Abstract
We developed an adverse drug event (ADE) monitor based on published rules, made modifications to improve its sensitivity and positive predictive value, and compared it to manual chart review and self report. Over the study period, the ADE monitor (COMP) identified 279 ADEs while the manual chart review and self-report mechanism (CHART) identified 421; 79 cases overlapped. Cases identified by COMP were more likely to be severe (p=0.03) but were similarly preventable (P=0.10). Changes in the knowledge base substantially improved the sensitivity and positive predictive value of the rules.