skip to Main Content

Development and validation of a prediction model of time to suicide attempt

Prediction models have the potential to transform targeted prevention strategies for suicide, a leading cause of death in the U.S. The work of our partners in the Mental Health Research Network demonstrated that predictive machine learning algorithms for suicide risk based on electronic health record data can accurately identify individuals with elevated suicide risk. However, a key limitation of current suicide prediction models is that the outcome window of prediction is 90 days post-visit. This means that current models do not necessarily distinguish between someone who is at imminent risk for suicide, and someone whose suicide attempt may occur months after the point of prediction. To address this limitation, we will develop models that predict time to suicide attempt and evaluate their overall performance, as well as performance across racial-ethnic categories where differences in suicide rates have been observed.

Investigator: Papini, Santiago

Funder: Northern California Community Benefit Programs

Explore all studies and publications

Back To Top