FHIR-GPT Enhances Health Interoperability with Large Language Models Article Swipe
YOU?
·
· 2023
· Open Access
·
· DOI: https://doi.org/10.1101/2023.10.17.23297028
· OA: W4387861191
Advancing health interoperability can significantly benefit health research, including phenotyping, clinical trial support, and public health surveillance. Federal agencies, including ONC, CDC, and CMS, have been collectively collaborating to promote interoperability by adopting Fast Healthcare Interoperability Resources (FHIR). However, the heterogeneous structures and formats of health data present challenges when transforming Electronic Health Record (EHR) data into FHIR resources. This challenge becomes more significant when critical health information is embedded in unstructured data rather than well-organized structured formats. Previous studies relied on multiple separate rule-based or deep learning-based NLP tools to complete the FHIR resource transformation, which demands substantial development costs, extensive training data, and meticulous integration of multiple individual NLP tools. In this study, we assessed the ability of large language models (LLMs) to transform clinical narratives into HL7 FHIR resources. We developed FHIR-GPT specifically for the transformation of clinical texts into FHIR medication statement resources. In our experiments using 3,671 snippets of clinical texts, FHIR-GPT demonstrated an exceptional exact match rate of over 90%, surpassing the performance of existing methods. FHIR-GPT improved the exact match rates of existing NLP pipelines by 3% for routes, 12% for dose quantities, 35% for reasons, 42% for forms, and over 50% for timing schedules. Our findings provide the foundations for leveraging LLMs to enhance health data interoperability. Future studies will aim to build upon these successes by extending the generation to additional FHIR resources.
