Predicting suicide death among veterans after psychiatric hospitalization using transformer based models with social determinants and NLP.

ABSTRACT: Predictions of suicide death of patients discharged from psychiatric hospitals (PDPH) can guide intervention efforts including intensive post-discharge case management programs, designed to reduce suicide risk among high-risk patients. This study aims to determine if additions of social and behavioral determinants of health (SBDH) as predictors could improve the prediction of suicide death of PDPH. We analyzed a cohort of 197,581 US Veterans discharged from 129 VHA psychiatric hospitals across the US between January 1, 2017, and July 1, 2019 with a total of 414,043 discharges. Predictive variables included administrative data and SBDH, the latter derived from unstructured clinical notes via a natural language processing (NLP) system and ICD codes, observed within a 365-day window prior to

SUBMITTER: Yang Z 

PROVIDER: S-EPMC12698767 | biostudies-literature | 2025 Nov

REPOSITORIES: biostudies-literature