Dr. Han’s Research

A Foundation Model for Perioperative Prediction and Causal Inference

Abstract: In the perioperative period, anesthesia providers are faced with increasingly complex medical decisions as patient care and surgical procedures continue to evolve. Development of computational tools that harness the wealth of generated perioperative data presents a significant opportunity to support anesthesia practice. Current approaches to guide preoperative and intraoperative decision making have been limited by low outcome generalizability, small patient datasets, biased features due to manual selection, and poor clinical actionability. Furthermore, intraoperative data remains heavily underutilized due to its varying, time-dependent, and resource-intensive nature. To address these challenges, Dr. Han proposes to build a foundation model for perioperative data, a type of unified artificial intelligence framework that can serve as a versatile “foundation” for a wide range of downstream tasks. To accomplish this, a global, reusable representation of patients and their procedures will be built, which Dr. Han will subsequently apply to outcome prediction, causal inference, and simulation of hypothetical clinical decisions via counterfactual modeling. With success, this approach will leverage a wealth of anesthetic data to improve risk stratification and identify potentially actionable interventions to further enhance personalized decision-making in the perioperative setting.