A text mining approach to categorize patient safety event reports by medication error type

Area of Science:

• Health Informatics
• Medical Informatics
• Patient Safety

Background:

• Patient safety reporting systems capture medication errors but often lack analysis.
• Undetected safety hazards pose risks in healthcare.

Purpose of the Study:

• To evaluate natural language processing (NLP) for improved categorization of medication-related patient safety event reports.
• To assess machine learning algorithms for identifying specific medication error types.

Main Methods:

• Utilized 3,861 annotated medication safety reports.
• Developed and tested three models: logistic regression, elastic net, and XGBoost.
• Analyzed model performance across predefined medication error categories.

Main Results:

• The XGBoost model demonstrated superior performance in categorizing medication errors.
• 'Wrong Drug', 'Wrong Dosage Form or Technique or Route', and 'Improper Dose/Dose Omission' were highly identifiable categories.
• Identified key terms associated with each error type and common co-occurring error categories.

Conclusions:

• Machine learning offers a semi-automated approach to classifying medication errors from free-text reports.
• Improved categorization can enhance the identification of critical medication safety patterns and trends.
• NLP techniques hold potential for advancing medication safety surveillance.