| Related Articles |
Prescriber response to computerized drug alerts for electronic prescriptions among hospitalized patients.
Int J Med Inform. 2017 Nov;107:70-75
Authors: Zenziper Straichman Y, Kurnik D, Matok I, Halkin H, Markovits N, Ziv A, Shamiss A, Loebstein R
Abstract
BACKGROUND: Clinical decision support systems (CDSS) reduce prescription errors, but their effectiveness is reduced by high alert rates, "alert fatigue", and indiscriminate rejection.
OBJECTIVES: To compare acceptance rates of alerts generated by the SafeRx(®) prescription CDSS among different alert types and departments in a tertiary care hospital, identify factors associated with alert acceptance, and determine whether alert overrides were justified.
METHODS: In a retrospective study, we compared acceptance rates of all prescription alerts generated in 2013 in 18 departments of Israel's largest tertiary care center. In a prospective study in 2 internal medicine departments, we collected data on factors potentially associated with alert override, and an expert panel evaluated the justification for each overridden alert. We used multivariate analyses to examine the association between patient and physician-related factors and alert acceptance.
RESULTS: In the retrospective study, of 390,841 prescriptions, 37.1% triggered at least one alert, 5.3% of which were accepted. Acceptance rates ranged from 7.9% for excessive dose alerts to 4.0% for duplicate drug and major drug-drug interactions alerts (p<0.001). In the prospective study, common reasons for alert overriding included "irrelevance to the specific condition" and "medication previously tolerated by the patient". Weekend shifts (incident rate ratio [IRR]=1.50 [95% CI, 1.01-2.22]) and a specific department (IRR=1.87 [1.23-2.87]) were associated with higher alert acceptance, while night shift (IRR=0.47 [0.26-0.85]) was associated with alert override. Most alert overrides (88.6%) were judged justified.
CONCLUSIONS: The vast majority of SafeRx(®) alerts are overridden, and overriding is justified in most cases. Minimizing the number of alerts is essential to reduce the likelihood of developing "alert fatigue". Our findings may inform a rational, department-specific approach for alert silencing.
PMID: 29029694 [PubMed - in process]