Microsoft Cloud for Healthcare — FHIR + Dragon Copilot (2026)

The FHIR Service in Azure Health Data Services is the HL7 FHIR R4-compliant ingestion and exchange layer for Microsoft Cloud for Healthcare. It accepts FHIR resources from Epic, Oracle Health Cerner, athenahealth, MEDITECH, NextGen, eClinicalWorks, and the long tail of digital health source systems, persists them in a HIPAA-eligible workspace covered by the Microsoft Business Associate Agreement, and exposes them to downstream Microsoft surfaces — Microsoft Fabric, Power BI, Dynamics 365 Patient Insights, the Care Team Copilot, and Azure Machine Learning — through governed FHIR APIs. Every read, write, and search is audit-trailed in Microsoft Sentinel under HIPAA-aligned retention.

The DICOM Service in Azure Health Data Services is the cloud-native medical imaging store that lands radiology, cardiology, pathology, ophthalmology, and dental imaging from a wide range of modalities and PACS systems — GE Healthcare, Philips, Siemens Healthineers, Canon Medical, Carestream, Sectra — into a HIPAA-eligible workspace. It exposes DICOMweb (WADO-RS, STOW-RS, QIDO-RS) for native image retrieval, links each study back to the FHIR patient resource through ImagingStudy, and surfaces the imaging plane to downstream Azure ML for radiomics, pathomics, and AI-assisted reading. The DICOM Service inherits the same BAA, the same Sentinel audit retention, and the same Purview labeling discipline as the FHIR Service.

Azure Health Data Services bundles the MedTech Service for IoMT device-event ingestion (continuous glucose monitors, cardiac telemetry, infusion pumps, smart inhalers, remote patient monitoring devices), a de-identification service that strips HIPAA Safe Harbor identifiers for analytics extracts, and the Microsoft Fabric Healthcare data foundation that lands FHIR plus DICOM plus MedTech feeds into OneLake under a clinical medallion architecture. The MedTech Service normalizes device payloads against the FHIR Observation resource. The de-identification service produces both expert-determination and Safe Harbor extracts for analytics and Copilot grounding. The Fabric Healthcare lakehouse is the single analytics plane for the provider.

Patient Insights is the Microsoft Cloud for Healthcare Dynamics 365 surface for patient relationship management — the longitudinal patient and household record on Dataverse, named pipelines for new-patient acquisition, service-line referral, care-gap closure, post-discharge follow-up, and patient-financial-services collections, plus a Teams-embedded virtual-clinic and care-coordination surface. Every care-team action — outreach, scheduling, documentation, escalation — is consent-gated against the patient Notice of Privacy Practices, role-bound through Microsoft Entra with attribute-based access control reflecting clinical role and care relationship, and audit-trailed into Microsoft Sentinel with HIPAA-aligned retention by default.

The Care Team Copilot is the Microsoft Cloud for Healthcare-bundled Copilot surface that sits inside Microsoft Teams for the clinical team — surfacing the patient longitudinal record at the point of huddle, drafting referral letters and care plans, summarizing prior-authorization packets, and helping the care manager triage the value-based-care risk panel. The Care Team Copilot grounds on the de-identified Fabric Healthcare gold layer for population queries, on the consented patient slice for one-patient queries, and on the named SharePoint policy library for organizational knowledge — with Purview sensitivity labels, named escalation triggers, and Sentinel transcript retention enforced at every turn.

Dragon Copilot is the Microsoft ambient clinical documentation product built on the Nuance Dragon Medical and DAX Copilot foundation that Microsoft acquired through Nuance in 2022. It listens to the patient encounter (in-room, telehealth, or hybrid), produces a structured, specialty-aware clinical note (SOAP, APSO, H&P, progress, procedure), and writes the note back to the electronic health record through the EHR-native integration. The clinician edits, signs, and the encounter is closed without after-hours note-typing. Every recording, transcript, and generated note is BAA-covered, encrypted in transit and at rest, audit-trailed in Sentinel, and bound to the HIPAA Privacy Rule minimum-necessary standard at the EHR write-back.

Beyond ambient note generation, Dragon Copilot delivers voice-driven clinical Q&A grounded on the patient longitudinal record and the organization knowledge base, voice-driven problem-list and medication-list reconciliation, voice-driven after-visit-summary generation in patient-readable language at the appropriate health-literacy reading level, and voice-driven order entry with EHR-native interlock. The clinician speaks naturally; Dragon Copilot writes the structured artifact and writes it back to the EHR. The post-visit administrative burden moves from hours per day to minutes per encounter.

Dragon Copilot ships first-class integration with the four EHR platforms that cover the overwhelming majority of U.S. acute-care, ambulatory, and post-acute deployments — Epic (Hyperdrive, Haiku, Canto), Oracle Health Cerner Millennium (PowerChart, Power Mobile), athenahealth athenaOne, and MEDITECH Expanse — plus production-quality integration with NextGen Enterprise, eClinicalWorks, Allscripts Sunrise and Paragon, Greenway Health, Veradigm, and Practice Fusion. Each integration writes structured discrete data where the EHR supports it, embeds inside the native chart workspace rather than a separate tab, and inherits the EHR's own role-based access control plus the BAA-covered Microsoft tenant boundary.

Provider-organization consolidation through M&A is the single most common Microsoft Cloud for Healthcare engagement pattern at integrated delivery networks and private-equity-backed multi-site groups. The acquired entity arrives with a different EHR (Epic, Cerner, athenahealth, MEDITECH, NextGen, or a long-tail platform), a different identity directory, and a different consent posture. The EPC pattern lands every EHR source through the Azure Health Data Services FHIR Service, conforms the data to US Core plus USCDI v3 at the Fabric Healthcare silver layer, unifies the patient identity through Customer Insights Data with documented match-rate evidence, and presents the longitudinal patient record to Patient Insights, the Care Team Copilot, and Dragon Copilot — without forcing a hard EHR rip-and-replace.

Ambulatory clinic networks — primary-care groups, multi-specialty groups, urgent-care networks, retail-clinic networks, and direct-primary-care practices — face the highest clinician-burnout exposure from after-hours documentation burden. The EPC pattern deploys Dragon Copilot ambient clinical documentation across the clinician workforce, with specialty-aware note templates, EHR-native write-back to Epic, athenahealth, eClinicalWorks, NextGen, or the host EHR, named documentation-quality gates, named clinician-attestation interlocks, and Sentinel audit retention. The pattern compresses documentation time from 90 to 120 minutes per day to under 15 minutes per day at scale, with named throughput, quality, and burnout-survey-score evidence.

Integrated delivery networks operating across acute, ambulatory, post-acute, and home-health lines need a single population-health analytics plane that respects HIPAA minimum-necessary, 42 CFR Part 2 substance-use confidentiality, and the state-level overlays. The EPC pattern lands every EHR, every payer-feed, every social-determinants-of-health source, and every patient-generated-health-data source through Azure Health Data Services, conforms the data at the Fabric Healthcare silver layer, and assembles the population-segment, care-gap, risk-stratification, and utilization-management views at gold. Power BI delivers the population-health, service-line, and executive dashboards. Care managers operate from the Care Team Copilot and Patient Insights surfaces.

Health-plan payers — commercial carriers, Medicare Advantage organizations, Medicaid managed-care organizations, and self-funded employer-plan administrators — operate inside the same HIPAA boundary as the provider, with an additional overlay from the NAIC Insurance Data Security Model Law, state-DOI requirements, and the No Surprises Act. The EPC pattern lands claims, eligibility, prior-authorization, and appeals feeds into the Fabric Healthcare lakehouse; uses Azure ML for fraud, waste, and abuse scoring under documented SR-11-7-style model governance; uses Copilot Studio for member-service agent assist; and uses the Care Team Copilot for case-management triage on high-cost claimants and rising-risk members.

Life-sciences sponsors — pharmaceutical, biotech, medical-device, and contract-research-organization — operate on Microsoft Cloud for Healthcare alongside the FDA 21 CFR Part 11 electronic-records-and-signatures requirement, the ICH-GCP good-clinical-practice framework, and the HIPAA boundary that applies whenever the sponsor handles identifiable subject data. The EPC pattern lands EDC, IRT, lab, imaging, and ePRO feeds through Azure Health Data Services; uses the Fabric Healthcare lakehouse for the conformed trial data layer; uses Azure ML for digital-biomarker analysis and adaptive-trial signal detection under SR-11-7-style governance; uses the Care Team Copilot for site-monitor and clinical-research-coordinator support; and ships a Part-11-aligned audit binder at study close.

Value-based care arrangements — Medicare Shared Savings Program ACOs, Medicare Advantage capitated risk, commercial value-based contracts, and Medicaid managed-care risk arrangements — depend on accurate, defensible, and auditable risk stratification of the attributed panel. The EPC pattern lands every claims, EHR, social-determinants, and patient-generated feed through Azure Health Data Services, conforms the data at Fabric Healthcare silver, runs the risk-stratification model on Azure ML under SR-11-7-style governance, presents the rising-risk and high-cost-claimant panel to the care manager through the Care Team Copilot and Patient Insights, and audit-trails every prediction, intervention, and outcome through Sentinel for the next CMS review or actuarial validation.

• HIPAA Privacy Rule minimum-necessary standard enforced at the FHIR endpoint via SMART on FHIR scopes plus Entra attribute-based access control reflecting clinical role and care relationship — every read, write, and search consent-bound
• HIPAA Security Rule administrative, physical, and technical safeguards mapped to Microsoft control owners — Sentinel audit retention, Defender XDR endpoint and identity coverage, Purview sensitivity-label propagation, Entra MFA and conditional access for every privileged identity
• HIPAA Breach Notification Rule readiness — Sentinel disclosure-accounting query patterns, named breach-investigation runbook, and named Office for Civil Rights response cadence pre-built into the engagement
• HIPAA-aligned six-year audit retention by default on every FHIR, DICOM, MedTech, Dynamics 365, Copilot, and Sentinel surface inside the Microsoft Business Associate Agreement boundary

• 42 CFR Part 2 substance-use chart slice carried under a Purview sensitivity label distinct from the general HIPAA PHI label — Substance-Use-Restricted — with separate Entra access policy and separate consent-context query filter
• Care Team Copilot and Dragon Copilot grounding catalog explicitly excludes the Substance-Use-Restricted slice unless the patient has signed the Part-2-specific consent — with named exclusion rationale per grounding source
• Sentinel audit log of every substance-use-restricted read, every consent verification, and every disclosure under the Part-2-specific consent boundary — indexed for the Substance Abuse and Mental Health Services Administration audit pattern
• 42 CFR Part 2 consent workflow built into Patient Insights with named clinical-role permission set, named consent-form library, and named re-consent cadence per care-relationship change

• California Confidentiality of Medical Information Act overlay applied through a CMIA-Restricted Purview label with named California-specific consent workflow and named California breach-notification cadence
• New York SHIELD Act safeguards mapped to Microsoft control owners with named breach-notification cadence to the New York Attorney General and the Department of State
• Texas Medical Records Privacy Act overlay applied through a TMRPA-Restricted Purview label with named Texas-specific consent workflow and named Texas breach-notification cadence
• Multi-state operator playbook — Massachusetts 201 CMR 17, Illinois Personal Information Protection Act, Colorado Privacy Act, Virginia Consumer Data Protection Act — mapped to the Microsoft control plane at the state-of-residence level for the patient population

• Microsoft Online Services Business Associate Agreement signed at the tenant level naming every Microsoft service in scope — Azure Health Data Services, Microsoft Fabric, Dynamics 365, Microsoft 365, Copilot, Sentinel, Defender, Purview, Power Platform — with sub-processor disclosure
• EPC Group operates under a downstream Master Services Agreement plus Business Associate Agreement equivalent that mirrors the Microsoft Online Services BAA scope with named-control-owner accountability per Microsoft service in the data flow
• U.S.-region data residency by default — East US, East US 2, Central US, South Central US, West US 2 — with documented Microsoft sub-processor list and named OCR-disclosure-cadence per region
• Documented Microsoft Online Services Healthcare Amendment scope addendum signed at engagement kick-off covering Azure Health Data Services, Dragon Copilot, Care Team Copilot, and the Patient Insights surface