Enhancing Identification of Suicide Risk among Military Service Members and Veterans: A Machine Learning Approach to Suicidality
Principal Investigator: 

Andrew Littlefield

Texas Tech Universty

Using psychological survey data from the Military Suicide Research Consortium’s (MSRC) Common Data Elements (N = 5,977-6,058 across outcomes), three primary machine learning approaches and a traditional statistical approach showed similar classification performance of suicide thoughts and behaviors based on CDE items measuring hopelessness, thwarted belongingness, anxiety sensitivity, PTSD, insomnia, alcohol and substance use.

It has consistently been shown that there is a need for better methods to predict suicidal thoughts and behaviors. When examining the most prevalent means of assessing the likelihood of an individual engaging in suicidal thoughts and behaviors, researchers have shown that little progress has been made over the past 50 years. Many of the factors that have been examined were determined to be weak or inaccurate predictors. This has lead researchers to believe that we need to implement new approaches to examining what leads someone to having suicidal thoughts or behaviors.

Dr. Andrew Littlefield and colleagues hopes to utilize multiple techniques labeled as machine learning to better examine predictors of suicide. With his collaborators, he plans to utilize these techniques within a military population, as they have increased likelihood of suicide. The way in which machine learning differs from prior approaches to identifying factors that lead to suicide is by its focus on prediction over explanation. Machin learning algorithms are able to analyze an immense number of different factors and account for complex relationships between them. As such, it does not need an explanatory theory-based approach but is able to assimilate vast quantities of data at once. Utilizing the common data elements (CDE) collected by the MSRC, Dr. Littlefield and his colleagues hope to find ways to identify those at heightened risk of engaging in suicidal behaviors or thoughts within the military. While many prior studies have utilized the CDE data to test for explanatory relationships, no prior study has included all the CDE variables to predict suicide related outcomes. Dr. Littlefield and his collaborators believe that this approach will allow him to identify optimal classification algorithms that can be utilized in the future to identify those at risk.

The study will be utilizing machine learning to identify and classify people across six different types of suicidality. Specifically, it will be trying to identify attempters, those with passive suicidal ideation, those with active suicidal ideation, those with a suicide plan, those with a history of suicide attempts, and those with heightened risk of lethal suicide attempts. It will compare the algorithms identified by machine learning against those using classic methods to test is machine learning is able to provide better identification of those at risk of suicidality.

The goal of this study is to provide better classification mechanisms to identify crucial components of suicide risk for Services members. Through thorough analysis of the CDE, these better classification mechanisms can be utilized as screening tools and improve the mental health of those in the military. Hopefully this better identification will allow us to prevent further suicides.

No news on file at this time.

1 Publications Listed
Alt Metrics
23 readers on Mendeley
Littlefield, A. K., Cooke, J. T., Bagge, C., Glenn, C., Kleiman, E. M., Jacobucci, R., Millner, A. J., & Steinley, D.
Clinical Psychological Science,