Skip to main content
ABCT 2025 59th Annual Convention Main banner
Psychology Tools Ad
Final Program


Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Presentation Icons
Recorded Session
Recorded Session
Ticketed Session
Ticketed Session
ViewAttendees 4

Symposia

Suicide and Self-Injury

Advancing Suicide Risk Assessment for Veterans: Leveraging AI, Assessing Novel Risks, and Developing New Strategies for Risk Management of Veteran Suicide

2 - (SYM 30) Effectiveness of Generative Ai-large Language Models' Recognition of Veteran Suicide Risk: A Comparison with Human Mental Health Providers Using a Risk Stratification Model

Saturday, November 22, 2025
10:22 AM - 10:41 AM CST
Location: Celestin G, Level 3
Keywords: Suicide, Risk / Vulnerability Factors, Assessment
Recommended Readings: Litschi, M. A., Lancaster, S. L., Linkh, D. J., Renno, S., Neal‐Walden, T. A., Lawless, C. E., & Breetz, A. A. (2024). Evaluating suicide risk stratification in outpatient settings: A vignette‐based approach. Clinical Psychology & Psychotherapy, 31(2), e2965. https://doi.org/10.1002/cpp.2965,   Edwards, E. R., Fortuna, A., Holliday, R., Addison, H., & Tsai, J. (2024). Prevalence of mental health conditions, substance use disorders, suicidal ideation and attempts, and experiences of homelessness among Veterans with criminal-legal involvement: A meta-analysis. Clinical Psychology Review, 102533. https://doi.org/10.1016/j.cpr.2024.102533,   Matarazzo, B. B., Eagan, A., Landes, S. J., Mina, L. K., Clark, K., Gerard, G. R., ... & Reger, M. A. (2023). The Veterans Health Administration REACH VET program: suicide predictive modeling in practice. Psychiatric services, 74(2), 206-209. https://doi.org/10.1176/appi.ps.2021006,   ,  
Abstract Body
With over 6,300 U.S. military veterans dying by suicide annually, the VHA is exploring AI-driven strategies for suicide risk assessment. While machine learning has been widely used, the potential of generative AI-large language models (GAI-LLMs) remains unexplored. This study evaluates the effectiveness of GAI-LLMs, specifically ChatGPT-3.5, ChatGPT-4o, and Google Gemini, in using the VHA’s Risk Stratification Table for identifying suicide risks and making treatment recommendations in response to standardized veteran vignettes. We compared the GAI-LLMs’ assessments and recommendations for both acute and chronic suicide risks to evaluations by mental health care providers (MHCPs). Four vignettes representing varying levels of suicide risk were used. The GAI-LLMs showed discrepancies with MHCPs in acute risk assessments, underrating the most acute case and overrating the least acute. Chronic risk evaluations were generally aligned, except for one case rated higher by a GAI-LLM. Variation across models was observed, with ChatGPT-3.5 assigning lower acute risk, while ChatGPT-4o identified higher chronic risk and consistently recommended hospitalization. Treatment planning by GAI-LLMs was predicted by chronic but not acute risk ratings. Findings demonstrate that while GAI-LLMs offer potential suicide risk assessments comparable to MHCPs, significant variation exists across different GAI-LLMs in both risk evaluation and treatment recommendations. These findings highlight the need for further research into optimizing GAI-LLMs for consistent and reliable use in clinical settings, ensuring they complement rather than replace human expertise.