Reduction of clinical support warnings through recategorization of severity levels.

PURPOSE Study results indicating that non-clinically significant alerts generated by clinical support systems can be reduced by recategorizing alert severity levels, thereby decreasing alert fatigue, are presented. METHODS In a single-site retrospective pre-post study, all drug-drug interaction alerts displayed by a medical center's clinical support system over a three-month baseline period were evaluated by a multidisciplinary expert panel. Based on a review of pharmaceutical package inserts and other evidence, the panel reduced the severity ranking of 99 of the 201 most frequently triggered alerts with the aim of reducing nonactionable alerts (i.e., alerts viewed but not resulting in a clinical intervention such as the discontinuation of a medication, a dose reduction, or the ordering of a laboratory test). Alert frequencies and types during the baseline period and a one-month period after adjustment of the alert severity rankings were compared to characterize the distribution of alert override responses. RESULTS Comparison of drug-drug interaction alerts before (n = 8023) and after (n = 7270) alert recategorization indicated significant differences in pharmacists' documentation of override responses in four evaluated categories (p < 0.001 for all comparisons); notably, alerts overrides in the "not clinically significant" category declined 22%. No significant difference was detected in the numbers of reported medication errors related to clinically significant drug-drug interaction alerts before and after alert recategorization. CONCLUSION Reducing the number of nonactionable drug-drug interaction alerts by recategorizing alert severity levels decreased alert overrides by 6%.