[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Figure 1.
Standardized Risk Estimates (SREs) of Suicide Attempt by Number of Past-Year Unit Suicide Attempts and Military Occupational Specialty (MOS)
Standardized Risk Estimates (SREs) of Suicide Attempt by Number of Past-Year Unit Suicide Attempts and Military Occupational Specialty (MOS)

The SREs (number of soldiers who attempted suicide per 100 000 person-years) were calculated assuming other predictors were at their sample-wide means. The SREs were calculated separately for each MOS based on logistic regression models that included basic sociodemographic and service-related variables (sex, age at entry into the army, current age, race/ethnicity, educational level, marital status, time in service, deployment status, unit size, and number of past-year unit suicide attempts) and included a dummy predictor variable for calendar month and year to control for secular trends.

Figure 2.
Standardized Risk Estimates (SREs) of Suicide Attempt by Number of Past-Year Unit Suicide Attempts and Unit Size
Standardized Risk Estimates (SREs) of Suicide Attempt by Number of Past-Year Unit Suicide Attempts and Unit Size

The SREs (number of soldiers who attempted suicide per 100 000 person-years) were calculated assuming other predictors were at their sample-wide means. The SREs were calculated separately for each unit size based on logistic regression models that included basic sociodemographic and service-related variables (sex, age at entry into the army, current age, race/ethnicity, educational level, marital status, time in service, deployment status, military occupation, and number of past-year unit suicide attempts) and included a dummy predictor variable for calendar month and year to control for secular trends.

Table 1.  
Univariate and Multivariate Association of Unit Suicide Attempts With Suicide Attempts Among Regular Army, Enlisted Soldiersa
Univariate and Multivariate Association of Unit Suicide Attempts With Suicide Attempts Among Regular Army, Enlisted Soldiersa
Table 2.  
Multivariate Associations of Unit Suicide Attempts With Suicide Attempt Among Regular Army, Enlisted Soldiers Stratified by Military Occupationa
Multivariate Associations of Unit Suicide Attempts With Suicide Attempt Among Regular Army, Enlisted Soldiers Stratified by Military Occupationa
Table 3.  
Multivariate Associations of Unit Suicide Attempts With Suicide Attempt Among Regular Army, Enlisted Soldiers Stratified by Unit Sizea
Multivariate Associations of Unit Suicide Attempts With Suicide Attempt Among Regular Army, Enlisted Soldiers Stratified by Unit Sizea
1.
Ursano  RJ, Kessler  RC, Heeringa  SG,  et al; Army STARRS Collaborators.  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):1-21.PubMedGoogle Scholar
2.
Ursano  RJ, Kessler  RC, Stein  MB,  et al; Army STARRS Collaborators.  Risk factors, methods, and timing of suicide attempts among US Army soldiers.  JAMA Psychiatry. 2016;73(7):741-749.PubMedGoogle ScholarCrossref
3.
Ursano  RJ, Kessler  RC, Stein  MB,  et al; Army Study to Assess Risk and Resilience in Servicemembers Collaborators.  Suicide attempts in the US Army during the wars in Afghanistan and Iraq, 2004–2009.  JAMA Psychiatry. 2015;72(9):917-926.PubMedGoogle ScholarCrossref
4.
Bryan  CJ, Griffith  JH, Pace  BT,  et al.  Combat exposure and risk for suicidal thoughts and behaviors among military personnel and veterans: a systematic review and meta-analysis.  Suicide Life Threat Behav. 2015;45(5):633-649.Google ScholarCrossref
5.
Kok  BC, Herrell  RK, Thomas  JL, Hoge  CW.  Posttraumatic stress disorder associated with combat service in Iraq or Afghanistan: reconciling prevalence differences between studies.  J Nerv Ment Dis. 2012;200(5):444-450.PubMedGoogle ScholarCrossref
6.
Englert  DR, Hunter  CL, Sweeney  BJ.  Mental health evaluations of U.S. Air Force basic military training and technical training students.  Mil Med. 2003;168(11):904-910.PubMedGoogle Scholar
7.
Britt  TW, Davison  J, Bliese  PD, Castro  CA.  How leaders can influence the impact that stressors have on soldiers.  Mil Med. 2004;169(7):541-545.PubMedGoogle ScholarCrossref
8.
Oliver  LW, Harman  J, Hoover  E, Hayes  SM, Pandhi  NA.  A quantitative integration of the military cohesion literature.  Mil Psychol. 1999;11(1):57-83.Google ScholarCrossref
9.
Williams  J, Brown  JM, Bray  RM, Anderson Goodell  EM, Olmsted  KR, Adler  AB.  Unit cohesion, resilience, and mental health of soldiers in basic combat training.  Mil Psychol. 2016;28(4):241-250.Google ScholarCrossref
10.
Wright  BK, Kelsall  HL, Sim  MR, Clarke  DM, Creamer  MC.  Support mechanisms and vulnerabilities in relation to PTSD in veterans of the Gulf War, Iraq War, and Afghanistan deployments: a systematic review.  J Trauma Stress. 2013;26(3):310-318.PubMedGoogle ScholarCrossref
11.
Mitchell  MM, Gallaway  MS, Millikan  AM, Bell  M.  Interaction of combat exposure and unit cohesion in predicting suicide-related ideation among post-deployment soldiers.  Suicide Life Threat Behav. 2012;42(5):486-494.PubMedGoogle ScholarCrossref
12.
Bartone  PT, Wright  KM.  Grief and group recovery following a military air disaster.  J Trauma Stress. 1990;3(4):523-539.Google ScholarCrossref
13.
Fullerton  CS, Ursano  RJ, Kao  T-C, Bharitya  VR.  Disaster-related bereavement: acute symptoms and subsequent depression.  Aviat Space Environ Med. 1999;70(9):902-909.PubMedGoogle Scholar
14.
Toblin  RL, Riviere  LA, Thomas  JL, Adler  AB, Kok  BC, Hoge  CW.  Grief and physical health outcomes in U.S. soldiers returning from combat.  J Affect Disord. 2012;136(3):469-475.PubMedGoogle ScholarCrossref
15.
Carr  RB.  When a soldier commits suicide in Iraq: impact on unit and caregivers.  Psychiatry. 2011;74(2):95-106.PubMedGoogle ScholarCrossref
16.
Swanson  SA, Colman  I.  Association between exposure to suicide and suicidality outcomes in youth.  CMAJ. 2013;185(10):870-877.PubMedGoogle ScholarCrossref
17.
Cox  GR, Robinson  J, Williamson  M, Lockley  A, Cheung  YTD, Pirkis  J.  Suicide clusters in young people: evidence for the effectiveness of postvention strategies.  Crisis. 2012;33(4):208-214.PubMedGoogle ScholarCrossref
18.
O’Carroll  PW, Mercy  JA, Steward  JA; Centers for Disease Control (CDC).  CDC recommendations for a community plan for the prevention and containment of suicide clusters.  MMWR Suppl. 1988;37(6):1-12.PubMedGoogle Scholar
19.
Kessler  RC, Stein  MB, Bliese  PD,  et al.  Occupational differences in US Army suicide rates.  Psychol Med. 2015;45(15):3293-3304.PubMedGoogle ScholarCrossref
20.
Henderson  WD. Assessing cohesion in small units. In: Henderson  WD, ed.  Cohesion, the Human Element in Combat: Leadership and Societal Influence in the Armies of the Soviet Union, the United States, North Vietnam, and Israel. Washington, DC: National Defense University Press; 1985:13-26.
21.
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(4):267-275.PubMedGoogle ScholarCrossref
22.
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(6):952-965.PubMedGoogle ScholarCrossref
23.
Gahm  GA, Reger  MA, Kinn  JT, Luxton  DD, Skopp  NA, Bush  NE.  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.PubMedGoogle ScholarCrossref
24.
Walkup  JT, Townsend  L, Crystal  S, Olfson  M.  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.PubMedGoogle ScholarCrossref
25.
Giovannelli  D.  Profile of the U.S. Army 2008: A Reference Handbook. Arlington, VA: Institute of Land Warfare, Association of the United States Army; 2008. https://www.ausa.org/sites/default/files/SR-2008-Profile-of-the-US-Army.pdf. Accessed June 16, 2017.
26.
SAS Institute Inc.  SAS 9.4 Software [computer program]. Cary, NC: SAS Institute Inc; 2013.
27.
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.PubMedGoogle ScholarCrossref
28.
Rothman  K, Greenland  S.  Modern Epidemiology. 2nd ed. Philadelphia, PA: Lippincott Williams & Wilkins; 1998.
29.
Andriessen  K.  Can postvention be prevention?  Crisis. 2009;30(1):43-47.PubMedGoogle ScholarCrossref
30.
Berman  AL.  Estimating the population of survivors of suicide: seeking an evidence base.  Suicide Life Threat Behav. 2011;41(1):110-116.PubMedGoogle ScholarCrossref
31.
Cerel  J, Maple  M, van de Venne  J, Moore  M, Flaherty  C, Brown  M.  Exposure to suicide in the community: prevalence and correlates in one US state.  Public Health Rep. 2016;131(1):100-107.PubMedGoogle ScholarCrossref
32.
Cerel  J, van de Venne  JG, Moore  MM, Maple  MJ, Flaherty  C, Brown  MM.  Veteran exposure to suicide: prevalence and correlates.  J Affect Disord. 2015;179:82-87.PubMedGoogle ScholarCrossref
33.
Cerel  J, McIntosh  JL, Neimeyer  RA, Maple  M, Marshall  D.  The continuum of “survivorship”: definitional issues in the aftermath of suicide.  Suicide Life Threat Behav. 2014;44(6):591-600.PubMedGoogle ScholarCrossref
34.
Bolton  JM, Au  W, Leslie  WD,  et al.  Parents bereaved by offspring suicide: a population-based longitudinal case-control study.  JAMA Psychiatry. 2013;70(2):158-167.PubMedGoogle ScholarCrossref
35.
Erlich  MD. Addressing the aftermath of suicide: why we need postvention. Psychiatric Times. 2016. http://www.psychiatrictimes.com/suicide/addressing-aftermath-suicide-why-we-need-postvention. Accessed January 30, 2017.
Original Investigation
September 2017

Risk of Suicide Attempt Among Soldiers in Army Units With a History of Suicide Attempts

Author Affiliations
  • 1Department of Psychiatry, Center for the Study of Traumatic Stress, Uniformed Services University of the Health Sciences, Bethesda, Maryland
  • 2Department of Health Care Policy, Harvard Medical School, Boston, Massachusetts
  • 3Darla School of Business, University of South Carolina, Columbia
  • 4Department of Preventive Medicine and Biostatistics, Uniformed Services University of the Health Sciences, Bethesda, Maryland
  • 5National Institute of Mental Health, Bethesda, Maryland
  • 6Institute for Social Research, University of Michigan, Ann Arbor
  • 7Department of Psychiatry, University of California, San Diego, La Jolla
  • 8Department of Family Medicine and Public Health, University of California, San Diego, La Jolla
  • 9Veterans Affairs San Diego Healthcare System, San Diego, California
JAMA Psychiatry. 2017;74(9):924-931. doi:10.1001/jamapsychiatry.2017.1925
Key Points

Question  Does a previous suicide attempt in a soldier’s US Army unit increase the risk of subsequent suicide attempts?

Findings  In this longitudinal cohort study of administrative records from 9650 medically documented soldiers who attempted suicide between the years 2004 and 2009 and 153 528 control person-months, the risk of suicide attempt among soldiers increased as the number of past-year suicide attempts within their unit increased. This finding was true for combat arms and other occupations and for units of any size but particularly for smaller units (1-40 soldiers).

Meaning  Units with a history of suicide attempts may be important targets for preventive interventions.

Abstract

Importance  Mental health of soldiers is adversely affected by the death and injury of other unit members, but whether risk of suicide attempt is influenced by previous suicide attempts in a soldier’s unit is unknown.

Objective  To examine whether a soldier’s risk of suicide attempt is influenced by previous suicide attempts in that soldier’s unit.

Design, Setting, and Participants  Using administrative data from the Army Study to Assess Risk and Resilience in Servicemembers (STARRS), this study identified person-month records for all active-duty, regular US Army, enlisted soldiers who attempted suicide from January 1, 2004, through December 31, 2009 (n = 9650), and an equal-probability sample of control person-months (n = 153 528). Data analysis was performed from August 8, 2016, to April 10, 2017.

Main Outcomes and Measures  Logistic regression analyses examined the number of past-year suicide attempts in a soldier’s unit as a predictor of subsequent suicide attempt, controlling for sociodemographic features, service-related characteristics, prior mental health diagnosis, and other unit variables, including suicide-, combat-, and unintentional injury–related unit deaths. The study also examined whether the influence of previous unit suicide attempts varied by military occupational specialty (MOS) and unit size.

Results  Of the final analytic sample of 9512 enlisted soldiers who attempted suicide and 151 526 control person-months, most were male (86.4%), 29 years or younger (68.4%), younger than 21 years when entering the army (62.2%), white (59.8%), high school educated (76.6%), and currently married (54.8%). In adjusted models, soldiers were more likely to attempt suicide if 1 or more suicide attempts occurred in their unit during the past year (odds ratios [ORs], 1.4-2.3; P < .001), with odds increasing as the number of unit attempts increased. The odds of suicide attempt among soldiers in a unit with 5 or more past-year attempts was more than twice that of soldiers in a unit with no previous attempts (OR, 2.3; 95% CI, 2.1-2.6). The association of previous unit suicide attempts with subsequent risk was significant whether soldiers had a combat arms MOS or other MOS (ORs, 1.4-2.3; P < .001) and regardless of unit size, with the highest risk among those in smaller units (1-40 soldiers) (ORs, 2.1-5.9; P < .001). The population-attributable risk proportion for 1 or more unit suicide attempts in the past year indicated that, if this risk could be reduced to no unit attempts, 18.2% of attempts would not occur.

Conclusions and Relevance  Risk of suicide attempt among soldiers increased as the number of past-year suicide attempts within their unit increased for combat arms and other MOSs and for units of any size but particularly for smaller units. Units with a history of suicide attempts may be important targets for preventive interventions.

Introduction

The suicide attempt (SA) rate among US Army soldiers increased substantially during the Iraq and Afghanistan wars.1 Medically documented SAs are more likely to occur among soldiers never or previously deployed (vs currently deployed) and soldiers who are female, younger, non-Hispanic white, less educated, and in the early stages of army service, particularly the initial months of training.2,3 The risk of SA may also be influenced by events within a soldier’s army unit, including suicidal behavior of other unit members.

Units are the foundational structure of the US Army. Soldiers have a variety of shared unit experiences, including exposure to external group stressors (eg, training, deployment, and combat) and internal group stressors (eg, leadership changes, bullying by peers or leaders, and injury or death of unit members). External group stressors, particularly combat, adversely affect the mental health and suicide risk of soldiers within a unit.4-6 Internal unit characteristics, including leadership quality and group cohesion, can exacerbate or mitigate the influence of external group stressors on performance and health-related outcomes,7-10 including suicide ideation.11

Injuries and deaths from combat and unintentional events become internal group stressors that adversely affect the mental and physical health of unit members.12-14 Less is known about the influence of suicidal behavior on other unit members. Personal accounts suggest that suicidal events can have profound negative effects within a unit.15 Suicide contagion, in which exposure to suicide is proposed to increase suicidal behaviors in others, even if the person who died was not personally known,16 may result in clustering or multiple suicidal behaviors exhibited in a short period and/or specific geographic area or community.17,18 Multiple suicidal events may also reflect high unit stress attributable to factors such as poor leadership or low group cohesion.

To the extent that there is an association between previous SAs within a unit and subsequent SA risk among soldiers in that unit, the strength of that association may be modified by individual and unit characteristics. Military occupation influences suicide risk, with combat soldiers having significantly elevated rates of suicide death compared with other soldiers.19 It is not known whether this increased risk is associated with the number of previous unit suicides or SAs. Unit size is another potentially relevant characteristic. Small military units, with their shared culture, structure, and cohesion, may profoundly affect soldiers’ functioning and well-being.20 It is possible that previous SAs within smaller units have a greater influence on subsequent risk among unit members.

We examined whether a soldier’s risk of SA is influenced by previous SAs in that soldier’s unit after adjusting for individual sociodemographic and service-related characteristics, history of mental health diagnosis, and other internal unit stressors, including previous unit deaths from combat, unintentional injuries, and suicide. We also examined whether the influence of previous unit SAs varies by occupation and unit size.

Methods
Sample

This longitudinal, retrospective cohort study used data from the Army Study to Assess Risk and Resilience in Servicemembers (STARRS) Historical Administrative Data Study (HADS), which integrates 38 US Army and US Department of Defense administrative data systems and includes individual-level person-month records for all soldiers on active duty from January 1, 2004, through December 31, 2009 (n = 1.66 million).21 Data analysis was performed from August 8, 2016, to April 10, 2017. The HADS 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. These institutional review boards determined that the present study did not constitute human participant research because it relies entirely on deidentified secondary data.

The HADS includes administrative records for all 957 057 regular army soldiers on active duty during the 2004 to 2009 study period (excluding activated US Army National Guard and Army Reserve), 9791 of whom had a documented SA. The current study focused on enlisted soldiers, which were previously found to account for nearly 99% of SAs from 2004 through 2009.3 Using all enlisted soldiers who attempted suicide (n = 9650 cases) and an equal-probability 1:200 sample of control person-months (n = 153 528), the study3 excluded person-months that were not linked to an army unit (n = 82) and outlier person-months with a unit size greater than 600 (n = 2058). The final analytic sample included 9512 enlisted soldiers who attempted suicide (hereinafter referred to as suicide attempters) and 151 526 control person-months. In selecting controls, the population was stratified by sex, rank, time in service, deployment status (never, currently, or previously deployed), and historical time. Control person-months excluded all soldiers with a documented SA or other nonfatal suicidal event (eg, suicidal ideation)1 and person-months in which a soldier died. Data were analyzed using a discrete-time survival framework with a person-month unit of analysis,22 such that each month in a soldier’s career was treated as a separate observational record. Control person-months were assigned a weight of 200 to adjust for undersampling.

Measures
Suicide Attempt

Suicide attempters were identified using US Army and US Department of Defense administrative records from the Department of Defense Suicide Event Report,23 a US Department of Defense–wide surveillance mechanism completed by medical professionals at US Department of Defense treatment facilities, and ICD-9-CM diagnostic codes E950 to E958 (indicating self-inflicted poisoning or injury with suicidal intent) from health care encounter data systems for military and civilian treatment facilities, combat operations, and aeromedical evacuations (eTable 1 in the Supplement). We excluded suicide deaths, Department of Defense Suicide Event Report records indicating only suicide ideation, and E959 codes (late effects of a self-inflicted injury) because this code confounds the temporal associations between the predictor variables and SA.24 Records from different data systems were cross-referenced, ensuring all cases represent unique soldiers. For soldiers with multiple SAs, we selected the first attempt using a hierarchical classification scheme that prioritized Department of Defense Suicide Event Report records because of that system’s more extensive reporting requirements.1

Sociodemographic and Service-Related Variables

Sociodemographic (sex, current age, race, educational level, and marital status) and service-related variables (age at army entry, time in service [1-2, 3-4, 5-10, or >10 years], deployment status [never, currently, or previously deployed], and occupation [combat arms vs other]) (eTable 2 in the Supplement) were drawn from administrative data.

Previous Mental Health Diagnosis

We created an indicator variable for any previous mental health diagnosis during army service by combining ICD-9-CM codes from administrative medical records (eg, major depression, bipolar disorder, posttraumatic stress disorder, and personality disorders), excluding postconcussion syndrome, tobacco use disorder, and supplemental V codes that are not disorders (eg, stressors or adversities, marital problems) (eTable 3 in the Supplement).

Unit Variables

Unit-related variables for each soldier were constructed from administrative records. Duty units were determined using deidentified Unit Identification Codes. Soldiers with the same Unit Identification Code in a given calendar month were considered to be members of the same unit (eAppendix in the Supplement).

Unit SAs in the Past Year

We constructed a total count of SAs occurring in each soldier’s unit during the year before the sampled person-month record. This count excluded SAs made by case soldiers. The resulting predictor variable was interpreted from the sampled soldier’s perspective (ie, number of SAs in “my” duty unit during the past year). The same method was used to obtain counts of past-year unit suicide, combat, and unintentional deaths.

Unit Size

The size of a soldier’s army unit was determined based on the mean number of soldiers in that unit during the past 3 months (including the sampled person-month). For instance, the size of a given soldier’s unit in March 2009 was obtained by averaging the number of soldiers in that unit in January, February, and March 2009. Unit sizes (1-40, 41-100, 101-200, and 201-600 soldiers) were categorized to approximate army operational units.25

Statistical Analysis

Analyses were conducted using SAS statistical software, version 9.4.26 The association of past-year unit SAs with SA risk was first examined in a univariable logistic regression analysis, followed by a series of multivariable logistic regression analyses, adjusting for sociodemographic variables (sex, current age, race/ethnicity, educational level, and marital status), service-related characteristics (age at army entry, time in service, deployment status, and occupation), prior mental health diagnosis, and other unit variables (unit size, suicide deaths, combat deaths, and unintentional deaths). We then separately examined the interaction of unit SAs with occupation and unit size to determine whether the association of unit SAs with subsequent SA risk varied according to these characteristics. These interactions were further explored by stratifying the sample. Logistic regression coefficients were exponentiated to obtain odds ratios (ORs) and 95% CIs. Final model coefficients were used to generate a standardized risk estimate27 (SRE; number of suicide attempters per 100 000 person-years) for each category of a predictor under a model that assumed other predictors were at their sample-wide means. All logistic regression models included a dummy predictor for calendar month and year to control for increasing rates of SA from 2004 through 2009.1 Coefficients of other predictors were consequently interpreted as averaged within-month associations based on the assumption that effects of other predictors do not vary over time. Population-attributable risk proportions28 were calculated to identify the proportion of observed SAs that would not have occurred if the effects attributable to unit SAs were reduced to a reference level, assuming that coefficients in the model represent causal effects of the predictors.

Results

Of the 9512 enlisted suicide attempters and 151 526 control person-months, most were male (86.4%), 29 years or younger (68.4%), younger than 21 years when entering the army (62.2%), white (59.8%), high school educated (76.6%), and currently married (54.8%). Almost three-quarters (72.2%) had more than 2 years of service, 40.3% had never deployed, and 76.7% were assigned to an occupation other than combat arms (eTable 4 in the Supplement).

In univariate analyses, soldiers were more likely to attempt suicide if assigned to a unit with 1 or more past-year SAs (χ25 = 1534.6, P < .001) (Table 1), with odds increasing monotonically as the number of unit SAs increased from 1 (OR, 1.6; 95% CI, 1.5-1.7) to 5 or more (OR, 3.9; 95% CI, 3.6-4.3). This association remained significant after adjusting for sociodemographic variables, age at army entry, time in service, deployment status, occupation, and unit size (χ25 = 543.9, P < .001), with soldiers being twice as likely to attempt suicide if there were 5 or more past-year unit attempts than if there were no attempts (OR, 2.3; 95% CI, 2.1-2.6). Of importance, the results were unchanged when (1) examining the number of past-year unit attempters (not attempts) as the predictor, (2) excluding soldiers in their first 6 months of service (a period of initial army training associated with particularly high risk of SA3), (3) excluding soldiers in Warrior Transition Units (intensive rehabilitative care units for soldiers with significant physical and/or mental health needs), and (4) adjusting for the number of soldiers within a unit who had a past-year mental health diagnosis (an indicator of unit-based differences in distress). Results were also replicated in a generalized linear mixed-effects model and a generalized estimating equation model to account for any potential nonindependence (detailed results available on request).

The SREs of SA increased from 319 per 100 000 person-years for soldiers in a unit with no past-year SAs to 747 per 100 000 person-years for soldiers in a unit with 5 or more SAs. The population-attributable risk proportion for at least 1 unit SA in the past year (ie, the proportion of observed SAs associated with the predictor) based on this model was 18.2%, indicating that, if the risk associated with units that had at least 1 past-year SA could be reduced to those with no attempts, 18.2% of attempts would not occur. The population-attributable risk proportion for units with a history of 5 or more SAs revealed that 4.1% of subsequent SAs could be eliminated if risk was reduced to that of units in which there were 4 SAs or fewer.

We further examined the robustness of unit SAs by adding prior mental health diagnosis and past-year unit deaths due to suicide, combat, and unintentional injury to the model. The SA was related to prior mental health diagnosis (χ21 = 7169.7, P < .001) but not to past-year unit deaths due to suicide (χ21 = 1.4, P = .23), combat (χ26 = 1.3, P = .97), or unintentional injury (χ22 = 0.3, P = .87). Of importance, the association of past-year unit SAs with soldiers’ risk of SA remained significant after inclusion of these factors (χ25 = 505.5, P < .001) (eTable 5 in the Supplement).

To identify whether the association of unit SA with risk of SA was modified by occupation and unit size, we included their 2-way interactions in separate multivariable models, adjusting for sociodemographic and service-related variables. The interactions of unit SAs with occupation (χ25 = 13.2, P = .02) and unit size (χ215 = 87.6, P < .001) were significant; the 3-way interaction was not (χ215 = 16.2, P = .37).

To better understand these 2-way interactions, we examined models stratified first by occupation and then, separately, by unit size. Among combat arms (χ25 = 119.3, P < .001) and other occupation groups (χ25 = 409.4, P < .001), the number of past-year unit SAs was significantly associated with risk of SA (Table 2). The odds of SA increased as the number of unit SAs increased in both groups (ORs, 1.4-2.3). Figure 1 shows the standardized risk of SA by occupation. Risk of SA in both occupation groups generally increased as a function of past-year unit SAs. Among soldiers in a unit with 5 or more past-year SAs, the standardized risk for those with a combat arms occupation was significantly higher than for those with other occupations (rate ratio [RR], 1.2; 95% CI, 1.1-1.2).

After stratifying by unit size, past-year unit SAs were consistently associated with higher odds of SA for each unit size group (Table 3). This association was particularly strong for smaller units of 1 through 40 soldiers (ORs, 2.1-5.9) compared with units of 41 through 100 (ORs, 1.4-2.2), 101 through 200 (ORs, 1.3-2.0), and 201 through 600 soldiers (ORs, 1.3-2.4). Figure 2 indicates that the SREs generally increased as a function of past-year unit SAs, with risk increasing most substantially for those in smaller units. Among soldiers in a unit with 5 or more past-year SAs, the SRE for those in a unit of 1 through 40 soldiers was significantly higher than the SREs for those in a unit of 41 through 100 (RR, 2.3; 95% CI, 2.2-2.4), 101 through 200 (RR, 2.8; 95% CI, 2.7-2.9), and 201 through 600 soldiers (RR, 2.8; 95% CI, 2.7-2.9). Among units in which there were at least 4 past-year SAs, soldiers in units of 1 through 40 were at least twice as likely to attempt suicide as those in larger units (ie, 41-600 troops; RRs, 2.3-3.6; 95% CI, 2.2-3.7; P < .001).

Discussion

Suicide attempts within an army unit are significant stressors. We found that a soldier’s risk of attempting suicide increased as the number of past-year unit SAs increased, a robust association that persisted even after adjusting for sociodemographic and service-related variables, unit size, previous mental health diagnosis, and past-year deaths due to suicide, combat, and unintentional injury. The association also persisted when examining number of past-year unit suicide attempters (instead of attempts) as the predictor, excluding soldiers in their first 6 months of service, excluding soldiers in Warrior Transition Units, and when adjusting for the number of soldiers within a soldier’s unit with a past-year mental health diagnosis.

Previous research has primarily focused on soldiers’ responses to suicide death, with less attention to responses after unit SAs. It is estimated that for every suicide, many people are profoundly affected.29,30 Nearly half of veterans report knowing someone who died by suicide,31 with 65% of those knowing more than 1 decedent.32 Exposure to suicide death (knowing of, knowing, or identifying with the person who died by suicide33) predicts subsequent suicide ideation and attempts in veterans and civilians,32,34 highlighting the importance of this risk factor among servicemembers. Contrary to previous findings, which relate generally to knowing someone who had died by suicide, unit deaths of any type, including suicide, were not significantly associated with risk of SA in our study. Unit deaths are a more specific and circumscribed group than knowing someone, and the causal mechanisms and responses (eg, of peers and leaders) may differ for unit suicide deaths vs attempts in an active-duty army unit. In addition, there is a lower frequency of unit deaths relative to unit SAs, which may make it harder to detect the influence of deaths in our sample. Intervention efforts that track unit SAs regardless of whether unit members were acquainted with the attempter may help identify at-risk groups.

We found that past-year unit SAs were associated with risk of SA regardless of occupation or unit size. Among combat arms and other occupations, risk of SA was twice as likely among those in units with at least 5 past-year SAs vs units with none. Risk was generally higher for soldiers in smaller vs larger units, regardless of the number of past-year unit SAs. Among soldiers assigned to smaller units (1-40 soldiers), the association of past-year unit SAs with subsequent risk of SA was particularly pronounced.

Unit characteristics that persist over time, including leadership quality, extent of social support and cohesion, and presence of bullying, may influence unit culture and affect the likelihood of and responses to unit SAs. Such characteristics may be more influential in smaller units. For example, unit cohesion has been associated with reduced distress, increased resilience, and positive states of mind.9 To the extent that SAs among fellow soldiers disrupt a soldier’s sense of unit cohesion, our findings suggest that the negative effects may be strongest in smaller units. Similarly, the potential for contagion, particularly within smaller units, is worthy of further study.

Limitations

The current study has several limitations. First, the SA data were from administrative and medical records and are therefore recognized SAs, although they perhaps include the most severe accounts. These records may be subject to diagnostic or coding errors. Second, these data focus exclusively on the 2004 through 2009 period, and findings may not generalize to earlier and later periods of the Iraq and Afghanistan wars or to other US military conflicts. Third, this study did not examine mechanisms of risk. Given the manner in which units were constructed for analysis, unit members may not have been located in the same geographic area or had direct contact with those who attempted suicide. It is important for future studies to explore potential mediators. Fourth, although we examined a range of individual and unit-level covariates, there may be additional important indicators to consider. For example, future studies should examine whether the association of previous unit SAs with subsequent SA risk is modified by other unit variables, including leadership quality and cohesion. Studies should also examine whether the recency of previous unit SAs affects subsequent risk.

Conclusions

Our study indicates that risk of SA among US Army soldiers is influenced by a history of SAs within a soldier’s unit and that higher numbers of unit SAs are related to greater individual suicide risk, particularly in smaller units. Attention to unit characteristics by leadership and service professionals may be a component in SA reduction efforts. Early unit-based postvention consisting of coordinated efforts to provide behavioral, psychosocial, spiritual, and public health support after SAs may be an essential tool in promoting recovery and suicide prevention in servicemembers.35

Back to top
Article Information

Corresponding Author: Robert J. Ursano, MD, Department of Psychiatry, Center for the Study of Traumatic Stress, Uniformed Services University of the Health Sciences, 4301 Jones Bridge Rd, Bethesda, MD 20814 (robert.ursano@usuhs.edu).

Accepted for Publication: May 24, 2017.

Published Online: July 26, 2017. doi:10.1001/jamapsychiatry.2017.1925

Author Contributions: Dr Ursano had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Ursano, Kessler, Naifeh, Herberman Mash, Fullerton, Zaslavsky, Heeringa, Stein.

Acquisition, analysis, or interpretation of data: Ursano, Kessler, Naifeh, Herberman Mash, Bliese, Ng, Aliaga, Wynn, Dinh, McCarroll, Sampson, Kao, Schoenbaum, Stein.

Drafting of the manuscript: Ursano, Naifeh, Herberman Mash, Bliese.

Critical revision of the manuscript for important intellectual content: Ursano, Kessler, Naifeh, Fullerton, Zaslavsky, Ng, Aliaga, Wynn, Dinh, McCarroll, Sampson, Kao, Schoenbaum, Heeringa, Stein.

Statistical analysis: Ursano, Kessler, Bliese, Zaslavsky, Ng, Sampson, Kao, Heeringa.

Obtained funding: Ursano, Kessler, Schoenbaum.

Administrative, technical, or material support: Ursano, Kessler, Naifeh, Herberman Mash, Fullerton, Dinh, Schoenbaum, Heeringa.

Study supervision: Ursano, Kessler.

Group Information: The US Army Study to Assess Risk and Resilience in Servicemembers (STARRS) team consisted of coprincipal investigators Robert J. Ursano, MD (Uniformed Services University of the Health Sciences, Bethesda, Maryland), and Murray B. Stein, MD, MPH (University of California, San Diego and Veterans Affairs San Diego Healthcare System, San Diego); site principal investigators Steven G. Heeringa, PhD (University of Michigan, Ann Arbor), James Wagner, PhD (University of Michigan), and Ronald C. Kessler, PhD (Harvard Medical School, Boston, Massachusetts); and army liaison/consultant Kenneth Cox, MD, MPH (US Army Public Health Center, Aberdeen Proving Ground, Maryland).

Other team members were as follows: Pablo A. Aliaga, MS (Uniformed Services University of the Health Sciences); David M. Benedek, MD (Uniformed Services University of the Health Sciences); Laura Campbell-Sills, PhD (University of California, San Diego); 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); 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); 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); Nancy A. Sampson, BA (Harvard Medical School); CDR Patcho Santiago, MD, MPH (Uniformed Services University of the Health Sciences); Gary H. Wynn, MD (Uniformed Services University of the Health Sciences); and Alan M. Zaslavsky, PhD (Harvard Medical School).

Conflict of Interest Disclosures: Dr Kessler reported receiving support for his epidemiologic studies from Sanofi; working as a consultant for Johnson & Johnson Wellness and Prevention, Shire, and Takeda; and serving on an advisory board for the Johnson & Johnson Services Inc Lake Nona Life Project. Dr Kessler reported being a co-owner of DataStat Inc, a market research firm that performs health care research. Dr Stein reported working as a consultant for Actelion Pharmaceuticals, Healthcare Management Technologies, Janssen Pharmaceuticals, Pfizer, Remedy Therapeutics, Oxeia Biopharmaceuticals, and Tonix Pharmaceuticals. No other disclosures were reported.

Funding/Support: The US Army STARRS was sponsored by the US Department of the Army and funded under cooperative agreement U01MH087981 (2009-2015) with the US Department of Health and Human Services, National Institutes of Health, National Institute of Mental Health. Subsequently, STARRS-Longitudinal Study was sponsored and funded by grant HU0001-15-2-0004 from Uniformed Services University of the Health Sciences from the US Department of Defense.

Role of the Funder/Sponsor: As a cooperative agreement, scientists employed by the National Institute of Mental Health (Lisa Colpe, PhD, MPH, and Michael Schoenbaum, PhD) and army liaisons/consultants (Steven Cersovsky, MD, MPH, and Kenneth Cox, MD, MPH) collaborated to develop the study protocol and data collection instruments, supervise data collection, interpret results, and prepare reports. Although a draft of the manuscript was submitted to the US Department of the Army and the National Institute for Mental Health for review and comment before submission, this was with the understanding that comments would be no more than advisory.

Disclaimer: The contents are solely the responsibility of the authors and do not necessarily represent the views of the US Department of Health and Human Services, the National Institute of Mental Health, the US Department of the Army, or the US Department of Defense.

References
1.
Ursano  RJ, Kessler  RC, Heeringa  SG,  et al; Army STARRS Collaborators.  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):1-21.PubMedGoogle Scholar
2.
Ursano  RJ, Kessler  RC, Stein  MB,  et al; Army STARRS Collaborators.  Risk factors, methods, and timing of suicide attempts among US Army soldiers.  JAMA Psychiatry. 2016;73(7):741-749.PubMedGoogle ScholarCrossref
3.
Ursano  RJ, Kessler  RC, Stein  MB,  et al; Army Study to Assess Risk and Resilience in Servicemembers Collaborators.  Suicide attempts in the US Army during the wars in Afghanistan and Iraq, 2004–2009.  JAMA Psychiatry. 2015;72(9):917-926.PubMedGoogle ScholarCrossref
4.
Bryan  CJ, Griffith  JH, Pace  BT,  et al.  Combat exposure and risk for suicidal thoughts and behaviors among military personnel and veterans: a systematic review and meta-analysis.  Suicide Life Threat Behav. 2015;45(5):633-649.Google ScholarCrossref
5.
Kok  BC, Herrell  RK, Thomas  JL, Hoge  CW.  Posttraumatic stress disorder associated with combat service in Iraq or Afghanistan: reconciling prevalence differences between studies.  J Nerv Ment Dis. 2012;200(5):444-450.PubMedGoogle ScholarCrossref
6.
Englert  DR, Hunter  CL, Sweeney  BJ.  Mental health evaluations of U.S. Air Force basic military training and technical training students.  Mil Med. 2003;168(11):904-910.PubMedGoogle Scholar
7.
Britt  TW, Davison  J, Bliese  PD, Castro  CA.  How leaders can influence the impact that stressors have on soldiers.  Mil Med. 2004;169(7):541-545.PubMedGoogle ScholarCrossref
8.
Oliver  LW, Harman  J, Hoover  E, Hayes  SM, Pandhi  NA.  A quantitative integration of the military cohesion literature.  Mil Psychol. 1999;11(1):57-83.Google ScholarCrossref
9.
Williams  J, Brown  JM, Bray  RM, Anderson Goodell  EM, Olmsted  KR, Adler  AB.  Unit cohesion, resilience, and mental health of soldiers in basic combat training.  Mil Psychol. 2016;28(4):241-250.Google ScholarCrossref
10.
Wright  BK, Kelsall  HL, Sim  MR, Clarke  DM, Creamer  MC.  Support mechanisms and vulnerabilities in relation to PTSD in veterans of the Gulf War, Iraq War, and Afghanistan deployments: a systematic review.  J Trauma Stress. 2013;26(3):310-318.PubMedGoogle ScholarCrossref
11.
Mitchell  MM, Gallaway  MS, Millikan  AM, Bell  M.  Interaction of combat exposure and unit cohesion in predicting suicide-related ideation among post-deployment soldiers.  Suicide Life Threat Behav. 2012;42(5):486-494.PubMedGoogle ScholarCrossref
12.
Bartone  PT, Wright  KM.  Grief and group recovery following a military air disaster.  J Trauma Stress. 1990;3(4):523-539.Google ScholarCrossref
13.
Fullerton  CS, Ursano  RJ, Kao  T-C, Bharitya  VR.  Disaster-related bereavement: acute symptoms and subsequent depression.  Aviat Space Environ Med. 1999;70(9):902-909.PubMedGoogle Scholar
14.
Toblin  RL, Riviere  LA, Thomas  JL, Adler  AB, Kok  BC, Hoge  CW.  Grief and physical health outcomes in U.S. soldiers returning from combat.  J Affect Disord. 2012;136(3):469-475.PubMedGoogle ScholarCrossref
15.
Carr  RB.  When a soldier commits suicide in Iraq: impact on unit and caregivers.  Psychiatry. 2011;74(2):95-106.PubMedGoogle ScholarCrossref
16.
Swanson  SA, Colman  I.  Association between exposure to suicide and suicidality outcomes in youth.  CMAJ. 2013;185(10):870-877.PubMedGoogle ScholarCrossref
17.
Cox  GR, Robinson  J, Williamson  M, Lockley  A, Cheung  YTD, Pirkis  J.  Suicide clusters in young people: evidence for the effectiveness of postvention strategies.  Crisis. 2012;33(4):208-214.PubMedGoogle ScholarCrossref
18.
O’Carroll  PW, Mercy  JA, Steward  JA; Centers for Disease Control (CDC).  CDC recommendations for a community plan for the prevention and containment of suicide clusters.  MMWR Suppl. 1988;37(6):1-12.PubMedGoogle Scholar
19.
Kessler  RC, Stein  MB, Bliese  PD,  et al.  Occupational differences in US Army suicide rates.  Psychol Med. 2015;45(15):3293-3304.PubMedGoogle ScholarCrossref
20.
Henderson  WD. Assessing cohesion in small units. In: Henderson  WD, ed.  Cohesion, the Human Element in Combat: Leadership and Societal Influence in the Armies of the Soviet Union, the United States, North Vietnam, and Israel. Washington, DC: National Defense University Press; 1985:13-26.
21.
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(4):267-275.PubMedGoogle ScholarCrossref
22.
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(6):952-965.PubMedGoogle ScholarCrossref
23.
Gahm  GA, Reger  MA, Kinn  JT, Luxton  DD, Skopp  NA, Bush  NE.  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.PubMedGoogle ScholarCrossref
24.
Walkup  JT, Townsend  L, Crystal  S, Olfson  M.  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.PubMedGoogle ScholarCrossref
25.
Giovannelli  D.  Profile of the U.S. Army 2008: A Reference Handbook. Arlington, VA: Institute of Land Warfare, Association of the United States Army; 2008. https://www.ausa.org/sites/default/files/SR-2008-Profile-of-the-US-Army.pdf. Accessed June 16, 2017.
26.
SAS Institute Inc.  SAS 9.4 Software [computer program]. Cary, NC: SAS Institute Inc; 2013.
27.
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.PubMedGoogle ScholarCrossref
28.
Rothman  K, Greenland  S.  Modern Epidemiology. 2nd ed. Philadelphia, PA: Lippincott Williams & Wilkins; 1998.
29.
Andriessen  K.  Can postvention be prevention?  Crisis. 2009;30(1):43-47.PubMedGoogle ScholarCrossref
30.
Berman  AL.  Estimating the population of survivors of suicide: seeking an evidence base.  Suicide Life Threat Behav. 2011;41(1):110-116.PubMedGoogle ScholarCrossref
31.
Cerel  J, Maple  M, van de Venne  J, Moore  M, Flaherty  C, Brown  M.  Exposure to suicide in the community: prevalence and correlates in one US state.  Public Health Rep. 2016;131(1):100-107.PubMedGoogle ScholarCrossref
32.
Cerel  J, van de Venne  JG, Moore  MM, Maple  MJ, Flaherty  C, Brown  MM.  Veteran exposure to suicide: prevalence and correlates.  J Affect Disord. 2015;179:82-87.PubMedGoogle ScholarCrossref
33.
Cerel  J, McIntosh  JL, Neimeyer  RA, Maple  M, Marshall  D.  The continuum of “survivorship”: definitional issues in the aftermath of suicide.  Suicide Life Threat Behav. 2014;44(6):591-600.PubMedGoogle ScholarCrossref
34.
Bolton  JM, Au  W, Leslie  WD,  et al.  Parents bereaved by offspring suicide: a population-based longitudinal case-control study.  JAMA Psychiatry. 2013;70(2):158-167.PubMedGoogle ScholarCrossref
35.
Erlich  MD. Addressing the aftermath of suicide: why we need postvention. Psychiatric Times. 2016. http://www.psychiatrictimes.com/suicide/addressing-aftermath-suicide-why-we-need-postvention. Accessed January 30, 2017.
×