From Zero to Production FHIR Server in 6 Months at a Large Northeast Healthsystem
summary
A large Northeast system stands up a fully managed cloud FHIR repo on Google Cloud Healthcare API and migrates over 10 TB of data in six months, ingesting ~2 million HL7‑derived bundles (10–15 million resources) with bursts of 20–75 bundles/sec and support for 15,000+ users. Data pipelines use Dataflow/Apache, Python and Liquid templates, favoring parallelized, batched bundle writes for throughput and cost. Patient merges currently use a $everything merge approach, reconciliation runs continuously, and data is encrypted at rest and in transit.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 0:58What makes Northwest Health a large health system
- 3:13Learning from the Past
- 5:57Precursors
- 9:45Lessons Learned
- 14:28Questions
related talks
22mFHIR R4 data ingestion pipelines driven by liquid templates
44mTransforming FHIR Data for horizontally scalable analytics
28mHerding Data to the Cloud
45mHow to solve scaling pains with FHIR and data streaming?
35mMix, Match, Merge
33mHealthcare "big" Data from Data Management to Web Apps
36mSynthea: Massive FHIR Data
54mUntangling Healthcare with Spark and Dataflow