The Overriding of Computerized Physician Order Entry (CPOE) Drug Safety Alerts Fired by the Clinical Decision Support (CDS) Tool: Evaluation of Appropriate Responses and Alert Fatigue Solutions

Abstract

Introduction Most computerized physician order entry (CPOE) software come with clinical decision-support components (CDS) that provide prescribers assistance and notify them about adverse drug reactions. An excessive number of alerts in a repeated and non-relevant manner leads to alert fatigue and enforces physicians and pharmacists to alert overrides. King Abdulaziz Medical City (KAMC) in Jeddah still reports a higher percentage of drug alerts overridden by clinicians and pharmacists. Thus, this study was conducted to evaluate CDS alerts overriding and to determine which alerts are clinically irrelevant and need modifications. Methods The study was carried out in the inpatient setting at KAMC in Jeddah, from September 1, 2020, to December 31, 2020. It was designed as a retrospective chart review study that included all red alerts that required comments and were overridden by a physician and pharmacist. Results Among 11350 red alerts, potential drug-drug interaction (pDDI), dose, and allergy alerts represent 57%, 41%, and 2%, respectively, of the total alerts. The most common drug-drug interactions (DDIs) in category X were proton pump inhibitors and clopidogrel (9.9%). The appropriate response by prescribers and pharmacists toward allergy alerts was associated with the highest odds compared with the other alerts (p < 0.05) while there is a significant decrease in the odds of appropriate action being taken by both prescribers and pharmacists in the dose screen alerts (p < 0.05). Among all clinical specialties, there is an increased odds of appropriate action being taken by residents and fellows for allergy and dose alerts, respectively, compared to other groups (p < 0.05). For diminishing the unnecessary alerts, we provided 14 alert refinements strategies and advised turning off four alerts. Applying this will terminate 32% of irrelevant alerts. Conclusion Our study's findings indicated that a substantial number of alerts are ignored, and the rate of appropriateness varies significantly by alert type and prescriber level.