A MACHINE LEARNING APPROACH FOR ALERT BEHAVIOR RESPONSE MODELING TO MITIGATE ALERT FATIGUE IN HEALTH INFORMATION SYSTEMS

Abstract

This research investigates novel approaches to reduce the burden of alert fatigue faced by primary care physicians using Clinical Decision Support Systems (CDSS) within EMR systems. CDSS issue a range of alerts to assist physicians in patient management with respect to clinical guidelines and institutional clinical pathways. However, the generation of alerts is usually suboptimal, and does not consider the physician’s clinical context. Our approach is to understand the physician’s practice to triage alert issuance, ensuring that alerts are adequately addressed by physicians without causing unnecessary alert fatigue. We utilize machine learning techniques to: cluster physicians into distinct practice groups based on their practice data, stratify the wide range of CDSS alerts based on key, defining attributes, and learn a classification based mapping between physician practice groups and alert types to develop an innovative alert issuance strategy that greatly reduces the volume of alerts presented to each physician group.