Characteristics and consequences of drug-allergy alert overrides
Characteristics and consequences of drug-allergy alert overrides in a computerized physician order entry system
Hsieh TC, Kuperman GJ, Jaggi T, Hojnowski-Diaz P, Fiskio J, Williams DH, et al. Characteristics and consequences of drug-allergy alert overrides in a computerized physician order entry system. J Am Med Inform Assoc. 2004;11:482–91.
Alert fatigue and high override rates for clinical alerts are popular discussion topics among developers and users of clinical information systems. As clinical decision support functionality continues to penetrate health information systems, finding the right balance between too many alerts (creating alert fatigue) and too few alerts (risking patient safety) is being actively investigated.
What reasons are given for overriding drug allergy alerts? How often does overriding a drug allergy alert result in an adverse drug event?
This study was designed to analyze the characteristics of drug allergy alert overrides and to determine how often overriding drug allergy alerts resulted in a preventable adverse drug event. At the study location, allergy documentation was completed by the admitting physician using pick-list or search functionality to populate the patient’s chart with coded allergy information. The type of reaction the patient experienced was stored as free text. Subsequently ordered medications were compared to the patient’s allergy list and generated an on screen allergy alert of “definite” if the drug match was exact or of the same class or “possible” if cross reactivity was the reason for the alert. A free text explanation was required if the provider overrode the alert. The investigators conducted a retrospective chart review of a three month time period to determine the category of drug ordered, the category of override reason, and the drug/drug allergy combinations that triggered the alerts. The records were then evaluated for possible adverse drug reactions which were then characterized as significant, severe, or life-threatening. A set of suggestions for improving drug allergy alerting systems were presented in the discussion.
Allergy alerts (n=7761) were triggered in 16% of admissions during the three month study period. Overrides occurred with 80% of the alerts. A subset of patient cases (n=347) underwent further analysis. In 1% of cases the medication was ordered as part of a desensitization protocol. In 7% of cases the patient never received the ordered medication. Definite (exact drug match) alerts comprised 10% of the overridden alerts. Possible drug allergy interactions based upon structural similarities or same pharmacologic family classification generated 90% of the overridden alerts. Override reasons included “Aware/will monitor” (55%), “Patient does not have this allergy/Tolerates” (33%), and “Patient taking already” (10%). Adverse events related to an overridden alert occurred in 19 patients (6%). None of the adverse events were fatal, 9 were serious and 10 were significant. Gastrointestinal symptoms including nausea and vomiting comprised the majority of the adverse events (63%). Non-exact drug allergy alerts were overridden in 95% of the adverse event reports. Clinical peer reviewers determined that the 19 adverse drug reactions were non-preventable and the original clinician’s decision to override the alert was justified. Of the cases where “Patient does not have this allergy/Tolerates” or “Patient taking already” were provided as override reasons, the patient’s allergy list was updated in only 2 instances.
While the override rate for drug allergy alerts remains unacceptably high, this study provides valuable insight into the reason for high override rates. The majority of alerts were generated from non-exact drug allergy matches and from a small number of medications. The investigators were able to show that the actual number of adverse drug events resulting after alert overrides were small (6%). The adverse events resulted mostly from non-exact drug allergy matches and were deemed non-preventable by clinical reviewers. The clinical reviewers agreed with the original clinician’s decision to override the alert in all cases because the risk of a serious allergic reaction was considered less critical than the patient’s clinical need for the medication.
The results of this article are most useful to clinical information system designers, clinical knowledge database designers, institutional informatics officers, and clinicians. This small, well designed study provides insight and clarity to the problem of high alert override rates. The unjustified overriding of clinical alerts would be concerning, however, in this study, overrides were deemed justifiable. This points to flawed alert functionality not flaws in clinical behavior. The authors provided several suggestions for improving alerting functionality in the discussion section of the paper.
The researchers reviewed a total of 7,761 drug allergy alerts generated by the CPOE system at Brigham and Women’s Hospital in a 3 month period. In total, 80% (6,182) of the alerts were overridden by the physician. The goals of the study included categorizing the reasons for the overrides, identifying when the overrides led to adverse drug events and to suggest methods to improve drug allergy alerting protocols.
A patient’s allergy list is entered into the system by the admitting physician upon admission of a patient into the hospital. When medications are ordered, the system compares the drug with the drugs on the allergy list and issues a “definite allergy” alert if the medication is an exact match or is in the same class as a drug on the allergy list. The system will also alert for “possible allergy” reactions based on known cross-reaction drug mappings. The system requires the clinician to enter a free-form text reason for overriding the alert. The researchers classified these reasons as: “Patient does not have this allergy/Tolerates,” “Aware/Will monitor,” “Patient taking already,” “Desensitization protocol,” and “Other.”
A total of 1,150 patients had overridden alerts. The study found that “90% of the overridden alerts were triggered by nonexact drug/allergy matches, in which the drug and allergy had structural similarities or were in the same family but were not identical.” The most frequent reason given for overrides was “Aware/Will monitor” (55%).
A total of 19 ADEs were identified as being caused by the administration of an overridden drug. However, none of these ADEs were considered by the physician reviewers as being preventable because the potential clinical benefits were judged to outweigh the risks of a serious allergic reaction.
The researchers concluded that “the high override rate appears to be attributable primarily to excessive and inappropriate alerting." Among their recommendations were making a distinction between drug allergies and drug sensitivities/intolerances. Interruptive alerts should be displayed only for true allergic reactions (not sensitivities) and the system should make serious or life-threatening warnings readily distinguishable from less dangerous warnings. Warnings of sensitivities should be displayed in a non-interruptive manner.
Other recommendations included generating alerts only when there is an exact drug match, generating no alert for an order for sulfonamide non-antibiotics when the allergy list indicates a “sulfa” allergy and improving the maintenance of patients’ allergy lists.
As noted by the researchers, it is important for healthcare organizations to include review of the overridden alerts as part of their quality improvement program and take actions to reduce the percentage of overridden alerts by fine-tuning the alerting protocols.