Using Graph-theory to empower patients
summary
Uses a directed acyclic graph over R4 FHIR resources to partition an individual’s record by source types (condition, encounter, EOB) and terminate on sinks (location, device, org), producing bounded subgraphs that answer concrete questions like “which meds, labs, practitioners relate to condition X.” Implementation is in an open‑source PHR (Go, standard graph library, not Neo4j), processes on the order of ~10,000 FHIR resources per patient, uses predecessor and adjacency maps, caches results for performance, and is practical for single‑patient views but costly for large cohorts.
AI-written summary · about
chapters (20)
- 0:00Introduction and speaker intro
- 1:51Personal health record challenges
- 4:41What is graph theory?
- 7:36FHIR resources and relationships
- 8:38FHIR relationship graphs explained
- 10:04Directed acyclic graphs and sources/sinks
- 12:11Creating subgraphs for conditions
- 13:13Adjacency and predecessor maps
- 15:17Generic algorithm for graph traversal
- 16:42Answering patient questions with graphs
- 17:27Medication history and contacts
- 18:30When graph theory is overkill
- 19:55Graph theory benefits and trade-offs
- 21:20Fasten Health project overview
- 22:01Q&A session begins
- 33:46EHR vs Fasten Health differences
- 35:50Future plans and accessibility
- 36:54Financial data and claims integration
- 37:58Static graph definitions in FHIR
- 38:41Next steps for patients
related talks
47mModeling a clinical scenario
17mEnhancing Benefit Adjudication Through Graph Node Embedding, Clustering, and Outlier Detection
32mUnderstanding Complexly Connected Data using Graph Databases
21mDemocratising FHIR with GraphQL
27mSupergraph Saving Lives
01h 41mGraph Intelligence: Enhance Reasoning and Retrieval Using Graph Analytics
29mA Graph Entity Resolution Playbook
48mArgonaut: Patient Lists