Developing EMR-based algorithms to Identify hospital adverse events for health system performance evaluation and...

Area of Science:

• Health Informatics
• Patient Safety Research
• Clinical Data Analytics

Background:

• Current adverse event (AE) detection relies on suboptimal administrative data lacking clinical detail.
• Electronic Medical Records (EMRs) offer comprehensive patient data for improved AE surveillance.
• Methodological tools for EMR-based AE detection require development and validation.

Purpose of the Study:

• To develop algorithms for detecting adverse events (AEs) using hospital EMR data.
• To assess the validity of these EMR-based AE detection algorithms within Canadian healthcare settings.

Main Methods:

• Developed AE algorithms by mapping AEs from literature/experts to EMR free text using Natural Language Processing (NLP).
• Validated algorithms on 10,000 EMRs, comparing NLP-identified AEs against trained reviewer chart analyses.
• Assessed algorithm performance using standard indicators like sensitivity, specificity, and PPV.

Main Results:

• Developed and validated novel NLP-driven algorithms for in-hospital AE detection from EMR data.
• Demonstrated the potential for improved accuracy and timeliness in AE surveillance compared to traditional methods.

Conclusions:

• The developed EMR-based AE algorithms can be implemented in healthcare systems for accurate and timely in-hospital AE detection.
• This approach enhances patient safety surveillance by leveraging rich clinical information within EMRs.