Summary
This presentation outlines the research conducted by Dr. Adam Bouras of Tritonis Inc. on the critical issue of harmonizing global medication data to align with the standards necessary for the Observational Health Data Sciences and Informatics (OHDSI) vocabulary. The core technical challenge addressed is the granularity gap between the fundamental International Nonproprietary Name (INN) system, commonly used in regions like Africa, and the highly detailed, computable standards of RxNorm. To bridge this disparity, the study implemented and compared two methodologies—a conventional RxNorm API pipeline and an innovative approach leveraging the capabilities of the OpenAI Large Language Model (LLM). Testing on a Moroccan medication list demonstrated that the OpenAI model achieved a perfect match rate, substantially surpassing the success rate of the traditional API method across various drug forms. While highly accurate, the high computational cost of using LLMs for large-scale data suggests that future efforts should prioritize hybrid, localized models and policies supporting global harmonization of pharmaceutical regulations.