Abstract
Introduction:
Few studies have examined characteristics distinguishing Veteran and civilian suicide decedents. An understanding of unique risk factors for Veteran suicide is critical to develop effective preventive interventions. This is particularly imperative for female Veterans, who have near double the suicide mortality rate of same-aged female civilians. The objectives of this study were to examine whether Veteran and civilian suicide decedents differed on risk factors and suicide-event characteristics, and to determine whether predictors changed based on sex.
Methods:
Data from 116,515 suicides collected by the National Violent Death Reporting System in 27 states between 2003 and 2015 were analyzed in 2018 in sex-stratified analyses. Logistic regression models examined population differences in risk factors and suicide-event characteristics.
Results:
Relative to male civilians, male Veterans were more likely to have a contributing physical health problem (AOR=1.10, 95% CI=1.06, 1.14) and to use a firearm for their suicide (AOR=1.41, 95% CI=1.36, 1.47); they were less likely to have substance use problems (AOR=0.70, 95% CI=0.66, 0.75), depressed mood (AOR=0.93, 95% CI=0.90, 0.97), or financial problems (AOR=0.91, 95% CI=0.86, 0.97). Female Veterans were more likely to use a firearm for their suicide (AOR=1.39, 95% CI=1.19, 1.63) relative to female civilians.
Conclusions:
Firearm use as a suicide method was a key distinguishing feature of Veteran suicide. Means restriction and firearm safety are pertinent to preventing Veteran suicide. Given low utilization of mental health care and frequent presence of physical health problems in this population, safe storage messages may have a greater preventative impact if delivered in primary care or other non-psychiatric settings.
INTRODUCTION
Veteran suicide is a pressing concern. In 2016, the age- and sex-adjusted suicide mortality rates for Veterans were 1.5 times higher than civilians, with female Veterans having nearly double (1.8 times) the suicide mortality rate of same-aged civilian females.1 Although previous studies have identified important risk factors for Veteran suicide,2,3 many of these studies have focused on Veterans Health Administration samples. Yet, among the 20 Veterans who die by suicide each day, approximately 14 have not received Veterans Health Administration services/care in the past year. Studies utilizing population-based samples are of great importance for better understanding Veteran suicide, yet very few exist. One population-based study found that firearm use as a suicide method distinguished male Veterans from matched civilian suicide decedents.4 Another study indicated female Veterans were also more likely to use firearms as a suicide method relative to female civilians, but utilized a small sample.5 Very little is known regarding the risk factors that differentiate female Veteran and female civilian suicide decedents,6 despite a growing female population in the military. Additional research on the precipitants and characteristics of suicide deaths is greatly needed for both male and female Veterans.
METHODS
Data came from the Restricted Access Database from the National Violent Death Reporting System (NVDRS) administered by the Centers for Disease Control and Prevention.7 The study was restricted to the secondary analysis of a fully de-identified dataset and IRB approval was not required. The NVDRS collects data from participating states regarding violent deaths, including suicides, and provides de- identified information obtained from death certificates, coroner/medical examiner reports, law enforcement reports, secondary sources, and abstractor reports. The NVDRS can reliably estimate Veteran suicide8 and this study included 116,515 suicide decedents with known military status and known circumstances for their suicide, occurring between 2003 and 2015 in 27 states. Veteran status was determined from national standardized death certificates in the section “Ever a member of U.S. Armed Forces.”
Measures
Demographic variables included age, race, ethnicity, residence (Northeast, Midwest, South, and West), education level, marital status, and year of death.
The following health and mental health variables were coded as yes or no/unknown: diagnosed mental health problem, depressed mood, alcohol dependence/problem, non-alcohol substance use dependence/problem, suicide attempt history, past mental health treatment, and contributing physical health problems.
The following social and occupational variables were coded as yes or no/unknown: problems with intimate partner, family member, or friend/associate (collapsed into single interpersonal problems category); problems with job/employment; criminal/civil legal issues; and recent death of family/friend.
Two suicide event variables coded as yes or no/unknown were: victim left suicide note/communication and recent disclosure of suicidal thoughts/plan. Location of suicide was dichotomized to residence and other. Suicide methods were grouped into: firearms, overdose/poisoning, hanging/strangulation/suffocation, and other.
Statistical Analysis
Data were analyzed utilizing SPSS, version 22.0 and Stata, version 15 in 2018. Chi-square and t-test analyses examined sex-stratified differences between Veterans and non-Veterans for demographic variables. A series of logistic regression models examined Veteran status and risk variables without adjustments (Model 1) and with demographic adjustments for age, sex, race, ethnicity, education level, marital status, year of death, and residence (Model 2). Propensity score-based Marginal Mean Weighting with Stratification9 was then used to balance Veterans and non-Veterans on the demographic variables outlined above and health/mental health, social/occupational, and suicide event/method variables (Model 3). Given this study’s interest in outcomes among males and females, separate analyses were conducted by sex. The adaptive False Discovery Rate method,10 which uses an adaptive stepwise procedure for determining significance for multiple comparisons, was used due to exploratory nature of these analyses.
RESULTS
Demographic characteristics and differences based on Veteran status appear in Table 1. Endorsement percentages and unadjusted/adjusted ORs for the health/mental health, social/occupational, and suicide event/method variables are presented for males in Table 2 and females in Table 3. In Model 3, male Veterans had greater adjusted odds of using a firearm as the suicide method, having a contributing physical health problem, and leaving a suicide note than male civilians. Male Veterans had lower adjusted odds of having substance use problems, depressed mood, financial problems, a recent death of a loved one, or disclosure of suicidal intent in the month prior to their death. Among females, the only significant difference in Model 3 was greater adjusted odds of using a firearm as the method for suicide for Veterans.
Table 1.
Demographics of Suicide Decedents From 2003–2015 Collected by the NVDRS
| Variablea | M civilian n=67,716 |
M Veteran n=22,707 |
F civilian n=25,251 |
F Veteran n=841 |
|---|---|---|---|---|
| Age, mean (SD) | 43.21 (23.4) | b58.84 (22.2) | c46.84 (21.4) | 44.69 (13.6) |
| Race/Ethnicity, % | ||||
| White, non-Hispanic | 82.6 | d90.1 | e86.1 | 81.1 |
| Black, non-Hispanic | 6.7 | 5.3 | 4.6 | 9.8 |
| Hispanic | 5.6 | 2.4 | 3.9 | 4.2 |
| Other | 5.1 | 2.2 | 5.4 | 4.9 |
| Education, % | ||||
| Less than high school | f19.2 | 10.1 | g12.5 | 3.6 |
| High school diploma/GED | 43.2 | 46.0 | 37.4 | 31.4 |
| Some college/Associates | 21.9 | 25.9 | 28.9 | 40.6 |
| Bachelor’s or higher | 15.7 | 18.0 | 21.2 | 24.4 |
| Marital status, % | ||||
| Single/Never married | h40.7 | 15.3 | 25.3 | 22.7 |
| Divorced/Separated | 22.3 | 26.5 | 30.1 | i35.3 |
| Married/Partnered | 33.3 | 46.0 | 35.9 | 37.8 |
| Widowed | 3.5 | 12.2 | 8.7 | 4.2 |
| Region, % | ||||
| Northeast | j16.8 | 14.6 | k16.0 | 9.9 |
| Midwest | 16.9 | 14.3 | 15.5 | 12.6 |
| South | 41.5 | 44.5 | 42.7 | 46.4 |
| West | 24.8 | 26.7 | 25.8 | 31.2 |
Note: Boldface indicates statistical significance (p<0.01) for t-tests and chi-square analyses examining post-hoc adjusted residuals of individual cells.
Tests were stratified by sex (male civilians to male Veterans; female civilians to female Veterans).
t(40848)=90.4, p<0.001.
t(983)=4.4, p<0.001.
χ2 (3)=847.4, p<0.001.
χ2 (3)=47.7, p<0.001.
χ2 (3)=692.1, p<0.001.
χ2 (3)=65.6, p<0.001.
χ2 (3)=6307.8, p<0.001.
χ2 (3)=32.2, p<0.001.
χ2 (3)=179.6, p<0.001.
χ2 (3)=35.2, p<0.001.
NVDRS, National Violent Death Reporting System; M, male; F, female.
Table 2.
ORs for Precipitants and Characteristics of Suicide: Comparisons of Male Veterans and Civilians
| Characteristic | M civilian n=67,716 % |
M Veteran n=22,707 % |
Model 1a,b
OR (95% CI) |
Model 2a,c
AOR (95% CI) |
Model 3a,d
AOR (95% CI) |
|---|---|---|---|---|---|
| Health/MH | |||||
| Current MH problem | 40.4 | 38.7 | 0.89 (0.87, 0.92) | 0.93 (0.90, 0.96) | 1.0 (0.97, 1.04) |
| Depressed mood | 40.4 | 41.1 | 1.02 (1.00, 1.06) | 0.90 (0.87, 0.93) | 0.93 (0.90, 0.97) |
| Alcohol problem | 20.3 | 15.7 | 0.73 (0.70, 0.76) | 0.80 (0.77, 0.84) | 0.91 (0.87, 0.96) |
| Other substance problem | 17.1 | 7.6 | 0.40 (0.38, 0.42) | 0.65 (0.61, 0.69) | 0.70 (0.66, 0.75) |
| Suicide attempt history | 17.5 | 12.7 | 0.69 (0.66, 0.72) | 0.91 (0.87, 0.95) | 0.97 (0.92, 1.03) |
| Treatment for MH (ever) | 34.6 | 31.3 | 0.86 (0.83, 0.90) | 0.90 (0.87, 0.94) | 0.97 (0.93, 1.00) |
| Physical health problem | 17.2 | 36.8 | 2.79 (2.70, 2.88) | 1.06 (1.02, 1.11) | 1.10 (1.06, 1.14) |
| Social/Occupational | |||||
| Interpersonal problem | 41.6 | 30.6 | 0.62 (0.60, 0.64) | 0.95 (0.91, 0.98) | 0.99 (0.95, 1.02) |
| Job problem | 15.0 | 11.1 | 0.71 (0.68, 0.74) | 0.82 (0.78, 0.86) | 1.02 (0.97, 1.08) |
| Financial problem | 12.7 | 10.9 | 0.84 (0.80, 0.88) | 0.79 (0.75, 0.83) | 0.91 (0.86, 0.97) |
| Legal problem | 15.8 | 10.7 | 0.64 (0.61, 0.67) | 0.85 (0.81, 0.90) | 0.95 (0.89, 1.00) |
| Recent loss | 7.4 | 8.7 | 1.20 (1.14, 1.27) | 0.87 (0.82, 0.93) | 0.89 (0.83, 0.95) |
| Suicide event | |||||
| Location (home) | 73.3 | 78.8 | 1.35 (1.30, 1.40) | 1.05 (1.01, 1.09) | 0.99 (0.95, 1.03) |
| Suicide note | 31.6 | 34.5 | 1.14 (1.10, 1.18) | 1.07 (1.04, 1.12) | 1.09 (1.05, 1.14) |
| Suicide intent disclosed | 28.2 | 26.7 | 0.93 (0.89, 0.96) | 0.94 (0.91, 0.98) | 0.95 (0.91, 0.99) |
| Suicide method | |||||
| Firearm | 53.3 | 70.4 | 2.08 (2.02, 2.15) | 1.47 (1.44, 1.55) | 1.41 (1.36, 1.47) |
| Hanging/Suffocation | 28.3 | 14.7 | 0.44 (0.42, 0.45) | 0.68 (0.65, 0.72) | 0.71 (0.67, 0.74) |
| Poison/Overdose | 11.7 | 10.4 | 0.87 (0.83, 0.92) | 0.83 (0.78, 0.87) | 0.96 (0.91, 1.02) |
| Other method | 6.7 | 4.5 | 0.65 (0.62, 0.71) | 0.73 (0.67, 0.79) | 0.74 (0.68, 0.80) |
Note: Boldface indicates statistical significance (p<0.05) for Veteran status following adjustment for multiple comparisons using the false discovery rate procedure, with civilians coded as the reference category.
Sample sizes for each analysis differed as information on all covariates and outcomes were not available for all participants. Missing data frequencies did not exceed 5% of initial sample utilized for this study.
Model 1 adjusted for no covariates.
Model 2 adjusted for demographic factors including age, race, ethnicity, education level, marital status, year of death, and region.
Model 3 adjusted for demographic variables outlined in Model 2 as well as health/mental health, social/occupational, and suicide event/method variables using propensity score-based marginal means weighting with stratification.
M, male; MH, mental health.
Table 3.
ORs for Precipitants and Characteristics of Suicide: Comparisons of Female Veterans and Civilians
| Characteristic | F civilian n=25,251 % |
F Veteran n=841 % |
Model 1a,b
OR (95% CI) |
Model 2a,c
AOR (95% CI) |
Model 3a,d
AOR (95% CI) |
|---|---|---|---|---|---|
| Health/MH | |||||
| Current MH problem | 63.4 | 61.5 | 0.92 (0.80, 1.06) | 0.90 (0.77, 1.03) | 0.99 (0.85, 1.17) |
| Depressed mood | 42.4 | 43.9 | 1.06 (0.92, 1.22) | 0.99 (0.85, 1.14) | 1.00 (0.85, 1.17) |
| Alcohol problem | 15.7 | 14.3 | 0.91 (0.75, 1.11) | 0.87 (0.71, 1.07) | 0.85 (0.68, 1.07) |
| Other substance problem | 18.2 | 14.6 | 0.77 (0.64, 0.94) | 0.76 (0.62, 0.93) | 0.79 (0.63, 0.99) |
| Suicide attempt history | 33.7 | 33.3 | 0.98 (0.85, 1.14) | 0.93 (0.80, 1.09) | 1.04 (0.88, 1.23) |
| Treatment for MH (ever) | 55.7 | 53.0 | 0.90 (0.78, 1.03) | 0.85 (0.74, 0.98) | 0.93 (0.79, 1.09) |
| Physical health problem | 23.3 | 20.6 | 0.85 (0.72, 1.01) | 0.88 (0.73, 1.05) | 1.04 (0.85, 1.28) |
| Social/Occupational | |||||
| Interpersonal problem | 35.9 | 39.1 | 1.14 (1.00, 1.32) | 1.02 (0.87, 1.18) | 0.93 (0.79, 1.09) |
| Job problem | 8.5 | 10.9 | 1.32 (1.06, 1.65) | 1.17 (0.93, 1.47) | 1.09 (0.85, 1.39) |
| Financial problem | 9.7 | 9.6 | 0.99 (0.78, 1.25) | 0.91 (0.71, 1.15) | 0.90 (0.69, 1.18) |
| Legal problem | 7.6 | 7.3 | 0.95 (0.73, 1.24) | 0.89 (0.68, 1.17) | 0.85 (0.64, 1.15) |
| Recent loss | 9.7 | 8.1 | 0.82 (0.64, 1.05) | 0.90 (0.70, 1.18) | 0.98 (0.72, 1.32) |
| Suicide event | |||||
| Location (Home) | 81.6 | 80.0 | 0.90 (0.76, 1.07) | 0.94 (0.79, 1.12) | 0.95 (0.74, 1.08) |
| Suicide note | 39.9 | 43.6 | 1.17 (1.02, 1.34) | 1.17 (1.02, 1.35) | 1.12 (0.96, 1.31) |
| Suicide intent disclosed | 27.0 | 28.9 | 1.10 (0.95, 1.28) | 1.04 (0.89, 1.21) | 1.11 (0.94, 1.32) |
| Suicide method | |||||
| Firearm | 31.4 | 41.0 | 1.52 (1.32, 1.75) | 1.44 (1.24, 1.67) | 1.39 (1.19, 1.63) |
| Hanging/Suffocation | 22.4 | 15.8 | 0.65 (0.54, 0.79) | 0.68 (0.56, 0.83) | 0.69 (0.56, 0.86) |
| Poison/Overdose | 38.4 | 36.5 | 0.92 (0.80, 1.07) | 0.94 (0.89, 1.10) | 0.96 (0.82, 1.13) |
| Other method | 7.9 | 6.7 | 0.84 (0.63, 1.10) | 0.87 (0.65, 1.15) | 0.86 (0.64, 1.17) |
Note: Boldface indicates statistical significance (p<0.05) for Veteran status following adjustment for multiple comparisons using the false discovery rate procedure, with civilians coded as the reference category.
Sample sizes for each analysis differed as information on all covariates and outcomes were not available for all participants. Missing data frequencies did not exceed 5% of initial sample utilized for this study.
Model 1 adjusted for no covariates.
Model 2 adjusted for demographic factors including age, race, ethnicity, education level, marital status, year of death, and region.
Model 3 adjusted for demographic variables outlined in Model 2 as well as health/mental health, social/occupational, and suicide event/method variables using propensity score-based marginal means weighting with stratification.
F, female; MH, mental health.
DISCUSSION
Consistent with past reports,4,5 the current findings showed that male Veterans were more likely to use firearms as the method of suicide compared with civilians. Furthermore, male Veterans were more likely to report contributing physical health problems. Although this may be a byproduct of greater overall prevalence of medical morbidity in the Veteran population,11 these findings suggest that suicide prevention efforts in physical health medical settings (e.g., primary care, pain clinics) are especially indicated. Surprisingly, male Veterans were less likely than male civilians to have had depressed mood, alcohol/substance use, or financial problems. This, combined with the low utilization of mental health care among suicide decedents, suggests that screening and prevention efforts in mental health settings is insufficient.
Female Veteran suicide decedents only differed from female civilians on one suicide characteristic—use of firearm as suicide method. The interpersonal theory of suicide stipulates that in order to die by suicide, one must have both desire and acquired capability to die by suicide.12 This finding suggests that differences in factors associated with acquired capability (e.g., exposure/access/comfort with firearms), rather than factors associated with suicidal desire (e.g., depressed mood, interpersonal problems), drives higher suicide mortality rates among Veterans compared with civilians for both males and females. As such, an increased focus on factors leading to greater capability for suicide and access to lethal means is especially warranted for Veterans.
Limitations
Data in the NVDRS only represent what was explicitly documented in death certificates, medical/coroner reports, and law enforcement reports, thus the circumstances and contributing stressors to suicide deaths are likely underreported. The U.S. states participating in the NVDRS had varying lengths of time of participation and are not nationally representative. The NVDRS does not distinguish active duty personnel from Veterans, and does not include military-related variables such as branch, service era, or deployments. Most variables were dichotomously coded, limiting ability to assess levels of severity.
Despite being the largest known study of female suicide decedents, there was limited power to detect differences for low-frequency variables (e.g., legal problems, recent loss) for female Veterans. Although suicide is a leading cause of death, base suicide mortality rates are quite low, making it difficult to predict. Statistical methods were used to balance possible confounds, yet unobserved confounders likely existed, which precludes the ability to make causal claims regarding impact of Veteran status on suicide-event variables.
CONCLUSIONS
Ownership of personal firearms in Veterans is greater than in the civilian population13 and presence of a gun in the home increases the likelihood of suicide.14 A growing body of evidence suggests means restriction is an effective method for reducing suicide,15 thus education regarding safe storage within the Veteran population is needed. Although the Veterans Administration has engaged in initiatives around this area, an expansion of reach and services is needed to capture Veterans not utilizing Veterans Administration care. Given the low utilization of mental health care and frequent presence of physical health problems in this population, safe storage messages may have a greater preventive impact if delivered in primary care or other non-psychiatric settings.
ACKNOWLEDGMENTS
This research uses data from National Violent Death Reporting System, a surveillance system designed by the Centers for Disease Control and Prevention’s (CDC) National Center for Injury Prevention and Control. The findings are based, in part, on the contributions of the 42 funded states and territories that collected violent death data and the contributions of the states’ partners, including law enforcement, vital records, medical examiners/coroners, and crime laboratories. The analyses, results, and conclusions presented here represent those of the authors and not necessarily those of the CDC. Dr. Adam Horwitz received a National Violent Death Reporting System Young Investigator Award from the American Public Health Association and CDC to support this research. Dr. Dale Smith has no conflicts of interest to report. Dr. Philip Held receives funding from the Boeing Company and the American Psychological Association. Dr. Zalta is supported by a career development award from the National Institute of Mental Health (K23-MH103394). The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the views of the funders, who had no role in the design, analysis, interpretation, or preparation of this manuscript.
Dr. Horwitz initiated the design and conceptualization of the project and led the writing of the manuscript. Dr. Smith led the statistical analyses and contributed to the method and results sections of the manuscript. Dr. Held aided with the conceptualization of the project and contributed to the introduction and discussion sections. Dr. Zalta oversaw the design and conceptualization of the study and contributed to the introduction and discussion sections. The findings of this article are not being considered and have not been published elsewhere. No financial disclosures were reported by the authors of this paper.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
REFERENCES
- 1.Department of Veterans Affairs Office of Mental Health and Suicide Prevention. Veteran Suicide Data Report, 2005–2016 www.mentalhealth.va.gov/docs/data-sheets/OMHSP_National_Suicide_Data_Report_2005-2016_508.pdf. Published September 2018. Accessed December 6, 2018.
- 2.Conner KR, Bohnert AS, McCarthy JF, et al. Mental disorder comorbidity and suicide among 2.96 million men receiving care in the Veterans Health Administration health system. J Abnorm Psychol 2013;122(1):256–263. 10.1037/a0030163. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Pigeon WR, Pinquart M, Conner K. Meta-analysis of sleep disturbance and suicidal thoughts and behaviors. J Clin Psychiatry 2012;73(9):e1160–e1167. 10.4088/JCP.11r07586. [DOI] [PubMed] [Google Scholar]
- 4.Kaplan MS, Huguet N, McFarland BH, Newsom JT. Suicide among male Veterans: a prospective population-based study. J Epidemiol Community Health 2007;61(7):619–624. 10.1136/jech.2006.054346. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Kaplan MS, McFarland BH, Huguet N. Firearm suicide among Veterans in the general population: findings from the National Violent Death Reporting System. J Trauma 2009;67(3):503–507. 10.1097/TA.0b013e3181b36521. [DOI] [PubMed] [Google Scholar]
- 6.McFarland BH, Kaplan MS, Huguet N. Datapoints: self-inflicted deaths among women with U.S. military service: a hidden epidemic? Psychiatr Serv 2010;61(12):1177 10.1176/ps.2010.61.12.1177. [DOI] [PubMed] [Google Scholar]
- 7.CDC. National Violent Death Reporting System (NVDRS) Coding Manual Revised. National Center for Injury Prevention, CDC www.cdc.gov/violenceprevention/pdf/NVDRS-WebCodingManual.pdf. Published 2016. Accessed December 6, 2018.
- 8.Huguet N, Kaplan MS, McFarland BH. The effects of misclassification biases on Veteran suicide rate estimates. Am J Public Health 2014;104(1):151–155. 10.2105/AJPH.2013.301450. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Hong G Marginal mean weighting through stratification: a generalized method for evaluating multivalued and multiple treatments with nonexperimental data. Psychol Methods 2012;17(1):44–60. 10.1037/a0024918. [DOI] [PubMed] [Google Scholar]
- 10.Benjamini Y, Hochberg Y. On the adaptive control of the false discovery rate in multiple testing with independent statistics. J Educ Behav Stat 2000;25(1):60–83. 10.3102/10769986025001060. [DOI] [Google Scholar]
- 11.Hoerster KD, Lehavot K, Simpson T, McFall M, Reiber G, Nelson KM. Health and health behavior differences: U.S. military, veteran, and civilian men. Am J Prev Med 2012;43(5):483–489. 10.1016/j.amepre.2012.07.029. [DOI] [PubMed] [Google Scholar]
- 12.Joiner T Why People Die by Suicide Harvard University Press; 2005. [Google Scholar]
- 13.Hepburn L, Miller M, Azrael D, Hemenway D. The U.S. gun stock: results from the 2004 National Firearms Survey. Inj Prev 2007;13(1):15–19. 10.1136/ip.2006.013607. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Miller M, Lippmann SJ, Azrael D, Hemenway D. Household firearm ownership and rates of suicide across the 50 United States. J Trauma 2007;62(4):1029–1034. 10.1097/01.ta.0000198214.24056.40. [DOI] [PubMed] [Google Scholar]
- 15.Zalsman G, Hawton K, Wasserman D, et al. Suicide prevention strategies revisited: 10-year systematic review. Lancet Psychiatry 2016;3(7):646–659. 10.1016/S2215-0366(16)30030-X. [DOI] [PubMed] [Google Scholar]