Journal study evaluates success of automated machine learning system to prevent medication prescribing errors

The Joint Commission Journal on Quality and Patient Safety published a study on the ability of a machine learning system to identify and prevent medication prescribing errors not previously identified by and programmed into an existing clinical decision support (CDS) system.

The Joint Commission Journal on Quality and Patient Safety published a study on the ability of a machine learning system to identify and prevent medication prescribing errors not previously identified by and programmed into an existing clinical decision support (CDS) system. Alerts were generated retrospectively by a machine learning system using existing outpatient data from Brigham and Women's Hospital and Massachusetts General Hospital in Boston from 2009 through 2013. The study analyzed whether the system—a platform based on advanced machine learning algorithms—generated clinically valid alerts, which were compared with alerts in the CDS system using a random sample of 300 alerts selected for medical record review. Findings indicated there were a total of 10,668 alerts during the 5-year period. Overall, 68.2% of the alerts would not have been generated by the existing CDS system, 80% were clinically valid, and 92% of a random sample of chart-reviewed alerts were accurate based on structured data available in the record. The estimated cost of adverse events potentially prevented in an outpatient setting was more than $60 per drug alert and $1.3 million when extrapolating the study's findings to the full patient population.