Risk of suicide-related behaviors is elevated among military personnel transitioning to civilian life. An earlier report showed that high-risk U.S. Army soldiers could be identified shortly before this transition with a machine learning model that included predictors from administrative systems, self-report surveys, and geospatial data. Based on this result, a Veterans Affairs and Army initiative was launched to evaluate a suicide-prevention intervention for high-risk transitioning soldiers. To make targeting practical, though, a streamlined model and risk calculator were needed that used only a short series of self-report survey questions.

We revised the original model in a sample of n = 8335 observations from the Study to Assess Risk and Resilience in Servicemembers-Longitudinal Study (STARRS-LS) who participated in one of three Army STARRS 2011–2014 baseline surveys while in service and in one or more subsequent panel surveys (LS1: 2016–2018, LS2: 2018–2019) after leaving service. We trained ensemble machine learning models with constrained numbers of item-level survey predictors in a 70% training sample. The outcome was self-reported post-transition suicide attempts (SA). The models were validated in the 30% test sample.

Twelve-month post-transition SA prevalence was 1.0% (s.e. = 0.1). The best constrained model, with only 17 predictors, had a test sample ROC-AUC of 0.85 (s.e. = 0.03). The 10–30% of respondents with the highest predicted risk included 44.9–92.5% of 12-month SAs.

An accurate SA risk calculator based on a short self-report survey can target transitioning soldiers shortly before leaving service for intervention to prevent post-transition SA.

Childhood adversities (CAs) predict heightened risks of posttraumatic stress disorder (PTSD) and major depressive episode (MDE) among people exposed to adult traumatic events. Identifying which CAs put individuals at greatest risk for these adverse posttraumatic neuropsychiatric sequelae (APNS) is important for targeting prevention interventions.

Data came from n = 999 patients ages 18–75 presenting to 29 U.S. emergency departments after a motor vehicle collision (MVC) and followed for 3 months, the amount of time traditionally used to define chronic PTSD, in the Advancing Understanding of Recovery After Trauma (AURORA) study. Six CA types were self-reported at baseline: physical abuse, sexual abuse, emotional abuse, physical neglect, emotional neglect and bullying. Both dichotomous measures of ever experiencing each CA type and numeric measures of exposure frequency were included in the analysis. Risk ratios (RRs) of these CA measures as well as complex interactions among these measures were examined as predictors of APNS 3 months post-MVC. APNS was defined as meeting self-reported criteria for either PTSD based on the PTSD Checklist for DSM-5 and/or MDE based on the PROMIS Depression Short-Form 8b. We controlled for pre-MVC lifetime histories of PTSD and MDE. We also examined mediating effects through peritraumatic symptoms assessed in the emergency department and PTSD and MDE assessed in 2-week and 8-week follow-up surveys. Analyses were carried out with robust Poisson regression models.

Most participants (90.9%) reported at least rarely having experienced some CA. Ever experiencing each CA other than emotional neglect was univariably associated with 3-month APNS (RRs = 1.31–1.60). Each CA frequency was also univariably associated with 3-month APNS (RRs = 1.65–2.45). In multivariable models, joint associations of CAs with 3-month APNS were additive, with frequency of emotional abuse (RR = 2.03; 95% CI = 1.43–2.87) and bullying (RR = 1.44; 95% CI = 0.99–2.10) being the strongest predictors. Control variable analyses found that these associations were largely explained by pre-MVC histories of PTSD and MDE.

Although individuals who experience frequent emotional abuse and bullying in childhood have a heightened risk of experiencing APNS after an adult MVC, these associations are largely mediated by prior histories of PTSD and MDE.

Research on post-traumatic stress disorder (PTSD) course finds a substantial proportion of cases remit within 6 months, a majority within 2 years, and a substantial minority persists for many years. Results are inconsistent about pre-trauma predictors.

The WHO World Mental Health surveys assessed lifetime DSM-IV PTSD presence-course after one randomly-selected trauma, allowing retrospective estimates of PTSD duration. Prior traumas, childhood adversities (CAs), and other lifetime DSM-IV mental disorders were examined as predictors using discrete-time person-month survival analysis among the 1575 respondents with lifetime PTSD.

20%, 27%, and 50% of cases recovered within 3, 6, and 24 months and 77% within 10 years (the longest duration allowing stable estimates). Time-related recall bias was found largely for recoveries after 24 months. Recovery was weakly related to most trauma types other than very low [odds-ratio (OR) 0.2–0.3] early-recovery (within 24 months) associated with purposefully injuring/torturing/killing and witnessing atrocities and very low later-recovery (25+ months) associated with being kidnapped. The significant ORs for prior traumas, CAs, and mental disorders were generally inconsistent between early- and later-recovery models. Cross-validated versions of final models nonetheless discriminated significantly between the 50% of respondents with highest and lowest predicted probabilities of both early-recovery (66–55% v. 43%) and later-recovery (75–68% v. 39%).

We found PTSD recovery trajectories similar to those in previous studies. The weak associations of pre-trauma factors with recovery, also consistent with previous studies, presumably are due to stronger influences of post-trauma factors.

Sexual assault is a global concern with post-traumatic stress disorder (PTSD), one of the common sequelae. Early intervention can help prevent PTSD, making identification of those at high risk for the disorder a priority. Lack of representative sampling of both sexual assault survivors and sexual assaults in prior studies might have reduced the ability to develop accurate prediction models for early identification of high-risk sexual assault survivors.

Data come from 12 face-to-face, cross-sectional surveys of community-dwelling adults conducted in 11 countries. Analysis was based on the data from the 411 women from these surveys for whom sexual assault was the randomly selected lifetime traumatic event (TE). Seven classes of predictors were assessed: socio-demographics, characteristics of the assault, the respondent's retrospective perception that she could have prevented the assault, other prior lifetime TEs, exposure to childhood family adversities and prior mental disorders.

Prevalence of Diagnostic and Statistical Manual of Mental Disorders IV (DSM-IV) PTSD associated with randomly selected sexual assaults was 20.2%. PTSD was more common for repeated than single-occurrence victimization and positively associated with prior TEs and childhood adversities. Respondent's perception that she could have prevented the assault interacted with history of mental disorder such that it reduced odds of PTSD, but only among women without prior disorders (odds ratio 0.2, 95% confidence interval 0.1–0.9). The final model estimated that 40.3% of women with PTSD would be found among the 10% with the highest predicted risk.

Whether counterfactual preventability cognitions are adaptive may depend on mental health history. Predictive modelling may be useful in targeting high-risk women for preventive interventions.

Traumatic events are common globally; however, comprehensive population-based cross-national data on the epidemiology of posttraumatic stress disorder (PTSD), the paradigmatic trauma-related mental disorder, are lacking.

Data were analyzed from 26 population surveys in the World Health Organization World Mental Health Surveys. A total of 71 083 respondents ages 18+ participated. The Composite International Diagnostic Interview assessed exposure to traumatic events as well as 30-day, 12-month, and lifetime PTSD. Respondents were also assessed for treatment in the 12 months preceding the survey. Age of onset distributions were examined by country income level. Associations of PTSD were examined with country income, world region, and respondent demographics.

The cross-national lifetime prevalence of PTSD was 3.9% in the total sample and 5.6% among the trauma exposed. Half of respondents with PTSD reported persistent symptoms. Treatment seeking in high-income countries (53.5%) was roughly double that in low-lower middle income (22.8%) and upper-middle income (28.7%) countries. Social disadvantage, including younger age, female sex, being unmarried, being less educated, having lower household income, and being unemployed, was associated with increased risk of lifetime PTSD among the trauma exposed.

PTSD is prevalent cross-nationally, with half of all global cases being persistent. Only half of those with severe PTSD report receiving any treatment and only a minority receive specialty mental health care. Striking disparities in PTSD treatment exist by country income level. Increasing access to effective treatment, especially in low- and middle-income countries, remains critical for reducing the population burden of PTSD.

The U.S. Army uses universal preventives interventions for several negative outcomes (e.g. suicide, violence, sexual assault) with especially high risks in the early years of service. More intensive interventions exist, but would be cost-effective only if targeted at high-risk soldiers. We report results of efforts to develop models for such targeting from self-report surveys administered at the beginning of Army service.

21 832 new soldiers completed a self-administered questionnaire (SAQ) in 2011–2012 and consented to link administrative data to SAQ responses. Penalized regression models were developed for 12 administratively-recorded outcomes occurring by December 2013: suicide attempt, mental hospitalization, positive drug test, traumatic brain injury (TBI), other severe injury, several types of violence perpetration and victimization, demotion, and attrition.

The best-performing models were for TBI (AUC = 0.80), major physical violence perpetration (AUC = 0.78), sexual assault perpetration (AUC = 0.78), and suicide attempt (AUC = 0.74). Although predicted risk scores were significantly correlated across outcomes, prediction was not improved by including risk scores for other outcomes in models. Of particular note: 40.5% of suicide attempts occurred among the 10% of new soldiers with highest predicted risk, 57.2% of male sexual assault perpetrations among the 15% with highest predicted risk, and 35.5% of female sexual assault victimizations among the 10% with highest predicted risk.

Data collected at the beginning of service in self-report surveys could be used to develop risk models that define small proportions of new soldiers accounting for high proportions of negative outcomes over the first few years of service.

Considerable research has documented that exposure to traumatic events has negative effects on physical and mental health. Much less research has examined the predictors of traumatic event exposure. Increased understanding of risk factors for exposure to traumatic events could be of considerable value in targeting preventive interventions and anticipating service needs.

General population surveys in 24 countries with a combined sample of 68 894 adult respondents across six continents assessed exposure to 29 traumatic event types. Differences in prevalence were examined with cross-tabulations. Exploratory factor analysis was conducted to determine whether traumatic event types clustered into interpretable factors. Survival analysis was carried out to examine associations of sociodemographic characteristics and prior traumatic events with subsequent exposure.

Over 70% of respondents reported a traumatic event; 30.5% were exposed to four or more. Five types – witnessing death or serious injury, the unexpected death of a loved one, being mugged, being in a life-threatening automobile accident, and experiencing a life-threatening illness or injury – accounted for over half of all exposures. Exposure varied by country, sociodemographics and history of prior traumatic events. Being married was the most consistent protective factor. Exposure to interpersonal violence had the strongest associations with subsequent traumatic events.

Given the near ubiquity of exposure, limited resources may best be dedicated to those that are more likely to be further exposed such as victims of interpersonal violence. Identifying mechanisms that account for the associations of prior interpersonal violence with subsequent trauma is critical to develop interventions to prevent revictimization.

Although interventions exist to reduce violent crime, optimal implementation requires accurate targeting. We report the results of an attempt to develop an actuarial model using machine learning methods to predict future violent crimes among US Army soldiers.

A consolidated administrative database for all 975 057 soldiers in the US Army in 2004–2009 was created in the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS). Of these soldiers, 5771 committed a first founded major physical violent crime (murder-manslaughter, kidnapping, aggravated arson, aggravated assault, robbery) over that time period. Temporally prior administrative records measuring socio-demographic, Army career, criminal justice, medical/pharmacy, and contextual variables were used to build an actuarial model for these crimes separately among men and women using machine learning methods (cross-validated stepwise regression, random forests, penalized regressions). The model was then validated in an independent 2011–2013 sample.

Key predictors were indicators of disadvantaged social/socioeconomic status, early career stage, prior crime, and mental disorder treatment. Area under the receiver-operating characteristic curve was 0.80–0.82 in 2004–2009 and 0.77 in the 2011–2013 validation sample. Of all administratively recorded crimes, 36.2–33.1% (male-female) were committed by the 5% of soldiers having the highest predicted risk in 2004–2009 and an even higher proportion (50.5%) in the 2011–2013 validation sample.

Although these results suggest that the models could be used to target soldiers at high risk of violent crime perpetration for preventive interventions, final implementation decisions would require further validation and weighing of predicted effectiveness against intervention costs and competing risks.

Civilian suicide rates vary by occupation in ways related to occupational stress exposure. Comparable military research finds suicide rates elevated in combat arms occupations. However, no research has evaluated variation in this pattern by deployment history, the indicator of occupation stress widely considered responsible for the recent rise in the military suicide rate.

The joint associations of Army occupation and deployment history in predicting suicides were analysed in an administrative dataset for the 729 337 male enlisted Regular Army soldiers in the US Army between 2004 and 2009.

There were 496 suicides over the study period (22.4/100 000 person-years). Only two occupational categories, both in combat arms, had significantly elevated suicide rates: infantrymen (37.2/100 000 person-years) and combat engineers (38.2/100 000 person-years). However, the suicide rates in these two categories were significantly lower when currently deployed (30.6/100 000 person-years) than never deployed or previously deployed (41.2–39.1/100 000 person-years), whereas the suicide rate of other soldiers was significantly higher when currently deployed and previously deployed (20.2–22.4/100 000 person-years) than never deployed (14.5/100 000 person-years), resulting in the adjusted suicide rate of infantrymen and combat engineers being most elevated when never deployed [odds ratio (OR) 2.9, 95% confidence interval (CI) 2.1–4.1], less so when previously deployed (OR 1.6, 95% CI 1.1–2.1), and not at all when currently deployed (OR 1.2, 95% CI 0.8–1.8). Adjustment for a differential ‘healthy warrior effect’ cannot explain this variation in the relative suicide rates of never-deployed infantrymen and combat engineers by deployment status.

Efforts are needed to elucidate the causal mechanisms underlying this interaction to guide preventive interventions for soldiers at high suicide risk.

Although variation in the long-term course of major depressive disorder (MDD) is not strongly predicted by existing symptom subtype distinctions, recent research suggests that prediction can be improved by using machine learning methods. However, it is not known whether these distinctions can be refined by added information about co-morbid conditions. The current report presents results on this question.

Data came from 8261 respondents with lifetime DSM-IV MDD in the World Health Organization (WHO) World Mental Health (WMH) Surveys. Outcomes included four retrospectively reported measures of persistence/severity of course (years in episode; years in chronic episodes; hospitalization for MDD; disability due to MDD). Machine learning methods (regression tree analysis; lasso, ridge and elastic net penalized regression) followed by k-means cluster analysis were used to augment previously detected subtypes with information about prior co-morbidity to predict these outcomes.

Predicted values were strongly correlated across outcomes. Cluster analysis of predicted values found three clusters with consistently high, intermediate or low values. The high-risk cluster (32.4% of cases) accounted for 56.6–72.9% of high persistence, high chronicity, hospitalization and disability. This high-risk cluster had both higher sensitivity and likelihood ratio positive (LR+; relative proportions of cases in the high-risk cluster versus other clusters having the adverse outcomes) than in a parallel analysis that excluded measures of co-morbidity as predictors.

Although the results using the retrospective data reported here suggest that useful MDD subtyping distinctions can be made with machine learning and clustering across multiple indicators of illness persistence/severity, replication with prospective data is needed to confirm this preliminary conclusion.

The US Army suicide rate has increased sharply in recent years. Identifying significant predictors of Army suicides in Army and Department of Defense (DoD) administrative records might help focus prevention efforts and guide intervention content. Previous studies of administrative data, although documenting significant predictors, were based on limited samples and models. A career history perspective is used here to develop more textured models.

The analysis was carried out as part of the Historical Administrative Data Study (HADS) of the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS). De-identified data were combined across numerous Army and DoD administrative data systems for all Regular Army soldiers on active duty in 2004–2009. Multivariate associations of sociodemographics and Army career variables with suicide were examined in subgroups defined by time in service, rank and deployment history.

Several novel results were found that could have intervention implications. The most notable of these were significantly elevated suicide rates (69.6–80.0 suicides per 100 000 person-years compared with 18.5 suicides per 100 000 person-years in the total Army) among enlisted soldiers deployed either during their first year of service or with less than expected (based on time in service) junior enlisted rank; a substantially greater rise in suicide among women than men during deployment; and a protective effect of marriage against suicide only during deployment.

A career history approach produces several actionable insights missed in less textured analyses of administrative data predictors. Expansion of analyses to a richer set of predictors might help refine understanding of intervention implications.

Research on the structure of co-morbidity among common mental disorders has largely focused on current prevalence rather than on the development of co-morbidity. This report presents preliminary results of the latter type of analysis based on the US National Comorbidity Survey Replication Adolescent Supplement (NCS-A).

A national survey was carried out of adolescent mental disorders. DSM-IV diagnoses were based on the Composite International Diagnostic Interview (CIDI) administered to adolescents and questionnaires self-administered to parents. Factor analysis examined co-morbidity among 15 lifetime DSM-IV disorders. Discrete-time survival analysis was used to predict first onset of each disorder from information about prior history of the other 14 disorders.

Factor analysis found four factors representing fear, distress, behavior and substance disorders. Associations of temporally primary disorders with the subsequent onset of other disorders, dated using retrospective age-of-onset (AOO) reports, were almost entirely positive. Within-class associations (e.g. distress disorders predicting subsequent onset of other distress disorders) were more consistently significant (63.2%) than between-class associations (33.0%). Strength of associations decreased as co-morbidity among disorders increased. The percentage of lifetime disorders explained (in a predictive rather than a causal sense) by temporally prior disorders was in the range 3.7–6.9% for earliest-onset disorders [specific phobia and attention deficit hyperactivity disorder (ADHD)] and much higher (23.1–64.3%) for later-onset disorders. Fear disorders were the strongest predictors of most other subsequent disorders.

Adolescent mental disorders are highly co-morbid. The strong associations of temporally primary fear disorders with many other later-onset disorders suggest that fear disorders might be promising targets for early interventions.

Interventions to treat mental disorders after natural disasters are important both for humanitarian reasons and also for successful post-disaster physical reconstruction that depends on the psychological functioning of the affected population. A major difficulty in developing such interventions, however, is that large between-disaster variation exists in the prevalence of post-disaster mental disorders, making it difficult to estimate need for services in designing interventions without carrying out a post-disaster mental health needs assessment survey. One of the daunting methodological challenges in implementing such surveys is that secondary stressors unique to the disaster often need to be discovered to understand the magnitude, type, and population segments most affected by post-disaster mental disorders.

This problem is examined in the current commentary by analyzing data from the WHO World Mental Health (WMH) Surveys. We analyze the extent to which people exposed to natural disasters throughout the world also experienced secondary stressors and the extent to which the mental disorders associated with disasters were more proximally due to these secondary stressors than to the disasters themselves.

Lifetime exposure to natural disasters was found to be high across countries (4.4–7.5%). 10.7–11.4% of those exposed to natural disasters reported the occurrence of other related stressors (e.g. death of a loved one and destruction of property). A monotonic relationship was found between the number of additional stressors and the subsequent onset of mental disorders

These results document the importance of secondary stressors in accounting for the effects of natural disasters on mental disorders. Implications for intervention planning are discussed.

Mental disorders may increase the risk of physical violence among married couples.

To estimate associations between premarital mental disorders and marital violence in a cross-national sample of married couples.

A total of 1821 married couples (3642 individuals) from 11 countries were interviewed as part of the World Health Organization's World Mental Health Survey Initiative. Sixteen mental disorders with onset prior to marriage were examined as predictors of marital violence reported by either spouse.

Any physical violence was reported by one or both spouses in 20% of couples, and was associated with husbands' externalising disorders (OR = 1.7, 95% CI 1.2–2.3). Overall, the population attributable risk for marital violence related to premarital mental disorders was estimated to be 17.2%.

Husbands' externalising disorders had a modest but consistent association with marital violence across diverse countries. This finding has implications for the development of targeted interventions to reduce risk of marital violence.

Burden-of-illness data, which are often used in setting healthcare policy-spending priorities, are unavailable for mental disorders in most countries.

To examine one central aspect of illness burden, the association of serious mental illness with earnings, in the World Health Organization (WHO) World Mental Health (WMH) Surveys.

The WMH Surveys were carried out in 10 high-income and 9 low- and middle-income countries. The associations of personal earnings with serious mental illness were estimated.

Respondents with serious mental illness earned on average a third less than median earnings, with no significant between-country differences (χ2(9) = 5.5–8.1, P = 0.52–0.79). These losses are equivalent to 0.3–0.8% of total national earnings. Reduced earnings among those with earnings and the increased probability of not earning are both important components of these associations.

These results add to a growing body of evidence that mental disorders have high societal costs. Decisions about healthcare resource allocation should take these costs into consideration.