Predicting Homelessness Among Transitioning U.S. Army Soldiers

Area of Science:

• Public Health
• Military Health
• Data Science

Background:

• Homelessness is a significant risk for transitioning service members (TSMs).
• Effective identification of at-risk TSMs is crucial for preventive interventions.
• Existing methods may not adequately identify all TSMs at highest risk.

Purpose of the Study:

• To develop and validate a practical machine learning-based method for triaging Army TSMs at highest risk of homelessness.
• To target preventive interventions towards TSMs most likely to experience homelessness post-service.
• To improve the efficiency and effectiveness of homelessness prevention programs for military personnel.

Main Methods:

• Utilized data from 4,790 soldiers in the Study to Assess Risk and Resilience in Servicemembers-Longitudinal Study (STARRS-LS).
• Trained two machine learning models: Stage-1 (administrative/geospatial data) and Stage-2 (self-reported survey data).
• Outcome variable was homelessness within 12 months post-transition.

Main Results:

• The 12-month post-transition homelessness prevalence was 5.0%.
• Stage-1 model identified 30% of TSMs, accounting for 52% of homelessness.
• Stage-2 model identified 10% of all TSMs (33% of Stage-1 high-risk group), accounting for 35% of all homelessness.

Conclusions:

• Machine learning models can effectively identify TSMs at high risk of homelessness.
• This approach enables targeted outreach and assessment for preventive interventions.
• The developed method offers a practical solution for reducing TSM homelessness.