Breaking Digital Health Barriers Through a Large Language Model-Based Tool for Automated Observational Medical Outcomes Partnership Mapping: Development and Validation Study.

BACKGROUND

The integration of diverse clinical data sources requires standardization through models such as Observational Medical Outcomes Partnership (OMOP). However, mapping data elements to OMOP concepts demands significant technical expertise and time. While large health care systems often have resources for OMOP conversion, smaller clinical trials and studies frequently lack such support, leaving valuable research data siloed.

OBJECTIVE

This study aims to develop and validate a user-friendly tool that leverages large language models to automate the OMOP conversion process for clinical trials, electronic health records, and registry data.

METHODS

We developed a 3-tiered semantic matching system using GPT-3 embeddings to transform heterogeneous clinical data to the OMOP Common Data Model. The system processes input terms by generating vector embeddings, computing cosine similarity against precomputed Observational Health Data Sciences and Informatics vocabulary embeddings, and ranking potential matches. We validated the system using two independent datasets: (1) a development set of 76 National Institutes of Health Helping to End Addiction Long-term Initiative clinical trial common data elements for chronic pain and opioid use disorders and (2) a separate validation set of electronic health record concepts from the National Institutes of Health National COVID Cohort Collaborative COVID-19 enclave. The architecture combines Unified Medical Language System semantic frameworks with asynchronous processing for efficient concept mapping, made available through an open-source implementation.

RESULTS

The system achieved an area under the receiver operating characteristic curve of 0.9975 for mapping clinical trial common data element terms. Precision ranged from 0.92 to 0.99 and recall ranged from 0.88 to 0.97 across similarity thresholds from 0.85 to 1.0. In practical application, the tool successfully automated mappings that previously required manual informatics expertise, reducing the technical barriers for research teams to participate in large-scale, data-sharing initiatives. Representative mappings demonstrated high accuracy, such as demographic terms achieving 100% similarity with corresponding Logical Observation Identifiers Names and Codes concepts. The implementation successfully processes diverse data types through both individual term mapping and batch processing capabilities.

CONCLUSIONS

Our validated large language model-based tool effectively automates the transformation of clinical data into the OMOP format while maintaining high accuracy. The combination of semantic matching capabilities and a researcher-friendly interface makes data harmonization accessible to smaller research teams without requiring extensive informatics support. This has direct implications for accelerating clinical research data standardization and enabling broader participation in initiatives such as the National Institutes of Health Helping to End Addiction Long-term Initiative Data Ecosystem.