Storage and analytics
summary
Compares storage strategies with concrete tradeoffs and working code: storing whole resources as blobs plus search parameter tables causes heavy reindexing and triple storage, while document stores like MongoDB lose SQL features and transactions. Shows an alternative using Postgres JSONB and functional indexes via fhirbase (Postgres + JSONB) to run FHIRPath-like queries in-database, avoid joins, and add selective indexes. Demonstrates falling back to SQL, describes Sequel-on-FHIR transformations (structured references, nested typed values, first-class extensions) and points to Bunsen, BigQuery pipelines for analytics.
AI-written summary · about
chapters (31)
- 0:00Introduction
- 1:45Relational schema
- 3:32Most popular schema
- 4:52Search parameters attributes
- 5:21Search parameter tables
- 6:15Reindexing
- 7:04Indexing
- 7:37Multiple Search Parameters
- 7:49Selective Search Parameters
- 8:58Document Databases
- 10:14Pros of Document Databases
- 10:37Cons of Document Databases
- 11:24Binary Jason
- 12:28Schema
- 12:57Postgres
- 13:43Firesearch
- 14:38JSON Life Extension
- 14:49Fire Path
- 15:29Index Expressions
- 16:35Index on Application Layer
- 16:54Magic in Postgres
- 17:32GS Query
- 18:01Summary
- 18:59Second API
- 20:22Sequel on Fire
- 21:06Bunsen
- 22:49Babybox
- 24:28Firstclass extensions
- 25:31Firebase tutorials
- 26:30Whats next
- 26:44Ask questions
related talks
29mSQL on FHIR 2.0
45mFHIR server scale and performance challenges
35mScalable Data Science with FHIR
01h 00mThe Modern Database Debate: Documents, Tables, or Documents in Tables?
44mTransforming FHIR Data for horizontally scalable analytics
55mThe Great Data Debate 2022
39mHow to integrate FHIR with your analytics pipeline
25mHybrid Applications with MongoDB and RDBMS