	4.0*	0.020			1.1	0.34	0.6	0.54	0.1	0.95
Currently deployed5	34.0	0.4	1.9*	(1.3-2.9)	1.9*	(1.3-2.9)	1.3	(0.9-2.0)	25.8	0.6	0.6	(0.2-1.8)	0.6	(0.2-1.6)	0.4	(0.2-1.3)
Previously deployed	23.9	0.4	2.4*	(1.5-3.9)	2.3*	(1.4-3.8)	1.2	(0.7-2.1)	19.9	0.7	1.4	(0.5-4.3)	1.4	(0.5-4.0)	0.9	(0.3-2.5)
Never deployed (reference)	42.2	0.5	-	-	-	-	-	-	54.3	1.0	-	-	-	-	-	-
F2,209			6.3*	0.002	6.0*	0.003	0.8	0.44			1.8	0.17	1.7	0.2	1.2	0.31

¹Race did not change over person-years.

²Multivariate model 2 included the same predictors as multivariate model 1 but controlled for the 8 mental disorders only if age-of-onset was before year of Army enlistment.

³Multivariate model 3 included the same predictors as multivariate models 1 & 2 but controlled for 8 mental disorders, whether or not age-of-onset was before year of Army enlistment.

⁴Distributions reflect the frequencies for each variable category among the sample of all post-enlistment person years.

⁵Distributions reflect the frequencies for each variable category among the sample of post-enlistment person years after 2000 for men (83.1% of total) and for women (86.6% of total).

^*p <.05.

All three associations remain significant after accounting for pre-enlistment mental disorders (Table 4, Model 2) and two of the three associations are still significant after accounting for both pre-enlistment and post-enlistment mental disorders (Table 4, Model 3). Specifically, black male soldiers continue to have significantly lower odds of post-enlistment suicide attempt after controlling for pre-enlistment mental disorders (OR = 0.4) and after controlling for mental disorders with either a pre-enlistment or post-enlistment onset (OR = 0.6). Among men, the higher OR associated with being a previously married remains significant after controlling for pre-enlistment disorders (OR = 2.1), but is no longer significant after taking into account pre-enlistment and post-enlistment disorders (OR = 1.4). Odds of post-enlistment suicide attempts among never-married women remains significantly lower compared with currently-married women after introducing controls for pre-enlistment disorders (OR = 0.4) and controls for both pre-enlistment and post-enlistment disorders (OR = 0.4).

Associations of Army career variables with post-enlistment attempts

Consistent with previous STARRS reports (Millner et al., in press), odds of post-enlistment suicide attempts in models that do not control for mental disorders are significantly higher for Regular Army than activated National Guard or Army Reserve soldiers among both men and women (OR = 3.4–3.8) and unrelated to the broad MOS categories considered here (Table 4, Model 1). Among men, junior enlisted soldiers have elevated odds of suicide attempt compared to officers (OR = 3.9) and currently-deployed soldiers have elevated odds of first suicide attempts while deployed (OR = 1.9) compared to never-deployed men. Previously-deployed soldiers have elevated odds of ever making a first suicide attempt after returning from deployment compared to never-deployed men (OR = 2.4). Among women, MOS, rank, and deployment status are unrelated to odds of suicide attempt.

Some of these significant associations change meaningfully when controls are introduced for pre-enlistment mental disorders (Table 4, Model 2) or for both pre-enlistment and post-enlistment mental disorders (Table 4, Model 3). More specifically, the elevated OR associated with being in the Regular Army rather than the G/R (OR = 3.4) remains significant among men with controls for pre-enlistment mental disorders (OR = 3.3), and is attenuated but also remains significant after controlling for both pre-enlistment and post-enlistment disorders (OR = 2.6). The significant OR for Regular Army versus Guard-Reserve among women in the model without controls (OR = 3.8) becomes non-significant in the models with controls for pre-enlistment (OR = 3.3) and both pre-enlistment and post-enlistment (OR = 2.4) disorders. Among men, the significant ORs for junior rank and current-prior deployment also become non-significant in the model that controls for both pre-enlistment and post-enlistment disorders (OR = 2.0, 1.3–1.2), but remain significant in the model that controls only for pre-enlistment disorders (OR = 3.6, 1.9–2.3).

DISCUSSION

This study has five noteworthy limitations. First, the AAS response rate and linkage rate with administrative data were both relatively low. If relationships among the variables in the current study are markedly different among the omitted soldiers, compared with the soldiers in the study, then the findings may not generalize across all soldiers. In order to improve the generalizability, we used post-stratification to adjust for differences between the sample and the population on administrative variables. Second, some respondents might have not reported current or past suicidal thoughts or behaviors due to stigma (Zinzow et al., 2013), fear of loss of confidentiality (Hamaguchi, 2014), or other reasons. If this non-reporting was systematically related to the predictors considered here (e.g., anxious suicidal soldiers fail to disclose suicidal behaviors more than non-anxious suicidal soldiers) it would introduce bias in tests of associations between predictors and outcomes. A similar bias would be introduced by systematic under-reporting of mental disorders. Third, participants may have been biased in their retrospective recall of AOO of suicidal behaviors, which would introduce imprecision and perhaps bias into estimates of associations. Fourth, we examined only a limited set of mental disorders and used screening questions rather than full diagnostic assessments (Kessler et al., 2013). Although, screening assessments have acceptable-to-good concordance with full diagnostic scales, they may have provided inaccurate diagnoses and prevalence estimates. Prevalence estimates for diagnoses measured with the screener with lower concordance rates (i.e. FHS) are doubtlessly inflated. Fifth, as this study focused only on active duty soldiers, we do not know if the patterns documented here continue to hold after separation.

Within the context of these limitations, the study has seven notable findings. First, female soldiers generally have higher lifetime prevalence of internalizing disorders and male soldiers of externalizing disorders, with the exceptions of BPD and ADHD, which have similar prevalence across the sexes. This finding is consistent with much previous evidence (Kessler, Berglund, et al., 2005). Additionally, substantial percentages of soldiers experience the onset of suicidal ideation, plans and attempts as well as mental disorders, both before and after enlistment. Although, prior reports have highlighted the noteworthy rates of pre-enlistment suicidal behaviors (Millner et al., in press; Nock et al., 2014), it is also significant to note that nearly one-third to almost one-half of suicidal thoughts and behaviors have their onset post-enlistment. Furthermore, onsets of most mental disorders occur for more soldiers during post-enlistment, compared with pre-enlistment. These results suggest that continued research into post-enlistment risk factors is important for understanding and preventing suicidal behaviors and mental disorders in the Army.

Second, although all of the lifetime mental disorders considered here are related to subsequent suicide attempts, these associations are largely mediated by suicide ideation, with only three disorders (IED, panic disorder, and substance use disorder) predicting the transition from ideation to attempt among men and none among women. These results are consistent with prior work in both civilian and Army samples that consistently found that many mental disorders have significant gross associations with subsequent incident suicide attempts, with depression among the most important of these disorders, but that these associations were largely mediated by suicide ideation (Nock et al., 2009; Nock, Hwang, Sampson, & Kessler, 2010).

Third, we examined the associations of pre-enlistment onset disorders with post-enlistment suicidal behaviors based on the practical question of whether early detection and treatment of pre-enlistment disorders after enlistment might be a useful approach to preventing subsequent onset of suicidal behaviors. The alternative is that only disorders with post-enlistment onsets are associated with onset of suicidal behaviors during the years of enlistment. We found that a number of pre-enlistment disorders (MDE, PTSD, PD, IED, and substance use disorder) significantly predicted post-enlistment suicide attempts among ideators. Although no pre-enlistment disorders predicted this transition among women, the small number of women included in that model reduced statistical power. These results suggest that it might be useful to target this subset of pre-enlistment disorders via screening for treatment services among new male soldiers.

Fourth, the only demographic factors associated with post-enlistment suicide attempts were race and marital status. Among men, black soldiers were at reduced risk of post-enlistment suicide attempts, compared with white soldiers. Consistent with this effect of race, black servicemembers consistently show reduced risk of both suicide attempts and suicide death across studies in the Army (Black et al., 2011; Ursano, Heeringa, et al., 2015; Ursano, Kessler, Heeringa, et al., 2015; Ursano, Kessler, Stein, et al., 2015) and in the overall U.S. military (Bachynski et al., 2012). Also among men, compared with married soldiers, previously married soldiers were at higher risk of post-enlistment attempts. The effect of marriage was different for women, with never-married women showing lower odds of a suicide attempt, compared with currently married women. The different associations of marital status with suicide attempt for men and women are new findings. Reports from other Army STARRS samples looking across all, predominantly male Army soldiers, found that being unmarried is generally associated with increased odds of suicide attempts (Ursano, Heeringa, et al., 2015; Ursano, Kessler, Heeringa, et al., 2015). The fact that the increased risk among previously-married male soldiers remains after controlling for pre-enlistment disorders but not pre- and post-enlistment disorders suggests that some combination of post-enlistment Army experiences, onset of mental disorders, and divorce are associated with an incident suicide attempt. The reason for the lower odds of suicide attempts among never married female soldiers is unclear and requires for further study.

Fifth, among Army career characteristics, we found that, relative to those in the G/R, soldiers in the Regular Army are at increased risk of first suicide attempts while on active duty and that this is true for both men and women. For men, increased risk of incident suicide attempts among those in the Regular Army compared to the G/R is present even after controlling for lifetime mental disorders with both pre-enlistment and post-enlistment onsets, although the OR becomes less elevated with these controls. For women, in comparison, the increased risk of incident suicide attempt among those in the Regular Army compared to G/R is explained statistically by controls for post-enlistment onset disorders, although the magnitudes of the ORs with and without controls are quite comparable for women and men. These results further clarify findings from a prior AAS report showing increased risk among Regular Army troops (Millner et al., in press). However, it differs from results from a study among new soldiers in BT that found lower prevalence of suicide ideation and attempts among soldiers in the Regular Army compared with G/R (Ursano, Heeringa, et al., 2015). There are many differences between activated soldiers of all ages and new recruits that could account for these divergent findings. For example, new G/R recruits with a history of suicidal behaviors may drop out in higher numbers than their counterparts in the Regular Army, leading to comparatively lower prevalence of suicidal behaviors in the G/R among activated soldiers. Alternatively, G/R soldiers with a history of suicidal behaviors or mental health issues may still be in the G/R but not activated and therefore omitted from the AAS sample. Future longitudinal studies that follow new recruits from BT through activation and deployment could identify component differences that moderate risk and help to clarify these inconsistent results.

Sixth, for men having junior enlisted rank, compared to being an officer, is significantly associated with post-enlistment suicide attempts, consistent with prior studies (Ursano, Kessler, Heeringa, et al., 2015). This relationship remains elevated when controlling for pre-enlistment but is attenuated and non-significant after controlling for both pre-enlistment and post-enlistment mental disorders. This suggests that pre-enlistment disorders cannot account for the association of junior rank with post-enlistment suicide attempts and the raises the possibility that, instead, vulnerabilities associated with experiences that occurred after enlistment might be involved in the elevated odds of suicide attempts among junior enlistment soldiers. Interestingly, senior enlisted soldiers do not have elevated odds of post-enlistment attempts, either with or without controls for mental disorders. This difference between junior and senior enlisted soldiers suggests either that the mental disorders that occur to enlisted soldiers as of promotion to the senior enlisted ranks become less important in suicidality subsequent to these promotions or that selection factors involving the associations of these disorders with suicidality lead to selection out of promotion. This general pattern of association between rank and suicidality is consistent with prior studies (Hoge, Auchterlonie, & Milliken, 2006), but we are unaware of any attempt to unpack the association to determine why it exists.

Seventh and finally, being currently or previously deployed is associated with elevated risk of subsequent first post-enlistment suicide attempts relative to the never-deployed, but these associations attenuate with controls for lifetime disorders suggesting that Army deployment experiences may have contributed to increased risk of attempts. Inconsistent with this result, a prior Army STARRS study found that, among enlisted soldiers, those that were never deployed had the highest odds of an attempt Ursano, Kessler, Stein et al., et al, 2015), although it is important to note that this other study included new soldiers in BT, while BT soldiers were explicitly omitted from the AAS. This is an important distinction given that this other study found that nonfatal suicide attempts are highly prevalent during BT. Yet another Army STARRS study found that soldiers in combat positions showed lower risk of suicide when deployed but soldiers in other positions showed increased risk (Kessler et al., 2015). This is a complex area of investigation due to the existence of a healthy warrior effect that selects soldiers with mental disorders out of deployment (Wilson et al., 2009) and due to the fact that deployment consists of many different experiences, some of which may increase and others decrease risk of subsequent suicidal behaviors. More fine-grained investigations are needed to clarify these inconsistent results.

In summary, this study provides new information about the prevalence of lifetime mental disorders in the Army as well as about the role of mental disorders, demographic characteristics, and Army career characteristics in the prediction of suicidal behaviors that begin during the years of Army service. Future studies using the Consolidated AAS will further examine potential risk and protective factors for suicidal behavior among Army soldiers and will focus on identifying risk factors for the transition from suicide ideation to suicide attempt. Taken together, these studies aim to improve the understanding, prediction, and prevention of suicidal behavior among Army soldiers.

Appendix

The STARRS-LS Collaborators: The Army STARRS Team consists of Co-Principal Investigators: Robert J. Ursano, MD (Uniformed Services University of the Health Sciences) and Murray B. Stein, MD, MPH (University of California San Diego and VA San Diego Healthcare System). Site Principal Investigators: Steven Heeringa, PhD (University of Michigan) and Ronald C. Kessler, PhD (Harvard Medical School). National Institute of Mental Health (NIMH) collaborating scientists: Lisa J. Colpe, PhD, MPH and Michael Schoenbaum, PhD. Army liaisons/consultants: COL Steven Cersovsky, MD, MPH (USAPHC (Provisional)) and Kenneth Cox, MD, MPH (USAPHC (Provisional)). Other team members: Pablo A. Aliaga, MA (Uniformed Services University of the Health Sciences); COL David M. Benedek, MD (Uniformed Services University of the Health Sciences); K. Nikki Benevides, MA (Uniformed Services University of the Health Sciences); Paul D. Bliese, PhD (University of South Carolina); Susan Borja, PhD (NIMH); Evelyn J. Bromet, PhD (Stony Brook University School of Medicine); Gregory G. Brown, PhD (University of California San Diego); Laura Campbell-Sills, PhD (University of California San Diego); Catherine L. Dempsey, PhD, MPH (Uniformed Services University of the Health Sciences); Carol S. Fullerton, PhD (Uniformed Services University of the Health Sciences); Nancy Gebler, MA (University of Michigan); Robert K. Gifford, PhD (Uniformed Services University of the Health Sciences); Stephen E. Gilman, ScD (Harvard School of Public Health); Marjan G. Holloway, PhD (Uniformed Services University of the Health Sciences); Paul E. Hurwitz, MPH (Uniformed Services University of the Health Sciences); Sonia Jain, PhD (University of California San Diego); Tzu-Cheg Kao, PhD (Uniformed Services University of the Health Sciences); Karestan C. Koenen, PhD (Columbia University); Lisa Lewandowski- Romps, PhD (University of Michigan); Holly Herberman Mash, PhD (Uniformed Services University of the Health Sciences); James E. McCarroll, PhD, MPH (Uniformed Services University of the Health Sciences); James A. Naifeh, PhD (Uniformed Services University of the Health Sciences); Tsz Hin Hinz Ng, MPH (Uniformed Services University of the Health Sciences); Matthew K. Nock, PhD (Harvard University); Rema Raman, PhD (University of California San Diego); Holly J. Ramsawh, PhD (Uniformed Services University of the Health Sciences); Anthony Joseph Rosellini, PhD (Harvard Medical School); Nancy A. Sampson, BA (Harvard Medical School); CDR Patcho Santiago, MD, MPH (Uniformed Services University of the Health Sciences); Michaelle Scanlon, MBA (NIMH); Jordan W. Smoller, MD, ScD (Harvard Medical School); Amy Street, PhD (Boston University School of Medicine); Michael L. Thomas, PhD (University of California San Diego); Leming Wang, MS (Uniformed Services University of the Health Sciences); Christina L. Wassel, PhD (University of Vermont); Simon Wessely, FMedSci (King’s College London); Christina L. Wryter, BA (Uniformed Services University of the Health Sciences); Hongyan Wu, MPH (Uniformed Services University of the Health Sciences); LTC Gary H. Wynn, MD (Uniformed Services University of the Health Sciences); and Alan M. Zaslavsky, PhD (Harvard Medical School).

Additional Information: A complete list of Army STARRS publications can be found at http://www.STARRS-LS.org.