Azure for HLS

Last reviewed: 2026-06-28 — see the freshness policy.

Learning objectives

After this chapter you will be able to:

• Use Azure Health Data Services as the interoperability layer of an Azure HLS design.
• Place Microsoft Fabric's healthcare data solutions in an analytics architecture.
• Reason about when Azure is the right platform — especially for Microsoft-first health systems.

Azure's HLS positioning

Microsoft's HLS strategy has two pillars: Azure Health Data Services for clinical-data interoperability (the transactional layer), and healthcare data solutions in Microsoft Fabric for analytics (the lakehouse layer). Azure's strongest pull is institutional: many health systems are already deeply invested in Microsoft (Entra ID, Teams, Office 365), and Epic runs on Azure in many deployments — making Azure the path of least resistance for provider organizations.

Azure signs a BAA and documents HIPAA/HITRUST coverage. As always, the platform is covered; your configuration (encryption, RBAC, network isolation, audit) is your responsibility. See HIPAA.

The two pillars

Azure Health Data Services (transactional / interoperability)

A managed, unified service spanning the three clinical modalities:

Healthcare data solutions in Microsoft Fabric (analytics)

An end-to-end SaaS analytics platform on OneLake that ingests, standardizes, and analyzes healthcare data. Key capabilities:

• FHIR ingestion into OneLake from the FHIR service.
• DICOM transformation into OneLake for imaging analytics.
• Text Analytics for Health (Azure AI Language) to structure unstructured clinical notes.
• OMOP transformations to prepare data for standardized analytics (see OMOP CDM).

Fabric unifies structured, imaging, device, and unstructured data in OneLake and exposes it through familiar tools — Power BI, Data Factory, Synapse-style analytics.

Reference architecture: interoperability + analytics on Azure

flowchart LR
  EHR["EHR (HL7v2 / C-CDA / FHIR)"] -->|"$convert-data"| FHIR["AHDS FHIR service"]
  PACS["Imaging"] --> DICOM["AHDS DICOM service"]
  Devices["IoMT devices"] --> MedTech["AHDS MedTech service"] --> FHIR
  FHIR -->|"ingest"| OneLake["OneLake (Fabric)"]
  DICOM --> OneLake
  Notes["Clinical notes"] --> TA["Text Analytics for Health"] --> OneLake
  OneLake --> OMOP["OMOP transform"]
  OneLake --> PBI["Power BI"]
  OneLake --> AOAI["Azure OpenAI / AI Foundry"]
  Audit["Azure Monitor / audit logs"] -.-> FHIR & DICOM

The pattern: Azure Health Data Services is the transactional clinical layer; Microsoft Fabric / OneLake is the analytics lakehouse; Azure OpenAI adds generative AI. The MedTech service makes Azure distinctive for connected-device and remote-monitoring scenarios.

Compute & batch

For genomics, imaging AI, and large ETL: Azure Batch schedules large batch/HPC jobs (with Spot VMs), and Azure CycleCloud orchestrates Slurm-style HPC clusters — useful for hybrid bursting from an on-prem cluster. AKS runs containerized pipelines and services; Azure ML handles GPU training and MLOps; ADLS Gen2 is the lake substrate feeding OneLake. Co-locate compute with storage and prefer Spot for fault-tolerant steps to manage cost.

HIPAA / governance posture on Azure

• Entra ID (formerly Azure AD) for identity, MFA, conditional access — usually already the org's identity provider.
• Private endpoints to keep PHI traffic off the public internet.
• Customer-managed keys via Key Vault for the FHIR/DICOM services.
• Microsoft Purview for data governance, classification, and lineage across the estate.
• Azure Monitor / diagnostic logs for the HIPAA-required audit trail.

When Azure is the right call

• The customer is a Microsoft-first health system (Entra ID, Teams, Office 365) — integration and identity are already there.
• Epic on Azure deployments where co-locating the analytics and interoperability layer reduces friction.
• Connected medical devices / IoMT — the MedTech service is purpose-built for device telemetry → FHIR.
• The analytics strategy already centers on Microsoft Fabric / Power BI.

Lab

There is no dedicated Azure spoke repo yet. The hls-fhir-interop lab includes an Azure Health Data Services deployment variant for the FHIR interoperability pattern.

Check yourself

1. A health system runs Epic on Azure and uses Entra ID and Power BI. Why is Azure likely the lowest-friction platform for their new clinical analytics project?
2. Which Azure Health Data Services component handles continuous device telemetry, and what does it normalize the data into?
3. In the Fabric-based architecture, where does standardized OMOP analytics happen, and what feeds it?

Reference architectures

• Microsoft for Healthcare — reference architectures (overview).
• End-to-end FHIR-based PHI store workload.
• microsoft/health-architectures (GitHub) — end-to-end FHIR reference implementations.
• Azure/ahds-reference-architecture (GitHub) — Azure Health Data Services integration patterns.

Further reading

• Azure Health Data Services
• Healthcare data solutions in Microsoft Fabric
• Text Analytics for Health
• Azure HIPAA/HITRUST compliance
• Azure Architecture Center — Health industry (reference architectures)