PI: Walid Gellad, MD, MPH
Funding Source: R.K. Mellon Foundation
July 2018 - June 2020, July 2020 - December 2025
This project, conducted in partnership with the Allegheny County Department of Human Services, develops and validates a machine learning algorithm that integrates Medicaid claims with human services and criminal justice data to predict individuals’ risk of opioid overdose within the 30 days from jail release. By leveraging a gradient boosting model with nearly 300 predictors—including social and justice-system factors—the study significantly improves prediction accuracy (C‑statistic = 0.92 vs. 0.87) and identifies high-risk subgroups to help target interventions effectively.
