Screening for medication errors using an outlier detection system.

Area of Science:

• Health Informatics
• Clinical Decision Support Systems
• Machine Learning in Healthcare

Background:

• Existing clinical decision support (CDS) systems may miss certain medication errors.
• Accurate and valid alerts are crucial for effective patient safety.

Purpose of the Study:

• To evaluate the accuracy, validity, and clinical usefulness of medication error alerts from an outlier detection screening system.
• To assess the performance of a novel CDS approach for identifying potential medication errors.

Main Methods:

• Extracted 5 years of clinical data from electronic health records for over 747,000 patients.
• Screened data using an outlier detection system to identify potential medication errors.
• Reviewed 300 generated alerts to determine accuracy, validity, and clinical usefulness.

Main Results:

• Three-quarters of the alerts generated by the screening system were valid, identifying potential medication errors.
• The majority (75.0%) of valid alerts were deemed clinically useful.
• The system identified potential errors that might be missed by current CDS systems.

Conclusions:

• Outlier detection screening can generate useful medication error alerts with a modest false positive rate.
• The system's effectiveness relies on high-quality and complete electronic health record data.
• This approach offers a valuable supplement to existing CDS systems for enhancing patient safety.