Methods: 15 different types of alerts (1 pre-built and 14 custom) were chosen for the initial ASP review. Pre-intervention, alerts were generated retrospectively from March 1-31, 2014. Post intervention, alerts were reviewed for ASP intervention by a pharmacist in real-time from May 19-June 20, 2014. Data collection included the total actionable (an intervention was or could have been attempted), acceptance of recommendation, and time spent for each type of alert. For de-escalation alerts, time to de-escalation to a more narrow spectrum agent was collected.
Results: 749 alerts were evaluated (373 pre) and (376 post). The primary service was medical (52% vs. 57%) and there was an infectious diseases consult in 48% and 47% of alerts in pre vs. post groups, respectively. Overall, 306 (41%) alerts were actionable (173 pre and 133 post). Custom built alerts were more actionable (53/95, 56%) vs. pre-built alerts (253/614, 41%), p<0.01. The most common alert types were drug-bug mismatch, custom de-escalation, and drug-interaction alerts. In the post group, an intervention was attempted in 97% of actionable alerts and 70% (91/131) of interventions were accepted. The average time spent per alert was 7 minutes, with actionable alerts taking average 15 min and non-actionable 6 min, p<0.01. For de-escalation alerts, time to de-escalation was 28.8 hours in pre vs. 4.7 hours in the post group, p<0.01.
Conclusion: ASP programs evaluating CDSS should consider customizing alerts to improve rate of actionable alerts and minimize time spent on non-actionable alerts. Overall, the use of CDSS as part of an ASP helped identify targets for de-escalation and decreased broad spectrum antimicrobial usage.
R. Ghamrawi, None
A. Kantorovich, None
A. Pallotta, None
J. Sekeres, None
S. Gordon, 3M: Scientific Advisor , Consulting fee
Thoratec: Scientific Advisor , Consulting fee