The Statistical Methods for Health Equity Series is a monthly online series co-hosted by the Data Science for Health Equity community, the Alan Turing Institute Health Equity Interest Group, and the Department of Statistical Science at University College London.

---

We are excited to announce the next instalment of our webinar series with speaker Vincent Jeanselme - Columbia University

Title: Advancing AI-Based Risk Prediction for Improved Decision-Making in Healthcare

Abstract: Accurate prediction of patient outcomes plays a critical role in healthcare decision-making, shaping treatment strategies, clinical guidelines, hospital operations, and insurance policy. As AI becomes embedded in these decisions, prediction errors can propagate into clinical and operational choices. Ensuring that AI models are both accurate and fair is therefore essential to benefit all patients.

A standard tool to model risk is survival analysis, which estimates the time until an event such as heart failure. However, patients often experience competing events, such as death from another cause, that prevent the outcome of interest. Practitioners often mishandle these events in current risk predictions, treating them as if the outcome were unobserved.

Vincent's research shows that this common practice has serious repercussions for decision-making. Mishandling competing risks systematically overestimates risk and produces unequal error across demographic groups. These errors may propagate into downstream decisions, affecting treatment and resource allocation, and ultimately, who benefits from AI in healthcare. In the analysis of cardiovascular management, this modeling error increases overtreatment by 1.8%, corresponding to a misallocation of millions of dollars and reinforcing disparities in care.

In this talk, Vincent will explain how this common modeling practice leads to biased and unfair risk estimation, introduce a neural network framework that correctly accounts for competing risks, and show how this approach improves both predictive accuracy and treatment recommendations.

More reading: https://arxiv.org/abs/2508.05435

Bio: Vincent Jeanselme is a postdoctoral researcher at the reAIM Lab, Columbia University, where he works with Prof. Shalmali Joshi at the intersection of machine learning, healthcare, and decision-making. His research focuses on developing and evaluating machine learning models to support clinical decision-making and reduce disparities in healthcare access and delivery.

--

Please direct any questions about the webinar series to info@datascienceforhealthequity.com.