Scroll to top

Machine Learning Strategies for Predicting the Risk of Suicide Using Clinical Note Text

Funding Source

Defense Advanced Research Projects Agency, Department of Defense, PR220927

Project Period


Principal Investigator

Gui Jiang, PhD (Geisel School of Medicine at Dartmouth)

Other Project Staff

Nicholas C. Jacobson (Co-I)

Project Summary

This retrospective cohort study will use data selected from the Veterans Affairs (VA) Corporate Data Warehouse from years 2015-2018, including over 25,000 VA users that died by suicide. 1:4 matching will be utilized in which cases and controls will be selected that share suicide risk thresholds based upon user scores on Recovery Engagement and Coordination for Health – Veterans Enhanced Treatment (REACH VET), VA’s currently used structured EMR suicide risk prediction metric. Using Text Integration Utility codes, we will obtain all clinical notes associated with coded psychotherapy encounters beginning 6-months before death by suicide and ending five days before death.