Predictive-analytic Models of Opioid Overdose and Reoffending (022)


Study Information

Predictive analytics is a powerful tool to determine the risk of fatal and nonfatal overdose and re-arrest and re-offending of criminal justice-involved individuals living with opioid use disorder.

The University of Chicago is developing open-source software that can be used by researchers and practitioners to predict overdose and re-offending risk of their population. This project will use large administrative datasets and machine-learning technology to develop a framework for transparent predictive models and simulations to help identify people at highest risk for overdose and/or re-offending and how populations will benefit from interventions, and explore the likely policy impact of observed relationships among emerging trends to improve outcomes.

Study Team


Harold Pollack, PhD
The University of Chicago

Logo of The University of Chicago

Study Aims

• Create a linked administrative dataset of first response, criminal justice, and treatment system data, with identified features associated with overdose and re-offending risk

• Implement machine-learning analysis for identified samples of individuals and locations

• Create open-source software to share with researchers and practitioners throughout JCOIN

Research Type

Modeling Project