SQL on FHIR
summary
Open-sourced at google/fhir, the project auto-generates protocol buffers from FHIR StructureDefinitions, hoists profiled extensions and codings into top-level fields and emits an analytic JSON for BigQuery so queries like “observations with specific LOINC or SNOMED” or “top mother’s maiden names by birth city” become simple SQL. Libraries exist in Java, C++, Python (Go coming), build rules use Bazel, conversion is lossless JSON↔proto, reduces storage ~10x, and converting and uploading a profiled resource can take two lines of code.
AI-written summary · about
chapters (21)
- 0:00Intro
- 1:30Download and Try
- 3:21BigQuery on FHIR Data
- 4:52Hard Queries
- 7:55Message Definitions compile into code
- 8:18Protocol Buffers - using generated code
- 8:49Protocol Buffers: Serialization
- 10:10Approach
- 13:44Primitive Types - Date Time
- 15:05Converting to/from JSON
- 15:54Inline CodeSystems
- 16:49A Patient Example
- 18:58Example: Extension
- 20:44Example: CodeableConcept
- 24:38Example: US-Core Ethnicity
- 25:15Example: US-Core Patient Profile
- 26:03Profiling Resources
- 27:43Querying Extensions
- 28:08Inlined CodeableConcepts
- 32:51Profiles Example
- 33:58Takeaways
related talks
29mSQL on FHIR 2.0
20mFHIR SQL Builder
47mStoring data using FHIR - a good idea?
34mFHIR Projects
24mUnleashing the power of FHIR data with knowledge graphs
43mEnglish introduction to FHIR
33mLearn to FSH: A friendly introduction to FHIR Shorthand
20mFHIR for scientific data: implementing the FAIR principles