AI Agent Extracts Medical Records from Hospital: A Case Study Based on FHIR and the 21st Century Cures Act
I gave an AI agent OAuth access to my hospital. It worked on the first try.
Summary
The author used an open-source health agent, 'Tula,' to extract their full medical history from their hospital (Beverly Hospital). This demonstrated that data access based on FHIR standards is possible through a simple text message and authentication process, without needing API integration or complex procedures.
Details
The article reports on the author's successful extraction of their complete medical history using an open-source health agent called 'Tula.' The system functions as a SMART on FHIR client skill set, leveraging patient rights for data access mandated by the 21st Century Cures Act. The process involved sending a request to the agent, authenticating via MyChart through a provided URL, and subsequently receiving encrypted FHIR R4 JSON batch data from the hospital side, which was then decrypted and saved locally. This method is highly accessible, requiring no API integration or major system overhaul, with an estimated low running cost of ~$30/month. The author highlights that real medical records present significant challenges, such as variations in data formats across different EHR endpoints, inconsistencies in LOINC codes, or edge cases like PDF attachments. Tula is built upon existing technology pioneered by key figures in SMART on FHIR, emphasizing its ability to handle real-world data extraction. The piece serves as a demonstration of the patient's right (under the 21st Century Cures Act) to access medical records not merely as a 'system dump,' but in a format that allows for longitudinal questioning using open-source AI agents, suggesting significant potential for future healthcare IT applications.
Original content copyright by respective publishers