Functional Regression Approach to Detect Clinically Meaningful Markers of Acute Events from an HER

• Activity Number: 323
• Type: Topic Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Benjamin Goldstein*+
• Companies: Stanford University
• Keywords: Functional Regression ; Electronic Health Records

Abstract

Cardiac events are the number one cause of death for patients undergoing hemodialysis. Electronic health records present a unique opportunity to observe granular information on patients evolving health status. An important clinical question is determining which metrics, observed over time, serve as markers for impending cardiac events. An analytic challenge is both appropriately sampling a cohort from the EHR and analyzing the data to detect viable markers. In this talk we present an intuitive functional regression model for detecting acute events. We first assess departures in functional patterns and then perform a series of nested tests to determine whether the markers conform to clinically meaningful patterns. We apply the method to an EHR dataset of hemodialysis patients.