Healthcare data platform reference architecture
This is the architecture DAX builds toward on every engagement. It is published in full because there is nothing proprietary about the shape of it — the value is in executing it against real payer files and real EHR exports.
Your environment. Your data. Your foundation.
Six layers, bottom to top
Source Systems
- EHR
- Claims
- Payer
- ADT / HIE
- CMS
- Operational
Ingestion
- API
- HL7
- FHIR
- SFTP
- Files
Owned Foundation
Deployed inside your environment.
- Raw
- Quality
- Identity
- Normalization
- Healthcare Models
- Warehouse / Lakehouse
Intelligence
- Patient
- Risk
- Quality
- Utilization
- Attribution
- Contract
- Financial
Application Choice
Interchangeable by design.
- DAX
- Existing Applications
- BI
- Custom Applications
- AI
Operations
- Care Management
- Quality
- HCC
- TCM
- Utilization
- Provider
- Contract
What each layer is responsible for
Source systems
EHR, payer claims, ADT/HIE, CMS files, and operational systems. The architecture assumes these disagree with each other, because they do.
Clinical integrationIngestion
API pulls, HL7 interfaces, FHIR resources, SFTP drops, and flat files — each with source-aware validation, durable raw capture, and replay.
Claims integrationOwned foundation
Raw landing, quality checks, identity resolution, normalization, healthcare domain models, and the warehouse or lakehouse itself. This is the part that has to be yours.
Warehouse scopeIntelligence
Versioned, attribution-aware semantic models for patient, risk, quality, utilization, contract, and financial measures — defined once, consumed everywhere.
Application choice
DAX's application, your existing tools, your BI stack, something your team builds, or an AI layer that does not exist yet. Interchangeable by design.
VBC applicationOperations
Care management, quality, HCC, TCM, utilization, provider performance, and contract management — the work the whole stack exists to support.
The constraints that make it an asset
Deployed in your environment
Your cloud account, your subscription, your credentials, your network boundary. Not a tenant inside somebody else's platform.
Raw data is retained
Every source file and message is kept in its original form. When a definition changes, you reprocess history instead of asking a vendor to.
Identity is resolved once
One enterprise patient identity across clinical, claims, and ADT, with crosswalks retained. Everything downstream depends on getting this right.
Definitions are versioned
A metric means the same thing in the dashboard, the extract, and the board deck — and you can see when and why it changed.
Quality is visible, not hidden
Completeness, lag, and exceptions surface to the people reading the numbers rather than being smoothed over before they see them.
The application tier is replaceable
Nothing in the foundation depends on which application reads it. That constraint is what makes the asset durable.
Your data team gets the keys.
DAX is not here to replace your team. We accelerate the architecture, bring healthcare-specific patterns, build alongside your engineers, document as we go, and hand over the operational knowledge deliberately.
We are not building a dependency on DAX. We are helping your team own and operate a capability the organization previously rented.
Bring your architect to the call.
This conversation goes better with the person who will inherit the environment in the room. Come with your source list, your cloud constraints, and your hardest reconciliation problem.