Screening for Medication Errors and Adverse Events Using Outlier Detection Screening Algorithms in an Inpatient...

Area of Science:

• Medical Informatics
• Clinical Decision Support Systems
• Patient Safety

Background:

• Medication-related risks pose a significant threat to patient safety in inpatient settings.
• Existing clinical decision support systems (CDSS) may generate high alert burdens, impacting physician workflow.
• Novel approaches are needed to accurately identify and mitigate medication risks.

Purpose of the Study:

• To evaluate a novel system utilizing outlier detection algorithms for identifying medication-related risks.
• To assess the accuracy, clinical relevance, and impact of system-generated alerts in an inpatient environment.

Main Methods:

• A two-phase study (retrospective and prospective) was conducted in a medical center.
• Outlier detection models were initially evaluated and then fine-tuned to local practice patterns.
• All generated alerts were reviewed by a clinical team for accuracy and relevance.

Main Results:

• In the retrospective phase, 1.2% of orders generated alerts, with 69% being clinically relevant.
• In the prospective phase, 1.6% of orders generated alerts, 72% of which were clinically relevant.
• 41% of prospective alerts led to changes in prescriber behavior, indicating impact on practice.

Conclusions:

• The novel system demonstrated potential in identifying medication risks in an inpatient setting.
• The system provided accurate and clinically valid alerts with a low alert burden.
• This CDSS-naïve system facilitated improvements in daily medical practice for physicians.