Generic database for FHIR

Who we are?

I'm Ryzhikov Nikolai

CTO & Co-founder of HealthSamurai

We are startup, pioneering FHIR in Health IT

• niquola (on twitter, github)
• nicola (RIO) (on fhir chat)

Our Beliefs

• We believe, that FHIR eco-system is a future
• We believe, tat generic FHIR platform as a useful brick in this eco-system
• We are proving this hypothesis with our clients, which are in production or going to production soon

Problem/ Challenged

Generic server needs Generic Database

Minimal Requirements

• Transactions (ACID)
• History tracking
• Sophisticated query (FHIR Search, Analytic etc)
• Extensibility (Extensions, Search Params, Compartments)

Approaches

• Relational DB
• Document DB
• Tripple Store
• ~ Graph DB
• Plygot
• Hybryd

Relational

• • more traditional (tools like ORM)
• • flexible queries (analytic)
• • partial updates
• • a lot of tables ~ 1K tables (first version of fhirbase)
• • expensive and complicated insert (resource -> ~10 tables)
• • expensive and complicated search (joins are not for free)
• • extensibility is tricky (dynamic schema)

Document DBs

• • one to one match Resource <-> Document
• • most of Doc Db's distributed and available
• • no ACID transactions (mongo, rethink) manual consistency is a hell
• • support queries thro different Resource Types (joins, window functions etc)

Polyglot: Key-Value + Indexes

For example: casandre or riak + elastic

• • scale & availability
• • flexible
• • complexity
• • consistency

Tripple store

Like datomic

• • extensibility
• • semantic
• • popular implementations
• • performance

Graph DBs

If you know some experiments - please let me know

Our solution - hybrid

Hybrid relational & document storage, using databases json support (especially PostgreSQL)

• fhirbase
• aidbox

Overview

• store FHIR resources in json column
• use fhirpath functions for access to fields
• use functional indexes to speedup queries

---

• Transactions
• Expressive power of SQL
• Advanced features of PostgreSQL indexing, extensions etc

What if not a postgres

SQL standard 2016 includes json type and utility functions

• mysql
• oracle
• mssql

Demo time

• CRUD
• History
• Search
• Indexes
• Terminology

code is available on github.com/niquola/madrid-2017-slides