Northwestern Medicine investigators have developed FHIR-GPT, an AI model that significantly improves the conversion of electronic health record (EHR) data into standardized Fast Healthcare Interoperability Resources (FHIR). This advancement, detailed in NEJM AI, enhances health data interoperability, research, and public health efforts. FHIR-GPT uses large language models to transform clinical data into FHIR resources, facilitating seamless data exchange across different health systems.

The model addresses the challenge of fragmented healthcare data, where each hospital often has bespoke EHR systems. By standardizing data, FHIR-GPT promotes better collaboration and patient care. In tests, it achieved a 90% success rate in matching EHR data with FHIR medication statements, outperforming current systems and improving accuracy across various metrics, such as medication timing schedules and dose quantities.

The system is more accurate, cost-effective, and scalable than existing tools, marking a significant step forward in the standardization of healthcare data. The team aims to further validate and deploy FHIR-GPT across U.S. healthcare systems to enhance data aggregation, improve patient care, and advance health equity. This research was supported by NIH grants and an American Heart Association fellowship.

To read the article at Northwestern Medicine, click here.

To read the entire research article at NEJM, click here.