Abstract: We investigated the usefulness of free-text override reasons to detect CDS alert malfunctions and prototyped techniques to automatically detect malfunctions by analyzing free-text override reasons. Although override reasons typically are never shown to EHR technical support, we found that many override reasons expressed frustration towards alerts. Out of 40 alerts flagged automatically based on override reasons, 23 were categorized by a human evaluator as “likely broken,” leading us to fix several (previously unknown) alert malfunctions.

Learning Objective 1: Formulate a simple approach to discover CDS alert malfunctions based on sentiments (such as frustration) expressed in free-text override reasons.


Skye Aaron (Presenter)
Brigham and Women's Hospital

Adam Wright, Brigham and Women's Hospital

Presentation Materials: