Annotation Guidelines for Medication Errors in Incident Reports: Validation Through a Mixed Methods Approach.

Area of Science:

• Health Informatics
• Patient Safety
• Natural Language Processing

Background:

• Lack of standardized annotation guidelines hinders effective learning from incident reports.
• Accurate data extraction from medication error reports is crucial for improving patient safety.

Purpose of the Study:

• To validate annotation guidelines for medication error incident reports.
• To establish a robust evidence base for these guidelines using mixed methods.

Main Methods:

• A mixed-methods explanatory sequential design was employed.
• Involved annotator training, performance evaluation (97% named entity, 90% attribute accuracy), user feedback, and expert interviews.
• Utilized quantitative and qualitative data for guideline validation.

Main Results:

• Achieved high accuracy in named entity (97%) and attribute (90%) identification.
• Participant feedback provided valuable insights for refining annotation methods.
• The mixed-methods approach yielded significant evidence supporting the annotation guidelines.

Conclusions:

• The developed annotation guidelines are validated for medication error incident reporting.
• The study provides a strong foundation for enhancing patient safety through improved incident report analysis.
• Future research should focus on guideline expansion and external validation.