Recorded on March 30, 2023 by the Stanford Center for Artificial Intelligence in Medicine and Imaging.

Abstract:
Patients present to the Emergency Department (ED) with diverse and rapidly progressive disease states for which the necessity and optimal timing of tests and interventions are difficult to predict. ED patients are routinely connected to sophisticated monitors that continuously measure several dimensions of physiology, which are known to predict a variety of disease outcomes. Yet the vast majority of monitoring data is discarded, and is inaccessible to clinicians and researchers. This talk presents recent work demonstrating the limitations of conventional clinical representations of patient trajectories, the diagnostic and prognostic benefits of modeling continuous monitoring and other multimodal time series data in the ED, and our efforts to make this approach useful for clinical decision-making in real time.

Speakers:
David Kim, MD PhD, is an Assistant Professor of Emergency Medicine at Stanford. His research focuses on predicting the short- and long-term trajectories of ED patients through multimodal modeling of physiologic, clinical, and administrative data.

Julia Reisler, BS, is a Master's student in Computer Science at Stanford. Prior to her graduate studies, Julia was a machine learning engineer at Apple, where she specialized in active learning and distribution shift.