Search

Machine learning is increasingly used in mental health research and has the potential to advance our understanding of how to characterize, predict, and treat mental disorders and associated adverse health outcomes (e.g., suicidal behavior). Machine learning offers new tools to overcome challenges for which traditional statistical methods are not well-suited. This paper provides an overview of machine learning with a specific focus on supervised learning (i.e., methods that are designed to predict or classify an outcome of interest). Several common supervised learning methods are described, along with applied examples from the published literature. We also provide an overview of supervised learning model building, validation, and performance evaluation. Finally, challenges in creating robust and generalizable machine learning algorithms are discussed. [ABSTRACT FROM AUTHOR]

Accidental death is a leading cause of mortality among military members and Veterans; however, knowledge is limited regarding time-dependent risk following deployment and if there are differences by type of accidental death. Longitudinal cohort study (N = 860,930) of soldiers returning from Afghanistan/Iraq deployments in fiscal years 2008–2014. Accidental deaths (i.e., motor vehicle accidents [MVA], accidental overdose, other accidental deaths), were identified through 2018. Crude and age-adjusted mortality rates, rate ratios, time-dependent hazard rates and trends postdeployment were compared across demographic and military characteristics. During the postdeployment observation period, over one-third of deaths were accidental; most were MVA (46.0 %) or overdoses (37.9 %). Across accidental mortality categories (all, MVA, overdose), younger soldiers (18–24, 25–29) were at higher risk compared to older soldiers (40+), and females at lower risk than males. MVA death rates were highest immediately postdeployment, with a significant decreasing hazard rate over time (annual percent change [APC]: –6.5 %). Conversely, accidental overdose death rates were lowest immediately following deployment, with a significant increasing hazard rate over time (APC: 9.9 %). Observed divergent trends in risk for the most common types of accidental deaths provide essential information to inform prevention and intervention planning for the immediate postdeployment transition and long-term. • Accidental deaths were the most common type of postdeployment mortality. • Accidental deaths were mostly motor vehicle accidents or overdose deaths. • Motor vehicle accident death rates were highest immediately postdeployment. • Accidental overdose death rates were lowest immediately postdeployment. • Hazard rates for types of accidental deaths had divergent trends over time. [ABSTRACT FROM AUTHOR]

Background: Epidemiologic studies have reported substance abuse and posttraumatic stress disorder (PTSD) diagnoses as risk factors for suicide among Veterans Health Administration (VHA) patients. Research on risk factors for suicide may not generalize to our understanding of non-fatal intentional self-harm (ISH), given the evidence that these outcomes have unique risk factors. The aims of this study were to examine (1) gender-stratified rates of non-fatal ISH in VHA patients with alcohol abuse/dependence, drug abuse/dependence, and PTSD and (2) gender-stratified interaction between alcohol abuse and dependence and drug abuse and dependence and PTSD in predicting non-fatal ISH.Methods: Participants include all VHA care users who received a PTSD diagnosis in Massachusetts from 2000 to 2008 (n=16,004) and an age- and gender-matched comparison group (n=52,502). Data were obtained from the VHA administrative registries.Results: We found evidence of stronger interactions between substance abuse diagnoses and PTSD in predicting non-fatal ISH for females than for males. The interaction contrast (IC) for alcohol abuse and dependence and PTSD in predicting non-fatal ISH among female VHA patients was 62.35/100,000 person-years; for male VHA patients the comparable IC was 21.49/100,000 person-years. For female VHA patients the IC for drug abuse and dependence and PTSD predicting ISH was 256.33/100,000 person-years; no interaction was observed for male VHA patients.Conclusions: This study contributes to the scant literature on gender differences in substance abuse and PTSD among VHA patients. The findings highlight comorbid diagnoses as particularly important risk factors for non-fatal ISH among female VHA patients. [ABSTRACT FROM AUTHOR]

Military personnel deployed to war zones are at increased risk of mental health problems such as posttraumatic stress disorder (PTSD) or depression. Early pre- or post-deployment identification of those at highest risk of such problems is crucial to target intervention to those in need. However, sufficiently accurate models predicting objectively assessed mental health outcomes have not been put forward. In a sample consisting of all Danish military personnel who deployed to war zones for the first (N = 27,594), second (N = 11,083) and third (N = 5,161) time between 1992 and 2013, we apply neural networks to predict psychiatric diagnoses or use of psychotropic medicine in the years following deployment. Models are based on pre-deployment registry data alone or on pre-deployment registry data in combination with post-deployment questionnaire data on deployment experiences or early post-deployment reactions. Further, we identified the most central predictors of importance for the first, second, and third deployment. Models based on pre-deployment registry data alone had lower accuracy (AUCs ranging from 0.61 (third deployment) to 0.67 (first deployment)) than models including pre- and post-deployment data (AUCs ranging from 0.70 (third deployment) to 0.74 (first deployment)). Age at deployment, deployment year and previous physical trauma were important across deployments. Post-deployment predictors varied across deployments but included deployment exposures as well as early post-deployment symptoms. The results suggest that neural network models combining pre- and early post-deployment data can be utilized for screening tools that identify individuals at risk of severe mental health problems in the years following military deployment. [ABSTRACT FROM AUTHOR]

Depression is associated with incident coronary heart disease (CHD) via a pathway that may be causal, but the mechanisms underlying this association are unclear. We assessed the extent to which metabolic syndrome (MetS) and its components (i.e., elevated waist circumference, low high-density lipoprotein [HDL] cholesterol, elevated triglycerides, elevated blood pressure, and elevated fasting plasma glucose) may mediate this association. Data were Framingham Heart Study Research Materials obtained from the National Heart, Lung, and Blood Institute (NHLBI) Biologic Specimen and Data Repository Information Coordinating Center. We used Cox proportional hazards regression to estimate adjusted hazard ratios (aHR) representing the total effect (aHRTE) of probable depression, measured via the Centers for Epidemiological Studies – Depression scale, on incident CHD over approximately 18 years. Using inverse odds ratio weighting, we decomposed this estimate into natural direct effects (aHRNDE) and natural indirect effects (aHRNIE) through potential mediators (measured approximately three years after depression). Probable depression was associated with incident CHD (aHRTE=1.45, 95% confidence interval [CI]: 0.93, 2.25), and elevated waist circumference partially mediated this association (aHRNDE=1.34, 95% CI: 0.76–2.32; aHRNIE=1.08, 95% CI: 0.63–1.91). We did not find evidence of additional mediation by additional MetS components. Elevated waist circumference appears to play a role in the association between depression and CHD. [ABSTRACT FROM AUTHOR]

Background: The literature regarding the associations between depression, anxiety, and fecundity is inconsistent. While cross-sectional studies suggest that depression and/or anxiety may adversely affect fecundity, the sole cohort study showed only a small association.Objective: We sought to evaluate the association of self-reported depressive symptoms, self-reported diagnoses of depression and anxiety, and psychotropic medication use with fecundability in a prospective cohort study.Study Design: Data were derived from Pregnancy Study Online (PRESTO), an Internet-based preconception cohort study of couples attempting to conceive in the United States and Canada. At baseline, female participants completed a survey that assessed demographic information, history of physician-diagnosed depression and anxiety, self-reported depressive symptoms (assessed by the Major Depression Inventory), and use of psychotropic medications. Women completed follow-up surveys every 8 weeks for up to 12 months or until reported conception to assess changes in exposures and pregnancy status. We estimated fecundability ratios and 95% confidence intervals using proportional probabilities regression models. The analysis was restricted to 2146 women who had been attempting to conceive for ≤6 cycles at study entry.Results: Severe depressive symptoms at baseline, regardless of treatment, were associated with decreased fecundability compared with no or low depressive symptoms (fecundability ratio, 0.62; 95% confidence interval, 0.43-0.91). The fecundability ratio associated with a 10-unit increase in Major Depression Inventory score was 0.90 (95% confidence interval, 0.83-0.97). Women who reported moderate to severe depressive symptoms and had never received psychotropic medications (fecundability ratio, 0.69; 95% confidence interval, 0.48-0.99) or who were currently being treated with psychotropic medications (fecundability ratio, 0.72; 95% confidence interval, 0.44-1.20) had decreased fecundability relative to women who had no/mild depressive symptoms and had never used psychotropic medications. Former users of psychotropic medications had increased fecundability regardless of the presence of no/mild depressive symptoms (fecundability ratio, 1.22; 95% confidence interval, 1.06-1.39) or moderate to severe depressive symptoms (fecundability ratio, 1.18; 95% confidence interval, 0.80-1.76).Conclusion: We found an inverse association between depressive symptoms and fecundability, independent of psychotropic medication use. Use of psychotropic medications did not appear to harm fecundability. [ABSTRACT FROM AUTHOR]

While suicide risk following psychiatric hospitalization has been studied extensively, risk following hospitalization for physical illness is less well understood. We used random forests to examine risk factors for suicide in the year following physical illness hospitalization in Denmark. In this case-cohort study, suicide cases were all individuals who died by suicide within one year of a hospitalization for a physical illness (n = 4563) and the comparison subcohort was a 5% random sample of individuals living in Denmark on January 1, 1995 who had a hospitalization for a physical illness between January 1, 1995 and December 31, 2015 (n = 177,664). We used random forests to examine identify the most important predictors of suicide stratified by sex. For women, the top 10 most important variables for random forest prediction were all related to psychiatric diagnoses. For men, many physical health conditions also appeared important to suicide prediction. Among the top 10 variables in the variable importance plot for men were influenza, injuries to the head, nervous system surgeries, and cerebrovascular diseases. Suicide prediction after a physical illness hospitalization requires comprehensive consideration of different and multiple factors for each sex. [ABSTRACT FROM AUTHOR]

Introduction: Little is known about predictors of military veterans' suicidal ideation as they transition from service to civilian life, a potentially high-risk period that represents a critical time for intervention. This study examined factors associated with veterans' suicidal ideation in the first year after military separation.Methods: A national sample of U.S. veterans (N=7,383) from The Veterans Metrics Initiative Study reported on their mental health, psychosocial well-being, and demographic/military characteristics in an online survey at 3 and 9 months after separation. Cross-validated random forest models and mean decrease in accuracy values were used to identify key predictors of suicidal ideation. Bivariate ORs were calculated to examine the magnitude and direction of main effects associations between predictors and suicidal ideation. Data were collected in 2016/2017 and analyzed in 2021.Results: In the first year after separation, 15.1% of veterans reported suicidal ideation. Endorsing depression symptoms and, to a lesser extent, identifying oneself as experiencing depression, were most predictive of suicidal ideation. Other psychopathology predictors included higher anxiety and posttraumatic stress disorder symptoms. Psychosocial well-being predictors included higher health satisfaction and functioning, community satisfaction and functioning, and psychological resilience. Logistic models performed similarly to random forest models, suggesting that relationships between predictors and suicidal ideation were better represented as main effects than interactions.Conclusions: Results highlight the potential value of bolstering key aspects of military veterans' mental health and psychosocial well-being to reduce their risk for suicidal ideation in the first year after separation. Findings can inform interventions aimed at helping veterans acclimate to civilian life. [ABSTRACT FROM AUTHOR]

Despite the prevalence and negative impact of posttraumatic stress disorder (PTSD), there are few medications approved by the U.S. Food and Drug Administration for treatment, and approved medications do not work well enough. We leveraged large-scale electronic health record data to identify existing medications that may be repurposed as PTSD treatments. We constructed a mechanistic tree of all Food and Drug Administration–approved medications and used the tree-based scan statistic to identify medications associated with greater than expected levels of clinically meaningful improvement in PTSD symptoms using electronic health record data from the U.S. Department of Veterans Affairs. Our cohort included patients with a diagnosis of PTSD who had repeated symptom measurements using the PTSD Checklist over a 20-year period (N = 168,941). We calculated observed numbers based on patients taking each drug or mechanistically related class of drugs and the expected numbers based on the tree as a whole. Medications typically used to treat PTSD, such as the Food and Drug Administration–approved agent sertraline, were associated with improvement in PTSD symptoms, but the effects were small. Several, but not all, direct-acting antivirals used in the treatment of hepatitis C virus demonstrated a strong association with PTSD improvement. The finding was robust to a sensitivity analysis excluding patients who received established PTSD treatments, including trauma-focused psychotherapy, concurrent with hepatitis treatment. Our exploratory approach both demonstrated findings that are consistent with what is known about pharmacotherapy for PTSD and uncovered a novel class of medications that may improve PTSD symptoms. [ABSTRACT FROM AUTHOR]

Accurate identification of persons at risk of suicide is challenging because suicide is a rare outcome with a multifactorial origin. The purpose of this study was to predict suicide among persons with depression using machine learning methods. A case-cohort study was conducted in Denmark between January 1, 1995 and December 31, 2015. Cases were all persons who died by suicide and had an incident depression diagnosis in Denmark (n = 2,774). The comparison subcohort was a 5% random sample of all individuals in Denmark at baseline, restricted to persons with an incident depression diagnosis during the study period (n = 11,963). Classification trees and random forests were used to predict suicide. In men with depression, there was a high risk of suicide among those who were prescribed other analgesics and antipyretics (i.e. , non-opioid analgesics such as acetaminophen), prescribed hypnotics and sedatives, and diagnosed with a poisoning (n = 96; risk = 81%). In women with depression, there was an elevated risk of suicide among those who were prescribed other analgesics and antipyretics, anxiolytics, and hypnotics and sedatives, but were not diagnosed with poisoning nor cerebrovascular diseases (n = 338; risk = 58%). Psychiatric disorders and their associated medications were strongly indicative of suicide risk. Notably, anti-inflammatory medications (e.g., acetaminophen) prescriptions, which are used to treat chronic pain and illnesses, were associated with suicide risk in persons with depression. Machine learning may advance our ability to predict suicide deaths. [ABSTRACT FROM AUTHOR]

Some narrow patterns of posttraumatic psychiatric comorbidity are well-established (e.g., posttraumatic stress disorder and substance use). However, broad multi-diagnosis profiles of posttraumatic comorbidity are poorly characterized. The goal of the current study was to use latent class analysis (LCA) to identify profiles of posttraumatic psychopathology from 11 International Classification of Diseases (ICD-10) diagnostic categories (e.g., stress, substance, depressive, psychosis, personality). Danish national registries were used to identify 166,539 individuals (median age = 41 years, range = <1 to >100) who experienced a traumatic event between 1994 and 2016 and were diagnosed with one or more mental disorders within 5 years. Two through 14-class LCA solutions were evaluated. A 13-class solution (a) provided the best fit, with the Bayes and Akaike Information Criteria reaching a minimum, (b) was broadly consistent with prior LCA studies, and (c) included several novel classes reflecting differential patterns of posttraumatic psychopathology. Three classes were characterized by high comorbidity: broad high comorbidity (M # diagnoses = 4.3), depression with stress/substance use/personality/neurotic disorders (M # diagnoses = 3.8), and substance use with personality/stress/psychotic disorders (M # diagnoses = 3.1). The other 10 classes were characterized by distinct patterns of mild comorbidity or negligible comorbidity. Compared to the mild and negligible comorbidity classes, individuals in high comorbidity classes were younger, had lower income, and had more pre-event psychiatric disorders. Results suggest that several different comorbidity patterns should be assessed when studying and treating posttraumatic psychopathology. [ABSTRACT FROM AUTHOR]

Overview: Alcohol use is common following traumatic military deployment experiences. What is less clear is why, and for whom, particular deployment experiences lead to alcohol use.Method: The current study explored associations between deployment stressors (Warfare, Military Sexual Trauma, and Concerns about Life and Family Disruptions-"Life Disruptions"), PTSD (PCL-5), and alcohol use (CAGE) post-deployment, stratified by gender among 2344 male and female veterans (1137 men; Mage = 35). Conditional process analyses examined the indirect effect of traumatic deployment experiences on alcohol use, via PTSD symptom severity, with Life Disruptions as a moderator.Results: More severe Warfare and military sexual trauma (MST) were associated with greater PTSD symptom severity, which was associated with higher problematic alcohol use. PTSD symptom severity accounted for the associations between trauma type (i.e., MST or Warfare) and alcohol use. Among women, but not men, Life Disruptions moderated the associations between trauma type (i.e., MST, Warfare) and PTSD symptom severity, such that elevated Life Disruptions amplified the associations between trauma type and PTSD symptom severity. Moderated mediation was significant for MST among women, indicating that the strength of the indirect effect (MST ➔ PTSD ➔ problematic alcohol use) was moderated by Life Disruptions; problematic alcohol use was highest for women with greater PTSD symptom severity following exposure to more severe Life Disruptions and MST (Est. = 0.0007, SE = 0.0001, CI = 0.0002 to 0.0013).Conclusions: Taken together, alcohol use following potentially traumatic deployment experiences can be understood by considering PTSD symptom severity, gender, and Life Disruptions. [ABSTRACT FROM AUTHOR]

Posttraumatic stress disorder (PTSD) is associated with suicidal ideation and behavior, and is found to frequently co-occur with other conditions that exacerbate the risk for suicidal behavior. Despite these findings, few individuals with PTSD engage in suicidal acts, and there has been little research to examine those factors that protect against such behaviors. The current study used path analysis to examine the association among PTSD, depression, hazardous alcohol consumption, and beliefs about suicide (i.e., reasons for living) in a community sample with motor vehicle accident related-PTSD (N=50). Reasons for living were inversely associated with PTSD, depression, and alcohol use. Further, depression symptom severity accounted for the association between PTSD symptom severity and reasons for living. In contrast, hazardous alcohol consumption only demonstrated a trend for accounting for the association between PTSD and reasons for living. Our findings highlight the importance of clinicians assessing co-occurring depression symptoms and suggest the potential use of interventions that promote adaptive cognitions about suicide among people with PTSD. [ABSTRACT FROM AUTHOR]

Background: Substance use disorders (SUD) elevate the risk for COVID-19 hospitalization, but studies are inconsistent on the relationship of SUD to COVID-19 mortality.Methods: Veterans Health Administration (VHA) patients treated in 2019 and evaluated in 2020 for COVID-19 (n=5,556,315), of whom 62,303 (1.1%) tested positive for COVID-19 (COVID-19+). Outcomes were COVID-19+ by 11/01/20, hospitalization, ICU admission, or death within 60 days of a positive test. Main predictors were any ICD-10-CM SUDs, with substance-specific SUDs (cannabis, cocaine, opioid, stimulant, sedative) explored individually. Logistic regression produced unadjusted and covariate-adjusted odds ratios (OR; aOR).Results: Among COVID-19+ patients, 19.25% were hospitalized, 7.71% admitted to ICU, and 5.84% died. In unadjusted models, any SUD and all substance-specific SUDs except cannabis use disorder were associated with COVID-19+(ORs=1.06-1.85); adjusted models produced similar results. Any SUD and all substance-specific SUDs were associated with hospitalization (aORs: 1.24-1.91). Any SUD, cocaine and opioid disorder were associated with ICU admission in unadjusted but not adjusted models. Any SUD, cannabis, cocaine, and stimulant disorders were inversely associated with mortality in unadjusted models (OR=0.27-0.46). After adjustment, associations with mortality were no longer significant. In ad hoc analyses, adjusted odds of mortality were lower among the 49.9% of COVID-19+ patients with SUD who had SUD treatment in 2019, but not among those without such treatment.Conclusions: In VHA patients, SUDs are associated with COVID-19 hospitalization but not COVID-19 mortality. SUD treatment may provide closer monitoring of care, ensuring that these patients received needed medical attention, enabling them to ultimately survive serious illness. [ABSTRACT FROM AUTHOR]

Cognitive Processing Therapy (CPT) is an effective treatment for posttraumatic stress disorder (PTSD); however, not every client achieves optimal outcomes. Data were pooled from four randomized trials in which female interpersonal trauma survivors completed CPT (N = 179). Random forests of classification trees were used to investigate the role of both baseline (e.g., demographics, trauma history, comorbid disorders) and session PTSD and depressive symptom scores on predicting trajectory and outcome. Of particular focus was whether those on track for poor outcome (e.g., non-response, partial treatment response) could be identified early in therapy. Results demonstrated inconsistent findings for discrimination between delayed responders (no early change but full response after 12 weeks of therapy) and those who either showed a partial response to treatment or did not respond at all; level of discrimination depended on the assessment point under study and the chosen comparison group. Those defined as clear and early responders , however, could be reliably differentiated from the other groups by session 4. Although it is possible to identify clients who will recover from PTSD by the middle of the CPT protocol, further work is needed to accurately identify those who will ultimately not recover from PTSD during a course of CPT. • Just over 50% of completers demonstrated early and sustained treatment response. • Early responders were differentiated from other groups by session 6. • Delayed versus poor response could not be reliably identified until after session 10. • Nonresponse to CPT cannot be identified in the early-to-middle stages of therapy. [ABSTRACT FROM AUTHOR]

Objective: There is some evidence that posttraumatic stress disorder (PTSD) is associated with increased risk of infections, and it is unknown whether adjustment disorder is as well. We assessed the association between adjustment disorder and subsequent infections, and assessed additive interaction with sex.Methods: The study population included a nationwide cohort of all Danish-born residents of Denmark diagnosed with adjustment disorder between 1995 and 2011, and an age- and sex-matched general population comparison cohort. We compared rates of infections requiring inpatient or outpatient hospitalization in the two cohorts. We fit Cox proportional hazards models to compute adjusted hazard ratios (aHR) for the associations between adjustment disorder and 32 types of infections, and calculated interaction contrasts to assess interaction between adjustment disorder and sex.Results: Adjustment disorder was associated with increased rates of infections overall (n = 19,838 infections, aHR = 1.8, 95% confidence interval = 1.8. 1.9), and increased rates of each individual infection type (aHRs for 30 infections ranged from 1.5 to 2.3), adjusting for baseline psychiatric and somatic comorbidities and marital status. For many infection types (e.g., skin infections, pneumonia), interaction contrasts indicated rate differences were greater among men than women, while for two (urinary tract infections and sexually transmitted infections), rate differences were greater for women.Conclusions: These findings are consistent with studies examining the relationship between psychological stress and infections, and between PTSD and infections. They may be explained by a combination of the triggering of unhealthy behaviors as well as immune responses to stress. [ABSTRACT FROM AUTHOR]