Association of Premilitary Mental Health With Suicide Attempts During US Army Service

Key Points Question Is premilitary mental health associated with suicide attempts among soldiers who do not receive a mental health diagnosis before their suicide attempt? Findings This cohort study of 21 772 Regular Army enlisted soldiers found that suicide attempt risk peaked at the end of the first year of service for those who did and did not receive a mental health diagnosis. Premilitary mental health risk factors differed in the 2 groups. Meaning The findings of this study suggest that time course and preenlistment mental health status may aid in identifying suicide attempt risk in soldiers with and without a mental health diagnosis.

Traumatic stress 308.X found good concordance between CIDI-SC and modified PCL diagnoses and independent clinical diagnoses based on blinded Structured Clinical Interviews for DSM-IV (SCID). 10 An indicator variable was created to identify soldiers who met criteria for any of the above disorders.

Pre-enlistment lifetime self-injurious thoughts and behaviors (SITBs). Pre-enlistment history of SITBs
was assessed with a modified version of the Columbia-Suicide Severity Rating Scale, 11 including lifetime SI ("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?"), lifetime SA ("Did you ever make a suicide attempt; that is, purposefully hurt yourself with at least some intention to die?"), and lifetime non-suicidal self-injury ("Did you ever do something to hurt yourself on purpose, but without wanting to die [e.g., cutting yourself, hitting yourself, or burning yourself]?").

Analysis methods
Analyses were conducted using SAS version 9.4. 12 Data were analyzed using discrete-time survival analysis with person-month the unit of analysis and a logistic link function. 2,3 In order to control for changes in SA risk across time in service, we began by estimating risk (suicide attempters per 100,000 person-months) during each month of service. Splines (piecewise linear functions) were calculated based on the monthly risk estimates to identify nonlinearities in how risk changed over the course of the first 48 months of service.
Specifically, after fitting a linear function to the monthly risk estimates, we used chi-square difference tests, deviance, and the Akaike Information Criterion to assess whether knots (changes in slope) and additional linear segments improved model fit. Next, a series of logistic regression models examined self-reported survey variables at baseline (lifetime mental disorder, lifetime SI, lifetime SA, lifetime NSSI) as predictors of first documented SA during the first four years of Army service. Each predictor was examined separately in a model that adjusted for time in service (using the spline variables), socio-demographic variables, service-related variables, and administrative MH-Dx. In each of those multivariable models, we tested the two-way interaction between the baseline survey predictor and administrative MH-Dx. The sample was stratified based on previous MH-Dx and each baseline survey predictor was examined separately as a multivariable predictor among personmonths with and without a previous MH-Dx.
Logistic regression coefficients and their confidence limits were exponentiated to obtain estimated odds ratios (OR) and 95% confidence intervals (95% CI). Standard errors were estimated using the Taylor series method 13 to adjust for the weighting and clustering of the NSS data. Multivariate significance tests in the logistic regression analyses were made using Wald  2 tests based on coefficient variance-covariance matrices that were adjusted for design effects using the Taylor series method. Statistical significance was evaluated using two-sided design-based tests and the .05 level of significance.

Sample characteristics
In the total cohort, weighted person-months were mostly male (87.6%), White

Prevalence, risk, and incidence of MH-Dx
In the total cohort, 20.8% of person-months were associated with a history of MH-Dx (Table 1) (Figure 1a). The cumulative incidence of first MH-Dx in the total cohort was 15.5% after 12 months of service and 50.7% after 48 months (Figure 1b). Of those who attempted suicide with no MH-Dx, 42.0% reported a pre-enlistment mental disorder at baseline.

SA risk by time in service
We used a discrete-time survival model and linear splines to examine risk as a function of time in service ( Figure 2). We first estimated risk of SA in each month of service. Spline models identified two nonlinearities (knots), one at the fourth month of service and one at the nineth month. Specifically, SA risk decreased from the start of service until the fourth month, then increased until reaching a peak in the ninth month of service. Following this peak, risk steadily decreased through the 48 th month of service. Two-way interactions between administrative MH-Dx and each component of the spline model were nonsignificant, indicating that risk by time in service did not differ for those with and without a MH-Dx. It was also nonsignificant when we examined the interaction between diagnosis and a simplified time variable that dichotomized time in service as ≤ 12 months vs. > 12 months.

Administrative predictors of SA
In separate models that adjusted only for time (represented by the spline variables), significant socio-  a The survey respondents considered here were Regular Army enlisted soldiers (n = 21,772). Survey-linked administrative personmonth records were examined through 48 months of service. The number of available person-month records for a given soldier varied because of attrition from service. b < High School includes: General Educational Development credential (GED), home study diploma, occupational program certificate, correspondence school diploma, high school certificate of attendance, adult education diploma, and other non-traditional high school credentials. NSSI = nonsuicidal self-injury eFigure. Monthly risk and cumulative incidence of attrition in a cohort of Regular Army enlisted soldiers during their first four years of service.  The survey respondents considered here were Regular Army enlisted soldiers (n = 21,772). Survey-linked administrative personmonth records were examined through 48 months of service. The number of available person-month records for a given soldier varied because of attrition from service. b Each predictor was examined separately in a logistic regression model that adjusted only for time in service (spline variables). c Multivariable 1: Logistic regression model that included time in service (spline variables), socio-demographic variables (gender, race/ethnicity, education, marital status), and administratively documented mental health diagnosis. d Multivariable 2: Logistic regression model that included all the variables in the Multivariable 1 model, plus service-related variables (rank, deployment status). b < High School includes: General Educational Development credential (GED), home study diploma, occupational program certificate, correspondence school diploma, high school certificate of attendance, adult education diploma, and other non-traditional high school credentials. *p < 0.05 eTable 5. Univariable and multivariable associations of self-report survey variables with documented suicide attempts among a cohort of Regular Army enlisted soldiers over their first four years of service. a 9.3* (p=0.003) 3.4 (p=0.067) a The survey respondents considered here were Regular Army enlisted soldiers (n = 21,772). Survey-linked administrative personmonth records were examined through 48 months of service. The number of available person-month records for a given soldier varied because of attrition from service. b Each predictor was examined separately in a logistic regression model that adjusted only for time in service (spline variables). c Multivariable 1: Each predictor was examined separately in a logistic regression model that adjusted for time in service (spline variables), socio-demographic variables (gender, race/ethnicity, education, marital status), service-related variables (rank, deployment status), and administratively documented mental health diagnosis. d Multivariable 2: All predictors were examined together in a logistic regression that adjusted for time in service (spline variables), socio-demographic variables (gender, race/ethnicity, education, marital status), service-related variables (rank, deployment status), and administratively documented mental health diagnosis. *p < 0.05