Reduce Medication Errors by Applying a Probabilistic Model: A Pilot Study

摘要

Medication errors are common, life threatening, costly but preventable. This study was implemented a probabilistic model that can reduce medication errors. 1. Methods: The association rules mining techniques are utilized for 9.6 million prescriptions from 2007 to 2011 of three hospital claimed database. The dataset included 68.6 million ICD9-CM diagnoses codes and 91.2 million ATC medications codes. Disease-Medication (DM) and Medication-Medication (MM) associations were computed with their associations' strength which being referred as Q values. By considering the number of DMQs and MMQs, we developed the Appropriateness of Prescription (AOP) model that can determine the appropriateness of a given prescription. The AOP model was implemented in CPOE system at TMUH from May 27 to Jun. 1, 2013 via a web-service. The uncommon prescribed medications was shown as a reminder message in CPOE system.