Palo Alto, Calif.-based Stanford Health Care created a ChatGPT-like tool that allows clinicians to ask questions in a patient’s EHR.
Here is how the health system developed ChatEHR, whose user interface is embedded directly into Epic, according to an Oct. 22 Stanford article:
1. Stanford Health Care decided on Fast Healthcare Interoperability Resources, or FHIR, as the most effective data retrieval source and engineered a near-real-time data foundation.
2. The health system turned raw FHIR data into formats optimized for large language model processing, with multiple concurrent data domains including medications, labs and procedures and subsecond response times.
3. Stanford opted to convert system-level FHIR data into a clinician-friendly format before displaying it and preserve metadata for auditing and traceability.
4. The organization deployed a self-hosted gateway for all LLM interactions, centralizing authorization, logging and monitoring.
