Since the law of patient safety came into force in Denmark in 2004 it has been required by law to report adverse events to the Danish Patient Safety Database (DPSD). 35.000 incidents were reported in 2010, which is a tenfold increase since the introduction of the law. Adverse events are currently classified by a manual retrospective review by risk managers and employees at Patientombuddet.
This project investigated the potential of using automatic text classification to categorize IT-related adverse events reported to DPSD. A system called UTH-Finder was developed, which is able to process and classify DPSD reports by using a Naive Bayes algorithm.
A balanced training set containing 482 reports (241 IT-related and 241 other events) and three test set each containing 1.000 reports was used to test the system. The system classified IT-related adverse events with an accuracy and specificity of 95 %.
With the development of UTH-Finder it is concluded that automatic text classification has a potential to be used in categorization of adverse events from DPSD.