Joint models targeting U.S. Army soldiers at high-risk of post-separation unemployment, homelessness, and suicide-related behaviors

Shelby Borowski, Emily R. Edwards, Joseph C. Geraci, Sarah M. Gildea, Irving Hwang, Chris J. Kennedy, Howard Liu, Alex Luedtke, Nancy A. Sampson, David M. Benedek, Vincent F. Capaldi, James A. Naifeh, Matthew K. Nock, James Wagner, Murray B. Stein, Robert J. Ursano, Ronald C. Kessler*

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

Abstract

Transitioning service members (TSMs) leaving military service have high risks of unemployment, homelessness, nonfatal suicide attempt (SA), and suicide death. Data from n = 7188 recently separated TSMs from the U.S. Army were used to update previously developed models for post-separation homelessness and SA based on data at the time of separation and to develop a new unemployment model. Predicted probabilities of suicide from a model developed elsewhere were imputed for comparison purposes. Cross-validated predictions were significant for the homelessness (AU-ROC = 0.68) and SA (AU-ROC = 0.78) models but not the unemployment model (AU-ROC = 0.60). Elevated cross-validated risk was found for the 10% of TSMs at the highest predicted risk of homelessness (SN = 26.6%), 20% for SA (SN = 60.9%), and 10% for suicide death (SN = 34.1%). 28% of TSMs were in the highest risk categories for at least one and 10% for more than one outcome. Findings regarding incomplete overlap highlight the complexities of risk targeting when multiple outcomes are of interest.

Original languageEnglish
Article number10
Journalnpj Mental Health Research
Volume5
Issue number1
DOIs
StatePublished - Dec 2026

Cite this