Extraction of Active Medications and Adherence Using Natural Language Processing for Glaucoma Patients

Area of Science:

• Health Informatics
• Clinical Informatics
• Medical Record Management

Background:

• Medication data accuracy in electronic health records (EHRs) is vital but frequently compromised by errors in medication lists.
• Physicians often document detailed medication information within clinical notes, presenting a potential source for improving list accuracy.
• Extracting current medications from unstructured clinical narratives poses a significant challenge for automated systems.

Purpose of the Study:

• To investigate discrepancies in medication documentation between medication lists and progress notes for glaucoma patients.
• To develop and validate a named entity recognition (NER) model for identifying current medications and adherence from clinical progress notes.
• To demonstrate a prototype tool for medication reconciliation leveraging the developed NER model.

Main Methods:

• Manual chart review to identify discrepancies in medication documentation for glaucoma patients.
• Development and validation of a named entity recognition model tailored for extracting medication and adherence information from free-text clinical notes.
• Creation of a prototype tool to showcase the practical application of the NER model in medication reconciliation.

Main Results:

• Significant discrepancies were observed between documented medication lists and information present in progress notes.
• The developed named entity recognition model demonstrated effectiveness in identifying current medications and adherence from clinical narratives.
• A functional prototype tool was successfully demonstrated, highlighting the potential for automated medication reconciliation.

Conclusions:

• Clinical notes contain valuable, often more accurate, medication information compared to structured lists.
• The developed NER model shows promise for improving the accuracy of medication data within EHR systems.
• Integration of such models into EHRs could facilitate real-time medication reconciliation, enhancing patient safety and research integrity.