Evaluation of Antimicrobial Stewardship-Related Alerts Using a Clinical Decision Support System

Background: Information technology, including clinical decision support systems (CDSS), have an increasingly important and growing role in identifying opportunities for antimicrobial stewardship-related interventions. Objective: The aim of this study was to describe and compare types and outcomes of CDSS-built antimicrobial stewardship alerts. Methods: Fifteen alerts were evaluated in the initial antimicrobial stewardship program (ASP) review. Preimplementation, alerts were reviewed retrospectively. Postimplementation, alerts were reviewed in real-time. Data collection included total number of actionable alerts, recommendation acceptance rates, and time spent on each alert. Time to de-escalation to narrower spectrum agents was collected. Results: In total, 749 alerts were evaluated. Overall, 306 (41%) alerts were actionable (173 preimplementation, 133 postimplementation). Rates of actionable alerts were similar for custom-built and prebuilt alert types (39% [53 of 135] vs 41% [253 of 614], P = .68]. In the postimplementation group, an intervention was attempted in 97% of actionable alerts and 70% of interventions were accepted. The median time spent per alert was 7 minutes (interquartile range [IQR], 5-13 minutes; 15 [12-17] minutes for actionable alerts vs 6 [5-7] minutes for nonactionable alerts, P < .001). In cases where the antimicrobial was eventually de-escalated, the median time to de-escalation was 28.8 hours (95% confidence interval [CI], 10.0-69.1 hours) preimplementation vs 4.7 hours (95% CI, 2.4-22.1 hours) postimplementation, P < .001. Conclusions: CDSS have played an important role in ASPs to help identify opportunities to optimize antimicrobial use through prebuilt and custom-built alerts. As ASP roles continue to expand, focusing time on customizing institution specific alerts will be of vital importance to help redistribute time needed to manage other ASP tasks and opportunities.