Effect of modifying dose alerts in an electronic health record on frequency of alerts.

Abstract

Results of a study to reduce the number of medication order-entry alerts and perceived alert fatigue by optimizing alert logic are reported. Data on dosage alerts registered throughout a health system over 2 days per study phase (preintervention and postintervention) were collected from the electronic health record. The 5 medications most frequently associated with dosage alerts during computerized prescriber order entry (CPOE) were evaluated for appropriateness in relation to patient-specific characteristics. Additionally, the 10 alerts most frequently marked by prescribers as "inaccurate warning" during alert override were evaluated for appropriateness. Recommendations were made for all alerts deemed inappropriate or unnecessary. The percent change in the number of alerts from the preintervention to the postintervention period was determined. To evaluate clinician perceptions of the alert updates, a pre-post survey was distributed to hospitalists and pharmacists at 1 facility within the health system. Changes were recommended for 8 alerts; 2 alerts within the dosage category overlapped with alerts in the inaccurate-warning group, resulting in a total of 6 recommended changes. Two recommended alert changes were made within the clinical drug information system, and 4 alerts were changed at the health-system level. As a result, a 3.6% dosage alert decrease occurred. The proportion of dose alerts, among all CPOE-generated alerts, decreased after some of the alerts were modified in accordance with institution-specific medication and population evaluations.