OMOP Extraction of Medical Text Using LLMs: Preliminary Results.

Area of Science:

• Medical Informatics
• Natural Language Processing
• Health Data Standards

Background:

• Electronic health records (EHRs) contain valuable medication information.
• Standardizing medication data is crucial for clinical research and data analysis.
• The Observational Medical Outcomes Partnership Common Data Model (OMOP CDM) provides a standardized framework.

Purpose of the Study:

• To evaluate the feasibility of using medium-sized large language models (LLMs) for medication data extraction.
• To assess the accuracy and performance of LLMs in harmonizing medication data into the OMOP CDM from German discharge letters.

Main Methods:

• Utilized medium-sized large language models (LLMs) for automated text processing.
• Extracted medication information from a corpus of German discharge letters.
• Mapped extracted data to the OMOP Common Data Model (CDM) structure.

Main Results:

• Achieved extraction accuracy exceeding 85%.
• Obtained an F1-score greater than 75% for medication data extraction.
• Demonstrated successful harmonization of medication data into the OMOP CDM.

Conclusions:

• Medium-sized LLMs are effective tools for extracting and standardizing medication data.
• The OMOP CDM can be populated with high accuracy using LLM-based extraction from clinical notes.
• This approach shows promise for improving health data interoperability and research.