Development of case-based medication alerting and recommender system: a new approach to prevention for medication error.

The purpose of this study was to develop a new alerting and recommender system for preventing medication errors. In recent years, alerting systems have been widely implemented, but because these systems apply a same static threshold for all patients in all cases, they produce excessive alerts and subject physicians to "alert fatigue". We believe that the most commonly-written prescription for a patient's status is the safest one. From this standpoint, we developed a real-time case-based medication alerting and recommender system linked to a database of past prescriptions. When a physician issues his or her prescription, our system dynamically compares it with past ones for similar patients in the database. An analysis of the 10 most frequently-used drugs in the University of Tokyo Hospital revealed that our system reduced the number of false alerts compared to the traditional static alert method. Our system contributes to the creation of alerts that are appropriate for patients' clinical conditions and based on physicians' empirical discretion.