Machine Learning for Overdose Risk Classification - From Clinical Decision Support to Real-World Deployment

Abstract:

This talk presents the development of machine learning models for opioid overdose risk classification in vulnerable populations. Using electronic health records from the BC Provincial Overdose Cohort (36,679 cases), I developed classification models achieving 88.77% accuracy and 91.12% AUROC. The research addresses key methodological challenges in clinical machine learning: handling high-dimensional temporal data, managing extreme class imbalance, and ensuring model interpretability for clinical stakeholders.

About Andy:

Andy Tai is a Postdoctoral Teaching and Learning Fellow in the Master of Data Science Program at the University of British Columbia’s Department of Statistics. His research focuses on developing trustworthy machine learning systems for healthcare applications, particularly clinical decision support for vulnerable populations. His doctoral work on opioid overdose risk prediction achieved high accuracy metrics and successful real-world deployment through the BC Provincial Health system.