Using Machine Learning for Military Service Members and Veterans at Risk for Suicide

Suicide rates are high among military members and veterans

Suicide is a major public health issue in the United States, especially among military service members and veterans. Suicide rates for this population have increased over recent years; they are twice as likely to die by suicide than the general population. Service members and veterans are exposed to many occupational and life stressors that the average person does not experience. These include emotional and physical traumas that put them at risk for depression, post-traumatic stress disorder, anxiety, sleep disorders, substance misuse, and isolating lifestyles, all of which can lead to increased suicide risk.

Dr. Chung-Yi Chiu, a certified rehabilitation counselor and rehabilitation psychologist at the University of Illinois Urbana-Champaign, wanted to see how we could leverage the power of technology to identify those potentially at risk for suicide. She and her colleague, Dr. Xiaotian Gao, a statistician with a background in machine learning and health data analysis, are collaborating to develop a tool for clinicians to flag potential risk factors for suicide ideation and suicide attempts among military service members and veterans. Dr. Kelly Clary, an assistant professor in social work at Texas State University, provides practice insight from social work perspectives.

Using machine learning for risk assessment

The researchers for this study are developing a personalized multifaceted reporting of an individual’s holistic health, incorporating 10 dimensions that can impact suicide ideation/suicide attempts. They will be using the Army Study to Assess Risk and Resilience in Servicemembers (STARRS dataset) in their algorithm. The dimensions include: 1) demographics and family history, 2) overall health, 3) overall mental health and illnesses, 4) mood, 5) spirituality, 6) treatment, 7) self-identity, 8) ownership of weapons, 9) social networks/personal relationship, and 10) military experience. The algorithm will assess these 10 clinically usable dimensions in STARRS to detect risk factors for suicide ideation and suicide attempts, flagging those at higher risk. This information will be visualized into profiles to assist in clinical screening, evaluation, and intervention.

“Using machine learning will give clinicians a more comprehensive understanding of patients through summarized clinical dimensions based on multiple questionnaires and help provide the most efficient and effective approach for early interventions.” – Dr. Chiu

Developing a model to save lives

This machine learning approach will provide clinicians and stakeholders with data-driven guidance on addressing suicide risk factors with observable and personal, contextual data. By incorporating this information into regular healthcare visits, clinicians could reduce suicide ideation and suicide attempts in this population.