Conference Program

Regulatory Example of Using Machine Learning in Average Treatment Effect Estimation (303673)

View Presentation

*Jaejoon Song, FDA/CDER 
Di Zhang, US Food and Drug Administration 

Keywords: propensity score

This presentation highlights a real regulatory submission using propensity score matching with machine learning to estimate the effect of antipsychotic drugs on stroke risk, among patients 65 years or older regardless of dementia status, in a retrospective observational cohort study. We will also discuss the current utility and challenges using machine learning with real-world data to estimate average treatment effect for regulatory decision making.