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.
•Machine learning may help characterize and predict psychiatric outcomes•Description of commonly used supervised learning methods•Introduction to model building, validation, and evaluation of algorithms•Discussion of challenges and opportunities to move field forward
Suicide is a public health problem, with multiple causes that are poorly understood. The increased focus on combining health care data with machine-learning approaches in psychiatry may help advance ...the understanding of suicide risk.
To examine sex-specific risk profiles for death from suicide using machine-learning methods and data from the population of Denmark.
A case-cohort study nested within 8 national Danish health and social registries was conducted from January 1, 1995, through December 31, 2015. The source population was all persons born or residing in Denmark as of January 1, 1995. Data were analyzed from November 5, 2018, through May 13, 2019.
Exposures included 1339 variables spanning domains of suicide risk factors.
Death from suicide from the Danish cause of death registry.
A total of 14 103 individuals died by suicide between 1995 and 2015 (10 152 men 72.0%; mean SD age, 43.5 18.8 years and 3951 women 28.0%; age, 47.6 18.8 years). The comparison subcohort was a 5% random sample (n = 265 183) of living individuals in Denmark on January 1, 1995 (130 591 men 49.2%; age, 37.4 21.8 years and 134 592 women 50.8%; age, 39.9 23.4 years). With use of classification trees and random forests, sex-specific differences were noted in risk for suicide, with physical health more important to men's suicide risk than women's suicide risk. Psychiatric disorders and possibly associated medications were important to suicide risk, with specific results that may increase clarity in the literature. Generally, diagnoses and medications measured 48 months before suicide were more important indicators of suicide risk than when measured 6 months earlier. Individuals in the top 5% of predicted suicide risk appeared to account for 32.0% of all suicide cases in men and 53.4% of all cases in women.
Despite decades of research on suicide risk factors, understanding of suicide remains poor. In this study, the first to date to develop risk profiles for suicide based on data from a full population, apparent consistency with what is known about suicide risk was noted, as well as potentially important, understudied risk factors with evidence of unique suicide risk profiles among specific subpopulations.
Coinciding with the development and revision of conceptual models of psychopathology, there has been a proliferation in the number of self-report clinical questionnaires and studies evaluating their ...psychometric properties. Unfortunately, many clinical measures are constructed and evaluated using suboptimal methods. This review provides current guidelines for the conceptualization, development, and psychometric validation of clinical questionnaires using latent variable methods. A two-stage exploratory-confirmatory framework is provided. The exploratory stage includes item selection and revision, initial structural evaluation, and preliminary tests of concurrent validity (e.g., convergent and discriminant). The confirmatory stage involves replicating factor structure using a more restrictive model, identifying areas of model strain, conducting additional tests of concurrent and predictive validity, and evaluating measurement invariance. Recommendations are provided for (
a
) item generation, (
b
) how to use different types of exploratory and confirmatory factor models to determine structure, and (
c
) evaluating reliability and validity using a latent variable measurement model approach.
Earthquakes are a common and deadly natural disaster, with roughly one-quarter of survivors subsequently developing posttraumatic stress disorder (PTSD). Despite progress identifying risk factors, ...limited research has examined how to combine variables into an optimized post-earthquake PTSD prediction tool that could be used to triage survivors to mental health services. The current study developed a post-earthquake PTSD risk score using machine learning methods designed to optimize prediction. The data were from a two-wave survey of Chileans exposed to the 8.8 magnitude earthquake that occurred in February 2010. Respondents (n = 23,907) were interviewed roughly three months prior to and again three months after the earthquake. Probable post-earthquake PTSD was assessed using the Davidson Trauma Scale. We applied super learning, an ensembling machine learning method, to develop the PTSD risk score from 67 risk factors that could be assessed within one week of earthquake occurrence. The super learner algorithm had better cross-validated performance than the 39 individual algorithms from which it was developed, including conventional logistic regression. The super learner also had a better area under the receiver operating characteristic curve (0.79) than existing post-disaster PTSD risk tools. Individuals in the top 5%, 10%, and 20% of the predicted risk distribution accounted for 17.5%, 32.2%, and 51.4% of all probable cases of PTSD, respectively. In addition to developing a risk score that could be implemented in the near future, these results more broadly support the utility of super learning to develop optimized prediction functions for mental health outcomes.
There has been limited progress evaluating the validity of dimensional approaches to emotional disorder classification. This has occurred in part because of a lack of standardized assessment tools ...developed with the specific intent of studying dimensional classification. The goal of the current study was to develop and validate the Multidimensional Emotional Disorder Inventory (MEDI) to efficiently assess nine empirically supported transdiagnostic dimensions proposed in the Brown and Barlow (2009) profile approach to emotional disorder classification: neurotic temperament, positive temperament, depression, autonomic arousal, somatic anxiety, social anxiety, intrusive cognitions, traumatic reexperiencing, and avoidance. The MEDI factor structure, reliability, and convergent/discriminant validity was evaluated in outpatients with emotional disorders (pilot sample = 227; validation sample = 780). The final 9-factor solution fit the data well. Intercorrelations among MEDI factors were consistent with previous research, and all MEDI dimensions had acceptable reliability. Correlations with common self-report questionnaires and DSM-5 diagnoses supported the convergent/discriminant validity of all nine MEDI dimensions. Collectively, these results support the use of 49-item MEDI in clinical research samples. The MEDI should be used in future research to evaluate the validity of the Brown and Barlow (2009) approach to emotional disorder classification. Because it provides an efficient assessment of several well-established emotional disorder traits and phenotypes, the MEDI also may have utility for general research or clinical purposes.
Public Significance Statement
The purpose of this study was to develop and validate an efficient self-report measure of nine distinct emotional disorder traits/symptoms. The measure was found to be reliable and valid, and may be useful for researchers or clinicians interested in a broad yet practical dimensional assessment of emotional disorder features.
Depression is a common mental disorder that may comprise distinct, underlying symptom patterns over time. Associations between stressful life events throughout the civilian lifecourse-including ...during childhood-and adult depression have been documented in many populations, but are less commonly assessed in military samples. We identified different trajectories of depression symptoms across four years in a military cohort using latent class growth analysis, and investigated the relationship between these trajectories and two domains of civilian life experiences: childhood adversity (e.g., being mistreated during childhood) and more proximal stressful experiences (e.g., divorce). A four-group depression model was identified, including a symptom-free group (62%), an increasing symptom group (13%), a decreasing symptom group (16%), and a "chronic" symptom group (9%). Compared to the symptom-free group, soldiers with childhood adversity were more likely to be in the chronic depression, decreasing, and increasing symptom groups. Time-varying adult stressors had the largest effect on depression symptoms for the increasing symptom group compared to other groups, particularly in the last two years of follow-up. This study indicates the importance of considering events from throughout the lifecourse-not only those from deployment-when studying the mental health of servicemembers.
The US Army experienced a sharp increase in soldier suicides beginning in 2004. Administrative data reveal that among those at highest risk are soldiers in the 12 months after inpatient treatment of ...a psychiatric disorder.
To develop an actuarial risk algorithm predicting suicide in the 12 months after US Army soldier inpatient treatment of a psychiatric disorder to target expanded posthospitalization care.
There were 53,769 hospitalizations of active duty soldiers from January 1, 2004, through December 31, 2009, with International Classification of Diseases, Ninth Revision, Clinical Modification psychiatric admission diagnoses. Administrative data available before hospital discharge abstracted from a wide range of data systems (sociodemographic, US Army career, criminal justice, and medical or pharmacy) were used to predict suicides in the subsequent 12 months using machine learning methods (regression trees and penalized regressions) designed to evaluate cross-validated linear, nonlinear, and interactive predictive associations.
Suicides of soldiers hospitalized with psychiatric disorders in the 12 months after hospital discharge.
Sixty-eight soldiers died by suicide within 12 months of hospital discharge (12.0% of all US Army suicides), equivalent to 263.9 suicides per 100,000 person-years compared with 18.5 suicides per 100,000 person-years in the total US Army. The strongest predictors included sociodemographics (male sex odds ratio (OR), 7.9; 95% CI, 1.9-32.6 and late age of enlistment OR, 1.9; 95% CI, 1.0-3.5), criminal offenses (verbal violence OR, 2.2; 95% CI, 1.2-4.0 and weapons possession OR, 5.6; 95% CI, 1.7-18.3), prior suicidality OR, 2.9; 95% CI, 1.7-4.9, aspects of prior psychiatric inpatient and outpatient treatment (eg, number of antidepressant prescriptions filled in the past 12 months OR, 1.3; 95% CI, 1.1-1.7), and disorders diagnosed during the focal hospitalizations (eg, nonaffective psychosis OR, 2.9; 95% CI, 1.2-7.0). A total of 52.9% of posthospitalization suicides occurred after the 5% of hospitalizations with highest predicted suicide risk (3824.1 suicides per 100,000 person-years). These highest-risk hospitalizations also accounted for significantly elevated proportions of several other adverse posthospitalization outcomes (unintentional injury deaths, suicide attempts, and subsequent hospitalizations).
The high concentration of risk of suicide and other adverse outcomes might justify targeting expanded posthospitalization interventions to soldiers classified as having highest posthospitalization suicide risk, although final determination requires careful consideration of intervention costs, comparative effectiveness, and possible adverse effects.
The present study evaluated the latent structure of the NEO Five-Factor Inventory (NEO FFI) and relations between the five-factor model (FFM) of personality and dimensions of DSM-IV anxiety and ...depressive disorders (panic disorder, generalized anxiety disorder GAD, obsessive—compulsive disorder, social phobia SOC, major depressive disorder MDD) in a large sample of outpatients (N = 1,980). Exploratory structural equation modeling (ESEM) was used to show that a five-factor solution provided acceptable model fit, albeit with some poorly functioning items. Neuroticism demonstrated significant positive associations with all but one of the disorder constructs whereas Extraversion was inversely related to SOC and MDD. Conscientiousness was inversely related to MDD but demonstrated a positive relationship with GAD. Results are discussed in regard to potential revisions to the NEO FFI, the evaluation of other NEO instruments using ESEM, and clinical implications of structural paths between FFM domains and specific emotional disorders.
Suicide risk is high in the 30 days after discharge from psychiatric hospital, but knowledge of the profiles of high-risk patients remains limited.
To examine sex-specific risk profiles for suicide ...in the 30 days after discharge from psychiatric hospital, using machine learning and Danish registry data.
We conducted a case-cohort study capturing all suicide cases occurring in the 30 days after psychiatric hospital discharge in Denmark from 1 January 1995 to 31 December 2015 (n = 1205). The comparison subcohort was a 5% random sample of all persons born or residing in Denmark on 1 January 1995, and who had a first psychiatric hospital admission between 1995 and 2015 (n = 24 559). Predictors included diagnoses, surgeries, prescribed medications and demographic information. The outcome was suicide death recorded in the Danish Cause of Death Registry.
For men, prescriptions for anxiolytics and drugs used in addictive disorders interacted with other characteristics in the risk profiles (e.g. alcohol-related disorders, hypnotics and sedatives) that led to higher risk of postdischarge suicide. In women, there was interaction between recurrent major depression and other characteristics (e.g. poisoning, low income) that led to increased risk of suicide. Random forests identified important suicide predictors: alcohol-related disorders and nicotine dependence in men and poisoning in women.
Our findings suggest that accurate prediction of suicide during the high-risk period immediately after psychiatric hospital discharge may require a complex evaluation of multiple factors for men and women.
The 2019 coronavirus (COVID-19) pandemic led to elevated levels of psychological distress on a global scale. Given that individuals with pre-existing physical conditions are at risk for worse ...COVID-19 outcomes, those dealing with the stress of physical health problems (including knowing someone with health problems) may experience more severe distress during the pandemic.
Patients with emotional disorders who completed a diagnostic assessment in the 6 months prior to COVID-19 were surveyed in May-June 2020 on their emotional reactions to COVID-19 (N = 77).
Multiple linear regression was used to test the hypothesis that chronic stress due to having and knowing others with physical health problems would predict COVID-related worries and behaviours, holding pre-COVID levels of depression, anxiety and worry about health constant. Chronic stress surrounding the health of others was significantly associated with experiencing more severe COVID-related worry and behaviours. In comparison, chronic stress due to one's own health problems had weak and non-significant associations with COVID-related worries and behaviours.
Results indicate that outpatients who report stress about surrounding loved one's health are at risk for experiencing more severe distress during a health pandemic and thus, may benefit from targeted outreach, assessment and intervention.