Supercharge FHIR Search with Large Language Models (LLMs)
summary
Demonstrations show two concrete patterns: push FHIR bundle into LLM to answer timeline and medication questions and produce concise patient summaries, and generate FHIR search/SQL from natural language to translate clinician queries into executable queries for FHIR servers or BigQuery. Prototypes run on Google Vertex AI and Healthcare API, include demo GIFs that produce correct timelines and summaries, and surface privacy and validation needs — working as a proof‑of‑concept, not a production solution.
AI-written summary · about
chapters (22)
- 0:00Introduction and welcome
- 1:50LLM use cases with FHIR
- 4:58Natural language queries in healthcare
- 6:22Approach one: Direct FHIR data to LLM
- 8:29Approach two: Generate FHIR queries with LLM
- 10:35Future of conversational FHIR search
- 11:40Google’s healthcare AI products
- 12:47Q&A session begins
- 13:09Benefits of FHIR in query generation
- 14:55Generative AI for FHIR updates
- 16:41Security and access control concerns
- 21:01End of Q&A and session wrap-up
- 26:08Introduction to next talk
- 27:12Automated detection of vaccine adverse events
- 30:06FDA BEST initiative overview
- 32:17Pilot phase data flow
- 33:44Automated adverse event reporting
- 35:51Clinical Query Language details
- 38:04Data staging and FHIR bundle creation
- 41:23Pilot success and future plans
- 43:33Next steps for blood product reporting
- 44:14Final Q&A and closing
related talks
33mOpening Keynote: LLMs, between fantasy and reality - the poin...
29mKeynote: LLMs, what are they good for?
49mImproving Accessibility through Leveraging Large Language Models (LLMs)
27mBuilding Useful Apps With Small, Local LLMs
40mTechniques to avoid common LLM pitfalls
25mHuman-Centered AI: What I've Learned From Building Two LLM-Powered Products Thi...
36mProduction-Ready AI Agents: From LLMs to Small Language Models
47mGive Your LLMs a Left Brain