Region Stockholm has conducted a feasibility study to explore how AI large language models (LLMs) can support the mapping of medical terms to standardized code systems e.g. Snomed CT. This work is important for both implementating new electronic health record (EHR) systems and meeting the requirements of the European Health Data Space (EHDS). In this study, we explored how large language models can interpret clinical terms, identify synonyms with multilingual support, detect relevant Snomed CT hierarchies, and suggest mappings. The results indicate that LLMs can provide accurate suggestions and have the potential to support health informaticians and shorten mapping lead times. The study also examined possible technical solutions and architectural approaches for integrating LLMs into a more efficient and quality-assured workflow.

The overall goal is to contribute to a more scalable and sustainable approach to semantic interoperability in healthcare. Semantic interoperability is essential to enable different systems to exchange and use health information while maintaining its meaning and context. Achieving this requires mapping information to standardized systems such as SNOMED CT, KVÅ, and ICD-10. Currently, the mapping process is manual, time-consuming, and requires specialized expertise.

During the presentation, Region Stockholm will share key insights from the study and outline the next steps in this work. The aim is to inspire innovation and demonstrate the potential of AI to support the digital transformation of healthcare.