Allergy Alerts in Computerized Physician Entry

Bates, David

Abstract

Computerized physician order entry (POE) has helped reduce medication errors, and a substantial amount of error reduction stems from its allergy alerting function. However, our experience has taught us that these systems should be improved for safer medication delivery. We recently performed an analysis of the user responsiveness to allergy alerts, and found that it was lower than expected (50% in 1994); moreover it had been declining for several years (to 20% in 1999). In our analysis, we learned several lessons. One is that the allergy screening software may be too richly cross-referenced, and therefore generates excessive alerts. In addition, some allergy alerts are consistently ignored based on accepted clinical practice, but their importance has not been tested. Also, present systems lack stratification of allergy severity, and fail to distinguish between medication tolerance and true allergies. By over-alerting, the overall integrity of the system is degraded. We seek to improve the computer-based intelligence of the allergy alerting process. Our study is designed to accomplish several goals. We will evaluate the frequency and significance of adverse drug events that resulted from overridden allergy alerts. Next, we will determine if distinguishing true allergic reactions from symptoms of medication intolerance will result in improved compliance to allergy alerts. Finally, we aim to design an allergy classification system that incorporates severity analysis that would otherwise be performed off-line by pharmacist oversight. We believe the results of our work will improve patient safety and the overall quality of pharmacotherapy. We also believe that the results of our study will be applicabe to other healthcare systems