A dataset for mapping the Japanese drugs to RxNorm standard concepts

Area of Science:

• Health Informatics
• Observational Health Data Sciences
• Pharmacovigilance
• Natural Language Processing

Background:

• International research relies on standardized data formats like the Observational Medical Outcomes Partnership Common Data Model (OMOP CDM) and terminologies such as RxNorm.
• Japanese real-world data (RWD) is underutilized internationally due to incompatible domestic terminologies and a wide variety of pharmaceutical products.
• Mapping Japanese pharmaceutical terms to RxNorm is crucial for integrating Japanese RWD into global research initiatives.

Purpose of the Study:

• To develop a method for mapping Japanese pharmaceutical terms to RxNorm using a Large Language Model (LLM).
• To create a valuable dataset for researchers in pharmacoepidemiology, pharmacoeconomics, and clinical decision support.
• To enable the use of Japanese RWD in large-scale international observational studies.

Main Methods:

• Utilized a Large Language Model (LLM) to perform ingredient-based mapping of Japanese pharmaceutical data to RxNorm.
• Conducted a sampling-based evaluation to confirm the accuracy of LLM-identified mapping candidates.
• Validated the LLM-generated mappings with pharmacists and a medical informatics researcher.

Main Results:

• Successfully generated an ingredient-based mapping of Japanese pharmaceutical terms to RxNorm.
• The LLM demonstrated high accuracy in identifying potential mapping candidates.
• The resulting dataset includes target drugs, translated names, LLM suggestions, and reference data.

Conclusions:

• LLM-based terminology mapping is effective for integrating Japanese pharmaceutical data into international research standards.
• This work facilitates the utilization of Japanese RWD for pharmacoepidemiology, pharmacoeconomics, and clinical decision support.
• The developed dataset and methodology support the advancement of NLP and machine learning in health terminology mapping.