Reproduce “Large Language Models for Automating Clinical Data Standardization: HL7 FHIR Use Case” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Alvaro Riquelme, Pedro Costa, Catalina Martinez
Researchers developed a semi-automatic pipeline leveraging large language models (LLMs) and techniques like Retrieval Augmented Generation (RAG) to standardize structured clinical data into HL7 FHIR resources. This pipeline demonstrated that GPT-4o could achieve up to 68.8% attribute-level mapping accuracy in a simulated real-world scenario and 100% resource-level identification in a baseline scenario.
Key insights
- Providing detailed, machine-readable context, especially through JSON schemas, significantly enhances LLM performance by reducing mapping ambiguity and narrowing confidence intervals.
- Semantic clustering can effectively enrich the contextual representation of structured clinical attributes, guiding LLMs toward more relevant FHIR resource subsets.
- Advanced LLMs like GPT-4o demonstrated superior accuracy and stability compared to Llama 3.2 405b for complex clinical data standardization tasks, even in challenging real-world data environments.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.