Secondary Logo

Journal Logo

Original Articles

Sex Differences in US Army Suicide Attempts During the Wars in Iraq and Afghanistan

Naifeh, James A. PhD*,†; Mash, Holly B. Herberman PhD*,†; Stein, Murray B. MD, MPH‡,§; Vance, Mary C. MD, MSc*,†; Aliaga, Pablo A. MS*,†; Fullerton, Carol S. PhD*; Dinh, Hieu M. BS*,†; Wynn, Gary H. MD*; Kao, Tzu-Cheg PhD; Sampson, Nancy A. BA; Kessler, Ronald C. PhD; Ursano, Robert J. MD*

Author Information
doi: 10.1097/MLR.0000000000001425

Abstract

Across diverse populations, women tend to have higher risk of making a nonfatal suicide attempt (SA) than men, and men tend to have higher risk of dying by suicide than women.1,2 The same sex differences are present in the US Army,3,4 where rates of suicidal behavior increased substantially during the wars in Iraq and Afghanistan5–7 and remain elevated.8,9 Representative survey data collected from 2011 to 2013 indicate that, among active duty soldiers, the lifetime prevalence of self-reported SA is 5.1% for women and 2.5% for men.10 Similarly, a study of administrative records from 2004 to 2009 found that the odds of medically documented SA among Regular Army enlisted soldiers are nearly 2.5 times higher for women versus men, and the risk for women remains higher regardless of deployment status; that is, among those never deployed, currently deployed (in-theater), and previously deployed.11 However, given that the majority of soldiers are men, with women comprising ∼15% of the US Army in 2018,12 female soldiers have been an understudied population in suicidal behaviors research. Little is known about the extent to which risk factors for SA may differ between women and men on active duty. Addressing this gap is necessary for the development and delivery of interventions tailored to the needs of female soldiers, whose experiences and stressors are both similar to, and different from, those of male soldiers.13

Although representative survey research has examined predictors of self-reported SAs separately among women and men in the US Air Force14 and US Army,10,15 less is known about sex-specific risk factors for medically documented SAs among active duty soldiers. Given that individuals may or may not have received medical care following self-reported SA, documented SAs have particular clinical importance as events diagnosed and treated within the military health care system. In the Regular Army, the vast majority of soldiers who make a documented SA are enlisted personnel (99%), with officers comprising only 1%.3 Documented SAs in the Regular Army enlisted population are more likely among soldiers who are younger, non-Hispanic White, less educated, in their first term of enlistment (particularly the first 2 y of service), either never deployed or previously deployed (vs. currently deployed), and those who previously received a mental health diagnosis (MHDx).3 However, these findings are based on an enlisted population predominantly composed of men. It is not known if the same risk factors are associated with documented SAs among women.

The current study used administrative data from the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS)16 to examine associations of sociodemographic, service-related, and mental health predictors with medically documented SAs among female Regular Army enlisted soldiers. For comparison, we examined the same predictors among male soldiers and the extent to which associations of those predictors with SA differed by sex. In addition, we examined sex differences in SA risk patterns as a function of time in service.

METHODS

Sample

This longitudinal, retrospective cohort study used data from the Army STARRS Historical Administrative Data Study (HADS), which integrates 38 Army and Department of Defense (DoD) administrative data systems, including every system that documents suicidal events. The HADS includes individual-level person-month records for all soldiers who were on active duty at some point between January 1, 2004 and December 31, 2009 (n=1.66 million soldiers).17 For soldiers who were on active duty during this period but began service before 2004, records date back to January 1, 2000. Person-month records were created by coding each month of a soldier’s career separately for each administrative variable and allowing values to change over time for an individual soldier.18,19 Creation and analysis of the consolidated and deidentified HADS database was approved by the Institutional Review Boards of the Uniformed Services University of the Health Sciences, University of Michigan Institute for Social Research, University of California—San Diego, and Harvard Medical School.

The analytic sample for this study included all 9650 Regular Army enlisted soldiers who attempted suicide from 2004 to 2009 (excluding Army National Guard and Reserve), plus an equal-probability sample of 153,528 control person-months (person-months in which a documented SA did not occur). Data were analyzed using a discrete-time survival framework with person-month as the unit of analysis, such that each month in a soldier’s career was treated as a separate observational record.18,19 Discrete-time survival coefficients can be estimated without bias when control person-months are randomly subsampled and weighted using the logic of case-control analysis.20 Therefore, in order to reduce computational intensity, we selected an equal-probability 1:200 sample of control person-months after stratifying the population by sex, rank, time in service, and deployment status (never, currently, or previously deployed). Control person-months excluded all soldiers with a documented SA or other nonfatal suicidal event (eg, suicide ideation)5 and person-months in which a soldier died. Each control person-month was assigned a weight of 200 (the inverse probability of selection) to adjust for undersampling.

Measures

SAs were identified using Army/DoD administrative records from: the DoD Suicide Event Report (DoDSER),21 a DoD-wide surveillance mechanism that aggregates information on suicidal behaviors by a standardized form completed by medical providers at DoD treatment facilities; and ICD-9-CM diagnostic codes E950–E958 (indicating self-inflicted poisoning or injury with suicidal intent) from the Military Health System Data Repository (MDR), Theater Medical Data Store (TMDS), and TRANSCOM (Transportation Command) Regulating and Command and Control Evacuating System (TRAC2ES), which together provide health care encounter information from military and civilian treatment facilities, combat operations, and aeromedical evacuations (eTable 1, Supplemental Digital Content 1, https://links.lww.com/MLR/C105). Although use of the DoDSER began in 2008, its database includes records from its Army-specific predecessor, the Army Suicide Event Report (ASER), which was active from 2004 to 2007. The ICD-9-CM E959 code (late effects of a self-inflicted injury) was excluded, as it confounds temporal relationships between predictor variables and the SA.22 For soldiers with multiple SAs, we selected the first attempt using a hierarchical classification scheme prioritizing DoDSER records (details available elsewhere5).

We used administrative records to construct indicators for sociodemographic characteristics (sex, current age, race/ethnicity, education, marital status), service-related characteristics [age at entry into Army service, time in service, deployment status (never deployed, currently deployed, previously deployed)], and presence/recency of MHDx (eTable 1, Supplemental Digital Content 1, https://links.lww.com/MLR/C105). The MHDx variable was created from ICD-9-CM mental disorder codes, excluding postconcussion syndrome, tobacco use disorder, and supplemental V-codes (eg, stressors/adversities, marital problems) (eTable 2, Supplemental Digital Content 1, https://links.lww.com/MLR/C105). Recency of diagnosis was determined based on number of months between the most recent diagnostic record and the SA (cases) or sampled person-month (controls).

Missing Values

During creation of the HADS integrated database, some item-level administrative data were missing for particular person-months. In most cases, these data could be recovered by cross-checking other data systems or other months in the same soldiers’ records. The few missing values that could not be recovered through cross-checking were assigned imputed values equal to subgroup modes.6

Analysis Methods

Analyses were conducted using SAS version 9.4.23 Person-month records were pooled over the 2004–2009 study period. The weighted sample was used to estimate population percentages for all variables. We used logistic regression models to examine the association of sex with SA in a given person-month before and after adjusting for all other predictors. We then examined the 2-way interaction between sex and each of the other predictors. Each interaction was examined in a separate multivariate model that adjusted for the main effects all other predictors. To correct for multiple comparisons, we conducted the false discovery rate (FDR) using PROC MULTTEST in SAS 9.4.23 The sample was then stratified by sex. Within each stratum, logistic regression analyses examined multivariate associations of the predictor variables with SA in a given person-month. Logistic regression coefficients were exponentiated to obtain odds ratio (OR) and 95% confidence interval (CI). Final model coefficients were then used to calculate standardized risk estimates24 (SREs; expressed as the number of suicide attempters per 100,000 person-years) for each category of each predictor variable. An SRE is an estimate of the SA rate expected under the model if the distributions of all other predictors were the same in each category of a given predictor. All logistic regression models included a dummy predictor for calendar month and year to control for secular trends. Coefficients of other predictors can consequently be interpreted as averaged within-month associations. To further examine changes in risk across time in service for women and men, discrete-time survival models were used to estimate risk (suicide attempters per 100,000 person-months) in each month since entering the Army.

RESULTS

Women comprised 13.7% of all enlisted person-months pooled over the study period (population percentages estimated based on the weighted sample) and 25.2% of person-months with a first SA. Across person-months, most women were younger than 30 (70.8%), Black, Hispanic, Asian, or other (non-White) race/ethnicity (59.6%), high school-educated (75.6%), and never married (54.2%). Most women had entered Army service before age 21 (61.3%), more than half (53.3%) were in their first term of enlistment (first 4 y of service), and 44.2% were in-theater (currently deployed) or had returned from deployment (previously deployed). More than 34% had previously received a MHDx (Table 1).

TABLE 1 - Distributions of Sociodemographic, Service-related, and Mental Health Predictors Among Regular Army Enlisted Women and Men*
Women (n=23,461 Person-Months) Men (n=139,717 Person-Months)
Predictor Variables Suicide Attempt Cases, N (%) Total Population, N (%) Suicide Attempt Cases, N (%) Total Population, N (%)
Sociodemographic predictors
 Current age
  <21 1020 (41.9) 747,420 (17.8) 2295 (31.8) 3,877,495 (14.6)
  21–24 818 (33.6) 1,290,018 (30.7) 2681 (37.2) 7,940,681 (30.0)
  25–29 367 (15.1) 939,567 (22.3) 1389 (19.3) 6,180,189 (23.3)
  30–34 137 (5.6) 519,337 (12.3) 498 (6.9) 3,720,098 (14.0)
  35–39 62 (2.5) 414,062 (9.8) 238 (3.3) 2,930,438 (11.1)
  40+ 32 (1.3) 297,032 (7.1) 113 (1.6) 1,858,913 (7.0)
 Race/ethnicity
  White 1407 (57.8) 1,696,807 (40.3) 5401 (74.9) 16,661,201 (62.9)
  Black 588 (24.1) 1,665,188 (39.6) 827 (11.5) 5,307,227 (20.0)
  Hispanic 281 (11.5) 540,481 (12.8) 698 (9.7) 3,016,698 (11.4)
  Asian 108 (4.4) 206,508 (4.9) 178 (2.5) 1,010,978 (3.8)
  Other 52 (2.1) 98,452 (2.3) 110 (1.5) 511,710 (1.9)
 Education
  <High school§ 583 (23.9) 344,983 (8.2) 2305 (32.0) 3,533,105 (13.3)
  High school 1727 (70.9) 3,180,127 (75.6) 4653 (64.5) 20,323,853 (76.7)
  Some college 67 (2.8) 342,467 (8.1) 124 (1.7) 1,362,124 (5.1)
  ≥College 59 (2.4) 339,859 (8.1) 132 (1.8) 1,288,732 (4.9)
 Marital status
  Never married 1603 (65.8) 2,280,403 (54.2) 3838 (53.2) 10,309,038 (38.9)
  Currently married 740 (30.4) 1,600,940 (38.1) 3234 (44.8) 15,213,834 (57.4)
  Previously married 93 (3.8) 326,093 (7.8) 142 (2.0) 984,942 (3.7)
Service-related predictors
 Age at Army entry
  <21 1650 (67.7) 2,580,050 (61.3) 4821 (66.8) 16,515,621 (62.3)
  21–24 506 (20.8) 977,106 (23.2) 1633 (22.6) 6,554,633 (24.7)
  25+ 280 (11.5) 650,280 (15.5) 760 (10.5) 3,437,560 (13.0)
 Time in service (y)
  1–2 1646 (67.6) 1,295,846 (30.8) 3770 (52.3) 7,264,770 (27.4)
  3–4 453 (18.6) 948,453 (22.5) 1825 (25.3) 5,871,425 (22.1)
  5–10 265 (10.9) 1,145,665 (27.2) 1277 (17.7) 7,177,077 (27.1)
  >10 72 (3.0) 817,472 (19.4) 342 (4.7) 6,194,542 (23.4)
 Deployment status
  Never deployed 1826 (75.0) 2,349,026 (55.8) 4068 (56.4) 10,082,268 (38.0)
  Currently deployed 223 (9.2) 678,023 (16.1) 717 (9.9) 6,495,117 (24.5)
  Previously deployed 387 (15.9) 1,180,387 (28.1) 2429 (33.7) 9,940,429 (37.5)
Mental health predictors
 Time since most recent mental health diagnosis (mo)
  No diagnosis 970 (39.8) 2,770,570 (65.8) 2906 (40.3) 20,385,706 (76.9)
  1 857 (35.2) 224,257 (5.3) 2659 (36.9) 926,659 (3.5)
  2–3 226 (9.3) 169,426 (4.0) 607 (8.4) 686,607 (2.6)
  4–12 252 (10.3) 399,652 (9.5) 635 (8.8) 1,590,235 (6.0)
  ≥13 131 (5.4) 643,531 (15.3) 407 (5.6) 2,918,607 (11.0)
Total 2436 (100) 4,207,436 (100) 7214 (100) 26,507,814 (100)
*The sample of person-months from Regular Army enlisted soldiers (n=9650 suicide attempt person-months, 153,528 control person-months) is a subset of the total Regular Army sample (n=193,617 person-months) from the Army STARRS Historical Administrative Data Study. Control person-months were assigned a weight of 200 to adjust for undersampling.
Includes both cases (ie, person-months with a first suicide attempt) and weighted control person-months.
Weighted percent of the population within each stratum (ie, females and males).
§<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 nontraditional high school credentials.
Army STARRS indicates Army Study to Assess Risk and Resilience in Servicemembers.

Enlisted men, who comprised 86.3% of all enlisted person-months and 74.8% of person-months with a first SA, were mostly younger than 30 (67.9%), non-Hispanic White (62.9%), high school-educated (76.7%), currently married (57.4%), and younger than 21 when they entered service (62.3%). Approximately half (49.5%) were in their first term of enlistment, and 62.0% were currently or previously deployed. Approximately 23% had a MHDx history (Table 1).

Annually, rates of first-recorded SA ranged from 363/100,000 person-years (2004) to 841/100,000 person-years (2008) for women, versus 183/100,000 person-years (2004) to 405/100,000 (2008) for men. In 2009, rates for women and men decreased 11.5% and 9.9%, respectively, relative to their 2008 peak during the study period (Fig. 1). The overall 2004–2009 SA crude rate was significantly higher for women (695/100,000 person-years) than men (327/100,000 person-years) (rate ratio=2.1; 95% CI, 2.1–2.2). Women had higher odds of SA in a given person-month (OR=2.2; 95% CI, 2.1–2.3), even after controlling for sociodemographic, service-related, and MHDx variables (OR=1.6; 95% CI, 1.5–1.7).

F1
FIGURE 1:
Annual rate of first-recorded suicide attempt among Regular Army enlisted women and men.* *This sample of female (n=2436 cases, 21,025 control person-months) and male (n=7214 cases, 132,503 control person-months) enlisted soldiers from the Army STARRS Historical Administrative Data Study includes all Regular Army soldiers (ie, excluding those in the US Army National Guard and Army Reserve) with a suicide attempt in their administrative records during the years 2004–2009, plus a 1:200 stratified probability sample of all other Regular Army enlisted person-months. Control person-months were assigned a weight of 200 to adjust for undersampling.

A within-sex multivariate logistic regression analysis found that, in a given person-month pooled over time, women with significantly elevated odds of SA were younger, non-Hispanic White, less educated, in their first term of enlistment, and never or previously deployed (vs. currently deployed) as of that person-month. They were also more likely to have received a MHDx, with odds of SA increasing monotonically as time since most recent diagnosis decreased (univariate results in eTable 3, Supplemental Digital Content 1, https://links.lww.com/MLR/C105). Women with the highest standardized risk in a given person-month were younger than 21 (953/100,000 person-years), less than high school-educated (1072/100,000 person-years), and in their first 2 years of service (1175/100,000 person-years). SREs for women with a MHDx history before a given person-month ranged from 4172/100,000 person-years for women diagnosed in the previous month to 520/100,000 person-years for those most recently diagnosed more than a year prior. Women with no MHDx history as of a given person-month had an SRE of 374/100,000 person-years (Table 2).

TABLE 2 - Multivariate Associations of Sociodemographic, Service-related, and Mental Health Predictors With First Suicide Attempt Among Regular Army Enlisted Women and Men
Women (n=23,461) Men (n=139,717)
Predictor Variables OR 95% CI SRE OR 95% CI SRE Predictor by Sex Interaction§
Sociodemographic predictors
 Current age χ2 5=24.1* (FDR P=0.0004)
  <21 1.8* 1.3–2.5 953 2.0* 1.7–2.4 475
  21–24 1.3 1.0–1.7 666 1.4* 1.2–1.6 323
  25–29 1.0 0.8–1.3 548 1.2* 1.0–1.3 275
  30–34 1.0 528 1.0 235
  35–39 0.7* 0.5–0.9 351 0.9 0.7–1.0 205
  40+ 0.6* 0.4–0.9 318 0.8* 0.6–1.0 184
 Within-sex association χ2 5=45.4* (P<0.0001) χ2 5=131.9* (P<0.0001)
  Race/ethnicity χ2 4=10.4* (FDR P=0.0465)
   White 1.0 732 1.0 344
   Black 0.9* 0.8–1.0 639 0.8* 0.7–0.9 274
   Hispanic 0.8* 0.7–1.0 618 0.9* 0.8–0.9 293
   Asian 1.1 0.9–1.3 770 0.8* 0.7–0.9 273
   Other 1.0 0.8–1.4 757 1.0 0.8–1.2 348
 Within-sex association χ2 4=12.8* (P=0.0123) χ2 4=52.3* (P<0.0001)
  Education χ2 3=3.8 (FDR P=0.32)
   <High school 1.7* 1.5–1.8 1072 1.6* 1.5–1.7 459
   High school 1.0 648 1.0 292
   Some college 0.8* 0.6–1.0 493 0.8* 0.7–1.0 246
   ≥College 0.6* 0.4–0.8 370 0.7* 0.6–0.9 215
 Within-sex association χ2 3=134.9* (P<0.0001) χ2 3=320.3* (P<0.0001)
  Marital status χ2 2=16.5* (FDR P=0.0004)
   Never married 1.1 1.0–1.2 709 1.0 0.9–1.0 322
   Currently married 1.0 662 1.0 333
   Previously married 1.1 0.9–1.4 743 0.9 0.8–1.1 311
 Within-sex association χ2 2=2.6 (P=0.27) χ2 2=1.8 (P=0.41)
Service-related predictors
 Age at Army entry χ2 2=1.1 (FDR P=0.58)
  <21 1.0 0.8–1.1 684 1.0 0.9–1.1 325
  21–24 1.0 716 1.0 328
  25+ 1.0 0.8–1.2 723 1.0 0.9–1.1 336
 Within-sex association χ2 2=0.45 (P=0.80) χ2 2=0.28 (P=0.87)
  Time in service (y) χ2 3=71.9* (FDR P<0.0001)
   1–2 3.7* 3.1–4.5 1175 2.3* 2.1–2.6 511
   3–4 1.6* 1.3–1.9 493 1.5* 1.3–1.6 324
   5–10 1.0 315 1.0 222
   >10 0.7* 0.5–0.9 210 0.5* 0.4–0.5 103
 Within-sex association χ2 3=264.3* (P<0.0001) χ2 3=349.3* (P<0.0001)
  Deployment status χ2 2=47.2* (FDR P<0.0001)
   Never deployed 1.3* 1.2–1.5 712 2.2* 2.0–2.4 371
   Currently deployed 1.0 535 1.0 171
   Previously deployed 1.4* 1.2–1.6 737 2.0* 1.9–2.2 350
 Within-sex association χ2 2=17.3* (P=0.0002) χ2 2=352.9* (P=0.0001)
Mental health predictors
 Time since most recent mental health diagnosis (mo) χ2 4=160.7* (FDR P<0.0001)
  No diagnosis 1.0 374 1.0 160
  1 11.2* 10.2–12.4 4174 20.2* 19.1–21.3 3210
  2–3 4.0* 3.5–4.6 1496 6.6* 6.0–7.2 1050
  4–12 2.2* 1.9–2.6 832 3.4* 3.2–3.8 549
  ≥13 1.4* 1.2–1.7 520 1.9* 1.7–2.1 306
 Within-sex association χ2 4=2567.8* (P<0.0001) χ2 4=11,860.7* (P<0.0001)
The sample of Regular Army enlisted soldiers (n=9650 cases, 153,528 control person-months) is a subset of the total Regular Army sample (n=193,617 person-months) from the Army STARRS Historical Administrative Data Study. Control person-months were assigned a weight of 200 to adjust for undersampling.
Each 2-way interaction was examined separately in a model that included all other predictor variables, but not other interactions. P-values for the interactions were corrected using FDR.
§SRE (number of soldiers with a suicide attempt per 100,000 person-years) were calculated assuming other predictors were at their sample-wide means.
<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 nontraditional high school credentials.
Army STARRS indicates Army Study to Assess Risk and Resilience in Servicemembers; CI, confidence interval; FDR, false discovery rate; OR, odds ratio; SRE, standardized risk estimates.
*P<0.05.

Interaction terms indicated that associations of current age (χ25=24.1; FDR P=0.0004), race/ethnicity (χ24=10.4; FDR P=0.0465), time in service (χ23=71.9; FDR P<0.0001), deployment status (χ22=47.2; FDR P<0.0001), and MHDx (χ24=160.7; FDR P<0.0001) with SA differed by sex (Table 2). Significant interactions were explored with pairwise analyses. Although there were statically significant pairwise differences on a number of risk factors, the ORs for women and men were in the same direction and most differences were small, as can be seen in Table 2. In a given person-month, Black (χ21=4.0; P=0.047) and Asian (χ21=7.4; P=0.006) race/ethnicity were more protective for men than women. The OR associated with time in service of 1–2 years was larger for women than men (χ21=40.7; P<0.0001). In contrast, the ORs were larger among men for deployment status of never deployed (χ21=15.2; P<0.0001) or previously deployed (χ21=47.0; P<0.0001), and for MHDx at all recency levels (χ21=22.5–146.4; all P’s<0.0001). Despite a significant interaction between sex and current age, pairwise analyses comparing women and men within the same age group were nonsignificant.

Women had higher SREs than men across all predictor categories (Table 2). However, once sex differences in the population base rate of SA were accounted for, most SREs represented an increase or decrease in risk that was proportional between women and men (ie, for most predictor categories, women and men had SREs that were comparably larger or smaller than their respective base rates). Receipt of a MHDx in the previous month was associated with the largest sex difference relative to the base rates of women and men. That is, the SRE for women with a previous-month MHDx (4174/100,000 person-years) was 6.0 times higher than women’s base rate (695/100,000 person-years), whereas the SRE for men with a previous-month MHDx (3210/100,000 person-years) was 9.8 times higher than men’s base rate (327/100,000 person-years). Also noteworthy, the SRE for currently deployed women (535/100,000 person-years) was 23% lower than women’s base rate, whereas the SRE for currently deployed men (171/100,000 person-years) was 48% lower than their base rate.

Discrete-time survival models further examined sex differences in SA risk as a function of time in service (Fig. 2). The overall pattern of risk was similar for women and men, peaking during the second month of service and then decreasing as time in service increased. During the first 12 months of service, women had an average monthly risk that was 2.6 times that of men. Sex differences in risk diminished as time in service increased, particularly during the first term of enlistment (approximately the first 48 mo of service). During the fifth year of service (the start of the second term of enlistment for many soldiers), the average monthly risk for women was 1.4 times that of men. The diminished difference between women and men over time was the result of a sharper decrease in risk among women than men. For example, from the point of peak risk (second month) to the end of the first term of enlistment (48th month), risk decreased 84.8% among women and 64.6% among men.

F2
FIGURE 2:
Risk of suicide attempt time in service among Regular Army enlisted women and men.* *This sample of female (n=2436 cases, 21,025 control person-months) and male (n=7214 cases, 132,503 control person-months) enlisted soldiers from the Army STARRS Historical Administrative Data Study includes all Regular Army soldiers (ie, excluding those in the US Army National Guard and Army Reserve) with a suicide attempt in their administrative records during the years 2004–2009, plus a 1:200 stratified probability sample of all other Regular Army enlisted person-months. Control person-months were assigned a weight of 200 to adjust for undersampling.

DISCUSSION

For both women and men in the Regular Army enlisted population, medically documented SA rates increased overall during the 2004–2009 study period. Women comprised <14% of the enlisted population but more than 25% of suicide attempters. Consistent with previous research,3 women were more likely than men to have a documented SA after adjusting for other sociodemographic characteristics, service-related variables, and MHDx. Importantly, the risk factors for SA were similar for women and men. Women who attempted suicide were more likely to be younger, non-Hispanic White, less educated, in their first term of enlistment, never or previously deployed, and previously given a MHDx. The same characteristics were risk factors among men. Although 6 of the 8 predictors statistically differed by sex, their associations with SA were in the same direction and generally similar in magnitude. Exploratory analyses indicated that the largest differences were related to time in service, deployment status, and recency of mental health diagnosis. Specifically, being in the first 2 years of service had a stronger association with SA among women, whereas deployment status and recent MHDx had a stronger association with SA among men. Additional research is needed to confirm and understand these differences. Consistent with the overall difference between women and men in SA base rates, women had higher standardized risk across all categories of the predictor variables. Thus, in each segment of the population defined by our predictors, the average woman was at greater risk than the average man.

Among both women and men, those who were never deployed or previously deployed had higher odds of SA relative to those currently deployed (in-theater), consistent with research on the total enlisted population.3,11 Standardized risk was comparable between never and previously deployed women (712 vs. 737/100,000 person-years, respectively), and the same was true for men (371 vs. 350/100,000 person-years, respectively). However, the relative decrease in risk during deployment was more substantial for men than women. Specifically, risk among currently deployed women was 25%–27% lower compared with women who were never or previously deployed, whereas risk among currently deployed men was 51%–54% lower compared with never or previously deployed men. Previous research has highlighted the importance of deployment in suicide deaths among women.4,25 Given that men are more likely than women to report exposure to combat and postbattle stressors (eg, seeing dead bodies, caring for wounded),26 our finding that risk decreased more substantially among men than women in-theater highlights the importance of considering other deployment-related experiences that may disproportionately affect women. In addition to women’s elevated risk of experiencing sexual trauma in-theater,27,28 studies have found that women are more likely to report lower social support from unit members before and during deployment, as well as lower perceived preparedness for deployment,26,29 factors which are associated with elevated mental health risk.27,29–33 Our finding that a smaller proportion of women had deployed relative to men (44% vs. 62%) was likely due to sex differences in factors that are associated with selection for deployment, particularly military occupation. The extent to which this selection process affected our findings is not yet known and may have important implications for understanding sex-related similarities and differences in mental health and SA risk during the deployment cycle. Addressing this issue is particularly important given the significant change in Army policy in 2013 that lifted the ban on women serving in ground combat units.34 As more women begin to serve in direct combat roles, there will be even greater need for research on mental health and suicide risk in this population.

Approximately one third (34%) of all women had a MHDx history, compared with 23% of men. In contrast, ∼60% of both female and male suicide attempters had previously received a MHDx. Thus, although a larger proportion of enlisted women received a MHDx, women and men who attempted suicide were equally likely to have previously received attention from the mental health care system. Although the reason for this is not yet known, the types of MHDx received by women versus men may be related to SA risk following diagnosis. Previous research indicates that male soldiers have a higher prevalence of intermittent explosive disorder and substance use disorder, both of which are associated with SA risk.15 Men with those externalizing disorders may engage in behaviors that are more likely to be noticed by leaders and clinicians before a SA.

The association of recent MHDx with SA was stronger among men than women. Whereas women with a previous-month diagnosis had standardized SA risk that was 6 times higher than women’s population base rate, men with a previous-month diagnosis had risk that was nearly 10 times their base rate. Previous research has found that male soldiers with a self-reported mental disorder are less likely than women to perceive a need for treatment.35 This raises the possibility that male soldiers who come to the attention of clinicians and receive a MHDx are more severely distressed and/or impaired (resulting in higher SA risk) than women who receive a MHDx. However, this is an empirical question requiring investigation. As our findings also highlight, it is important for future research to address the significant challenge of identifying and intervening with the 40% female and male soldiers who were never diagnosed before attempting suicide.36

Time in service was associated with important similarities and differences in the patterns of risk for women and men. Previous research on the total enlisted population during the same time period demonstrated greatly elevated risk during the first term of enlistment, with the highest risk occurring during the second month of service, toward the end of basic training.3 We found that this general pattern was present in both sexes and that women and men differed primarily in level of risk. The higher SA risk among women was most pronounced during the first year of service, with risk differences between women and men diminishing as time in service increased. Women are more likely than men to begin Army service with a self-reported history of mental disorders and suicidal thoughts and behaviors,37,38 which may help account for our finding that sex differences in SA risk were greatest toward the beginning of service.

This study has 5 noteworthy limitations. First, administrative records are limited to events that come to the Army’s attention and are subject to classification and coding errors. Unreported and/or undiagnosed SAs and mental disorders were not captured. Because of this, sex differences in health care utilization and/or delivery39 may have affected our findings. In addition, efforts to identify and document suicidal behavior among service members increased during the 2004–2009 study period,5,21 and it is not known whether those changes differentially affected women and men. Second, we were unable to account for nonsuicidal self-injury,40 a clinically important outcome relevant to SA,41 for which there is no specific ICD-9-CM code. Third, pre-enlistment risk factors,42–44 which may help identify soldiers at greatest risk,45 are not captured in administrative records. Fourth, our findings may not generalize to other time periods (eg, different phases of Iraq/Afghanistan wars, other military conflicts) or populations (eg, other military branches, veterans, civilians). Fifth, soldiers were categorized based on the binary classification of gender in Army administrative records, which does not account for transgender soldiers or those with other, nonbinary gender identities who may experience unique stressors and elevated suicide risk.46,47

As demonstrated in the current study, accounting for sociodemographic characteristics, service-related characteristics, and MHDx attenuates, but does not eliminate, the higher odds of documented SA among women in the US Army. It is important for future research to expand the range of risk factors that may help explain the elevated risk among female soldiers to more effectively target interventions. Particular consideration should be given to stressors disproportionately experienced by military women. For example, female service members and veterans are more likely than men to report experiencing sexual assault, harassment, and gender discrimination during military service.48 Sexual assault and harassment during Army service are associated with increased risk for suicidal thoughts and behaviors, as well as other negative health and career outcomes (eg, demotion, attrition from service).49,50 Even after service has ended, veterans with a history of military sexual trauma have elevated risk of dying by suicide.51 It may also be valuable to examine familial factors that differ between women and men. For example, there is evidence that active duty women are more likely than men to be single parents, and nearly half of married military women (48%) have a spouse who is also a service member, versus only 7% of military men who are married.52 Importantly, documented family violence during service is associated with increased SA risk among both female and male soldiers, but those associations may be modified depending a soldier’s role as a victim or perpetrator of family violence.53 For example, male perpetrators were found to have higher odds of SA than male victims, but female perpetrators and victims did not differ in risk.53 It is also important to recognize the importance of premilitary trauma exposure.54 Women entering Army service are more likely than men to report frequent childhood sexual, physical, and emotional maltreatment, experiences associated with increased SA risk before enlistment.42

CONCLUSIONS

The Regular Army enlisted population examined in this study accounts for the majority of documented US Army SAs. However, it is noteworthy that similar sex-based risk discrepancies have been observed in Regular Army officers and activated officers and enlisted personnel in the US Army Reserve Components (Army National Guard and Army Reserve),3,55 indicating a need for focused research on women throughout the total Army population. Although the elevated risk among women may be partially explained by sociodemographic, service-related, and mental health variables, women continued to have higher odds of attempt even after accounting for these variables. Importantly, we found that risk factors for SA were generally similar regardless of sex. These findings reinforce the importance of examining sex differences across a broader array of potential risk factors (eg, specific mental disorders, stressors before and during military service, familial and other interpersonal relationships). Research focused on the mental health of military women will become even more important as women begin to occupy combat roles not historically available to them.

REFERENCES

1. Fox KR, Millner AJ, Mukerji CE, et al. Examining the role of sex in self-injurious thoughts and behaviors. Clin Psychol Rev. 2018;66:3–11.
2. Nock MK, Borges G, Bromet EJ, et al. Suicide and suicidal behavior. Epidemiol Rev. 2008;30:133–154.
3. Ursano RJ, Kessler RC, Stein MB, et al. Suicide Attempts in the U.S. Army during the wars in Afghanistan and Iraq, 2004–2009. JAMA Psychiatry. 2015;72:917–926.
4. Gilman SE, Bromet EJ, Cox KL, et al. Socio-demographic and career history predictors of suicide and suicide mortality in the United States Army 2004-2009. Psychol Med. 2014;44:2579–2592.
5. Ursano RJ, Kessler RC, Heeringa SG, et al. Nonfatal suicidal behaviors in U.S. Army administrative records, 2004–2009: results from the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS). Psychiatry. 2015;78:1–21.
6. Schoenbaum M, Kessler RC, Gilman SE, et al. Predictors of suicide and accident death in the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS). JAMA Psychiatry. 2014;71:493–503.
7. Black SA, Gallaway MS, Bell MR, et al. Prevalence and risk factors associated with suicides of army soldiers 2001–2009. Mil Psychol. 2011;23:433–451.
8. Pruitt LD, Smolenski DJ, Bush NE, et al. DoDSER: Department of Defense Suicide Event Report: Calendar Year 2016 Annual Report. Falls Church, VA: Psychological Health Center of Excellence, Defense Health Agency; 2017.
9. Gibson N, Corrigan E, Kateley K, et al. Surveillance of Suicidal Behavior: January Through December 2016. Aberdeen Proving Ground, MD: Division of Behavioral and Social Health Outcomes Practice, Clinical Public Health and Epidemiology Directorate, U.S. Army Public Health Center; 2017.
10. Millner AJ, Ursano RJ, Hwang I, et al. Lifetime suicidal behaviors and career characteristics among U.S. Army soldiers: results from the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS). Suicide Life Threat Behav. 2018;48:230–250.
11. Ursano RJ, Kessler RC, Stein MB, et al. Risk factors, methods, and timing of suicide attempts among U.S. Army soldiers. JAMA Psychiatry. 2016;73:741–749.
12. Department of Defense. 2018 Demographics: Profile of the Military Community. Alexandria, VA: Department of Defense, Office of the Deputy Assistant Secretary of Defense for Military Community and Family Policy; 2018.
13. Street AE, Vogt D, Dutra L. A new generation of women veterans: stressors faced by women deployed to Iraq and Afghanistan. Clin Psychol Rev. 2009;29:685–694.
14. Snarr JD, Heyman RE, Smith AM. Recent suicidal ideation and suicide attempts in a large-scale survey of the U.S. Air Force: prevalences and demographic risk factors. Suicide Life Threat Behav. 2010;40:544–552.
15. Millner AJ, Ursano RJ, Hwang I, et al. 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). Suicide Life Threat Behav. 2019;49:3–22.
16. Ursano RJ, Colpe LJ, Heeringa SG, et al. The Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS). Psychiatry. 2014;72:107–119.
17. Kessler RC, Colpe LJ, Fullerton CS, et al. Design of the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS). Int J Methods Psychiatr Res. 2013;22:267–275.
18. Willett JB, Singer JD. Investigating onset, cessation, relapse, and recovery: why you should, and how you can, use discrete-time survival analysis to examine event occurrence. J Consult Clin Psychol. 1993;61:952–965.
19. Singer JD, Willett JB. Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence. New York, NY: Oxford University Press; 2003.
20. Schlesselman JJ. Case-Control Studies: Design, Conduct, Analysis. New York, NY: Oxford University Press; 1982.
21. Gahm GA, Reger MA, Kinn JT, et al. Addressing the surveillance goal in the National Strategy for Suicide Prevention: the Department of Defense Suicide Event Report. Am J Public Health. 2012;102(suppl 1):S24–S28.
22. Walkup JT, Townsend L, Crystal S, et al. A systematic review of validated methods for identifying suicide or suicidal ideation using administrative or claims data. Pharmacoepidemiol Drug Saf. 2012;21(suppl 1):174–182.
23. SAS Institute Inc. SAS®9.4 Software. Cary, NC: SAS Institute Inc.; 2013.
24. Roalfe AK, Holder RL, Wilson S. Standardisation of rates using logistic regression: a comparison with the direct method. BMC Health Serv Res. 2008;8:275.
25. Street AE, Gilman SE, Rosellini AJ, et al. Understanding the elevated suicide risk of female soldiers during deployments. Psychol Med. 2015;45:717–726.
26. Street AE, Gradus JL, Giasson HL, et al. Gender differences among veterans deployed in support of the wars in Afghanistan and Iraq. J Intern Med. 2013;28(suppl 2):S556–S562.
27. Banducci AN, McCaughey VK, Gradus JL, et al. The associations between deployment experiences, PTSD, and alcohol use among male and female veterans. Addict Behav. 2019;98:106032.
28. Walsh K, Galea S, Cerda M, et al. Unit support protects against sexual harassment and assault among National Guard soldiers. Womens Health Issues. 2014;24:600–604.
29. Carter-Visscher R, Polusny MA, Murdoch M, et al. Predeployment gender differences in stressors and mental health among U.S. National Guard troops poised for Operation Iraqi Freedom deployment. J Trauma Stress. 2010;23:78–85.
30. Franz MR, Wolf EJ, MacDonald HZ, et al. Relationships among predeployment risk factors, warzone-threat appraisal, and postdeployment PTSD symptoms. J Trauma Stress. 2013;26:498–506.
31. Mott JM, Graham DP, Teng EJ. Perceived threat during deployment: risk factors and relation to axis i disorders. Psychol Trauma Theory Res Pract Policy. 2012;4:587–595.
32. Renshaw KD. An integrated model of risk and protective factors for post-deployment PTSD symptoms in OEF/OIF era combat veterans. J Affect Disord. 2011;128:321–326.
33. Han SC, Castro F, Lee LO, et al. Military unit support, postdeployment social support, and PTSD symptoms among active duty and National Guard soldiers deployed to Iraq. J Anxiety Disord. 2014;28:446–453.
34. Servick K. Setting the bar. Science. 2015;349:468–471.
35. Naifeh JA, Colpe LJ, Aliaga PA, et al. Barriers to initiating and continuing mental health treatment among soldiers in the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS). Mil Med. 2016;181:1021–1032.
36. Ursano RJ, Kessler RC, Naifeh JA, et al. Risk factors associated with attempted suicide among U.S. Army soldiers without a history of mental health diagnosis. JAMA Psychiatry. 2018;75:1022–1032.
37. Ursano RJ, Heeringa SG, Stein MB, et al. Prevalence and correlates of suicidal behavior among new soldiers in the US Army: results from the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS). Depress Anxiety. 2015;32:3–12.
38. Rosellini AJ, Heeringa SG, Stein MB, et al. Lifetime prevalence of DSM-IV mental disorders among new soldiers in the U.S. Army: results from the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS). Depress Anxiety. 2015;32:13–24.
39. Haskell SG, Mattocks K, Goulet JL, et al. The burden of illness in the first year home: do male and female VA users differ in health conditions and healthcare utilization. Womens Health Issues. 2011;21:92–97.
40. Nock MK. Understanding Nonsuicidal Self-injury: Origins, Assessment, and Treatment. Washington, DC: American Psychological Association; 2009.
41. Turner BJ, Kleinman EM, Nock MK. Non-suicidal self-injury prevalence, course, and association with suicidal thoughts and behaviors in two large, representative samples of US Army soldiers. Psychol Med. 2019;49:1470–1480.
42. Stein MB, Campbell-Sills L, Ursano RJ, et al. Childhood maltreatment and lifetime suicidal behaviors among new soldiers in the US Army: results from the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS). J Clin Psychiatry. 2018;79:70–77.
43. Nock MK, Ursano RJ, Heeringa SG, et al. Mental disorders, comorbidity, and pre-enlistment suicidal behavior among new soldiers in the U.S. Army: results from the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS). Suicide Life Threat Behav. 2015;45:588–599.
44. Campbell-Sills L, Kessler RC, Ursano RJ, et al. Associations of childhood bullying victimization with lifetime suicidal behaviors among new U.S. Army soldiers. Depress Anxiety. 2017;34:701–710.
45. Bernecker SL, Rosellini AJ, Nock MK, et al. Improving risk prediction accuracy for new soldiers in the U.S. Army by adding self-report survey data to administrative data. BMC Psychiatry. 2018;18:87.
46. Tucker RP, Testa RJ, Reger MA, et al. Current and military-specific gender minority stress factors and their relationship with suicide ideation in transgender veterans. Suicide Life Threat Behav. 2019;49:155–166.
47. Blosnich JR, Brown GR, Shipherd JC, et al. Prevalence of gender identity disorder and suicide risk among transgender veterans utilizing Veterans Health Administration care. Am J Public Health. 2013;103:e27–e32.
48. Breslin RA, Davis L, Hylton K, et al. 2018 Workplace and Gender Relations Survey of Active Duty Members: Overview Report. Alexandria, VA: Office of People Analytics, U.S. Department of Defense; 2019.
49. Rosellini AJ, Street AE, Ursano RJ, et al. Sexual assault victimization and mental health treatment, suicide attempts, and career outcomes among women in the U.S. Army. Am J Public Health. 2017;107:732–739.
50. Griffith J. The sexual harassment-suicide connection in the U.S. military: contextual effects of hostile work environment and trusted unit leaders. Suicide Life Threat Behav. 2017;49:41–53.
51. Kimerling R, Makin-Byrd K, Louzon S, et al. Military sexual trauma and suicide mortality. Am J Prev Med. 2016;50:684–691.
52. U.S. Department of defense. Demographics 2010: profile of the military community. Washington, DC: Office of the Deputy Under Secretary of Defense (Military Community and Family Policy); 2010. Available at: https://download.militaryonesource.mil/12038/MOS/Reports/2010-Demographics-Report.pdf.
53. Ursano RJ, Stein MB, Herberman Mash HB, et al. Documented family violence and risk of suicide attempt among U.S. Army soldiers. Psychiatry Res. 2018;262:575–582.
54. Bryan CJ, Bryan AO, Clemans TA. The association of military and premilitary sexual trauma with risk for suicide ideation, plans, and attempts. Psychiatry Res. 2015;227:246–252.
55. Naifeh JA, Ursano RJ, Kessler RC, et al. Suicide attempts among activated soldiers in the U.S. Army reserve components. BMC Psychiatry. 2019;19:31.
Keywords:

military; gender; women; men; suicide attempt

Supplemental Digital Content