Factors Leading to False Positive Computerized Provider Order Entry (CPOE) Opiate Allergy Alerts

Abstract

Background: Computerized provider order entry (CPOE) drug alert overrides generally exceed 80%, of which opiate allergies are significant contributors. With an increasingly complex, information dependent healthcare culture, providers do not have unlimited cognitive capacity to interpret and effectively act on high volumes of low value alerts. Strategies are needed to improve clinical value and reduce alert volume, both of which are significant factors in intentional and unintentional alert overrides. Purpose: This study examined the frequency and characteristics of opiate allergy alerts, and the influence of patient characteristics, reaction/severity, and provider role on alert override. Findings will be used to improve future allergy alerts. Method: This was a retrospective, quantitative analysis of all FY10 adult opiate allergy alerts, related orders and patient characteristics at a large academic medical center. Three reaction/severity groups were created: (1) Non-Allergic/Low Severity (NALS=15%), (2) Unknown reaction with Unknown severity (16%), and (3) All else (69%). Data quality limited the NALS group to gastrointestinal reactions only (nausea, constipation, etc.). Effect of age, race, gender, visit type, provider type, and reaction/severity on the likelihood of overriding the patient's first opiate alert (alert1) was analyzed using Generalized Estimating Equations (GEE). Results: Over half of all patients had opiates ordered. Those orders alone triggered 25461 allergy alerts for 2767 patients. This represented, minimally, a 9% opiate allergy prevalence across all inpatients (2767/30321). Opiate reaction severity was 2.5% mild, 6.5% moderate, 16.8% severe and 74.3% unknown. Codeine was rarely ordered, but accounted for 32% of the alerts due to cross reactivity algorithms. Of the factors tested, only prescriber role had a significant influence on alert1 overrides. Advanced practice nurses were generally less likely to override alert1 (80% OR) as compared with physicians (90% OR, GEE B= -.793, p=.001). Conclusion: Drug allergy alerting was one of the earliest and supposedly simplest forms of CPOE clinical decision support, yet has failed to attain acceptable override rates. Explicit allergy definitions, staff training, allergy data entry decision support and CPOE (GI) symptom management for all medications could significantly reduce alert volume and improve patient care.