Bayesian Learning for Disparities in Rare Surgical Outcomes: An Integrated Data Approach
Conference
65th ISI World Statistics Congress 2025
Format: IPS Abstract - WSC 2025
Keywords: bayesian hierarchical model, kernel methods
Session: IPS 917 - Harnessing the Power of Functional Data and Machine Learning in Biomedical Research
Tuesday 7 October 2 p.m. - 3:40 p.m. (Europe/Amsterdam)
Abstract
This study tackles under-explored hospital-level contributions to population-wide surgical outcome disparities. We address the limitations of existing statistical methods in isolating hospital effects due to rare events and complex racial/ethnic data. We propose novel machine learning approaches, including nuanced statistical adjustments beyond basic racial categorizations, probabilistic geo-spatial corrections, and multi-level kernel adjustments for the functional behavior of the risk components associated with the surgical outcomes. These approaches will improve risk calculators for patient protocols and facilitate the development of insightful hospital-level hierarchical disparity models. We acknowledge the challenges of analyzing national data for this intricate issue, but this work has the potential to significantly improve our understanding of how hospitals influence surgical outcome disparities.