RESUMO
PURPOSE: Medication-error alerts for warfarin orders detected by a bar-code-assisted medication administration (BCMA) system were evaluated. METHODS: All patients receiving warfarin who were admitted to a university medical center between July 1, 2008, and February 6, 2009, in inpatient units with BCMA systems were candidates for inclusion in this study. Medication-error alerts displayed to the nurse administering the warfarin were reviewed to determine whether a true potential error was detected. Each alert was converted to a scenario, and its potential to require treatment or cause patient harm was rated using a validated severity scale of 0-10, where a score of 0 indicated no probable effect on the patient and 10 indicated that the error would likely result in patient death. A severity score was obtained by averaging the scores of four pharmacist reviewers. RESULTS: Of the 18,393 warfarin doses ordered during the study period for 2,404 patients, error alerts associated with only 99 warfarin doses were found to be clinically meaningful. The mean ± S.D. severity rating of these alerts was low (2.93 ± 1.42), with a standardized Cronbach's coefficient alpha of 0.845. The mean ± S.D. warfarin dose attempted when the nurse received an alert was 4.10 ± 2.48 mg. The majority of doses with alerts (70%) were for patients who had an active order for warfarin. CONCLUSION: Of the large number of medication-error alerts generated through a BCMA system, only a small proportion were considered clinically significant. This indicated that the rate of false-positive alerts was unexpectedly high, increasing the risk of alert fatigue.