In our recent webinar, "Breaking the Silos: Understanding What It Takes to Build a Unified Data Platform for Actionable Healthcare Insights," Keyrus joined forces with Brittany Timpano, Manager of Decision Support and Data Management at Royal Victoria Regional Health Centre (RVH) and part of the CARE4 Partnership, alongside Microsoft, to discuss how a four-hospital healthcare partnership is turning years of accumulated clinical data into a governed, AI-ready platform.
For healthcare organizations today, the challenge isn't a shortage of data; it's making that data trustworthy enough to act on. In case you missed the live session, here's a recap of how the CARE4 Hospital Network Partnership is approaching governance, federated architecture, and AI on Microsoft Fabric. You can watch the replay here.
Industry Perspective: From Digitization to Data Maturity
Brittany Timpano opened the session by introducing CARE4, a regional digital partnership between four Ontario hospitals: Royal Victoria Regional Health Centre, Collingwood General & Marine Hospital, Georgian Bay General Hospital, and Headwaters Health Care Centre. Together, the partnership represents roughly 744 inpatient beds and manages over 266,000 patient days and 213,000 ER visits a year, all running on a single, shared Meditech Expanse instance.
Having each achieved HIMSS EMRAM Stage 6, the partnership had proven it could digitize. The next challenge was proving it could turn that digitized data into value.
Key Challenges Addressed
Inconsistent metrics: the same KPI defined differently across teams, with reports rebuilt from scratch each time.
Poor traceability: no reliable lineage, making it difficult to trace a disputed number back to its source.
Erosion of trust: as discrepancies piled up, users began to distrust the data itself, driving a growing backlog of ad hoc requests to the decision support team.
Regional complexity: four autonomous organizations with their own boards, C-suites, and starting points for analytics, all operating under shared privacy obligations.
The Federated Governance Model
Rather than choosing between a fully centralized model (slow, unresponsive to local needs) or a fully decentralized one (duplicated effort, no comparability), CARE4 landed on a federated approach: govern regionally, operate locally, learn collectively. Regional standards cover definitions, privacy, and technology, while each site retains control over local workflows and reporting needs, reducing duplication while preserving the autonomy each hospital needs.
Driving Innovation with AI on Microsoft's Stack
Marco Baltazar of Microsoft shared where healthcare organizations sit in the broader AI adoption curve, and one example stood out: an Ontario hospital that stood up a Copilot Studio knowledge agent, nicknamed "Nurse Nancy," in just seven minutes by pointing it at 640 existing standard operating procedures stored on a shared drive. The agent gives frontline staff instant, conversational access to protocols, the kind of resource a nurse needs at 2 a.m., not a 17-page PDF.
The same hospital later used an agentic "Co-Work" workflow to assess how an upcoming EMR migration would affect all 640 SOPs, a task the CIO estimated would have taken a year of manual review, completed instead for roughly the cost of the Copilot Studio credits consumed.
The Foundation: A Federated Fabric Architecture
Alex Perras, Senior Managing Consultant at Keyrus and lead on the CARE4 implementation, walked through the technical building blocks underpinning the platform:
Governance-first design: Microsoft Purview cataloging, PHI classification, and a shared business glossary embedded across the pipeline end-to-end, not bolted on after the fact.
Medallion architecture: Meditech Expanse data ingested via mirroring (with a traditional ETL framework for sources that don't support CDC), then refined through bronze, silver, and gold layers so each site can consume data at the level of maturity it's ready for.
Single shared Fabric tenant: one regional capacity with isolated site workspaces, chosen in part to unlock the F64 SKU threshold, where Power BI licensing becomes unlimited across thousands of end users — a meaningful economic consideration at regional scale.
Capacity monitoring: an Eventhouse-based approach to tracking usage more effectively than Microsoft's native capacity metrics app, paired with Activator jobs for real-time data quality alerts.
The Keyrus Perspective: Foundations Before Features
Before writing a single pipeline, Keyrus spent four to six weeks auditing CARE4's existing SSIS packages, data marts, and reporting estate, and interviewing each partner site to understand data maturity and pain points. That investigation shaped a phased plan built around three pillars: foundational governance, a federated operating model, and the platform itself.
Rather than starting with an AI use case, the team chose a "lighthouse" use case deliberately: rebuilding CARE4's most complex existing asset, a multi-sheet, sub-five-minute-latency SSRS dashboard, in Power BI. Proving that single, high-stakes migration, while ingesting data from all 26 Meditech modules used across the reporting estate, not just what one dashboard needed, gave the team a semantic model robust enough to support future AI use cases without revisiting ingestion later.
Throughout the build, Keyrus embedded its team directly alongside RVH's decision support and data management staff, prioritizing knowledge transfer so the internal team can fully own the platform once the partnership engagement winds down.
Q&A Highlight: Governance is the Hard (but Necessary) Part
When asked what she'd do differently, Brittany Timpano didn't point to technology; she pointed to timing. Starting governance conversations before the technology was in place, she noted, would have tempered the temptation to jump straight to the "fun, shiny" parts of the build. The business case rested less on new capability and more on eliminating waste: hundreds of previously built reports sitting unused, representing real, recurring cost.
On architecture trade-offs, Alex Perras noted that a single shared Fabric capacity versus per-site capacities are both defensible choices; the right answer depends on existing tenant structure, site scale, and licensing economics, not a one-size-fits-all rule. And on team size, his advice was reassuring: standing up a platform like this takes fewer people than most organizations assume, provided the architecture questions are resolved early, and a clear-minded vision guides the rollout.
Ready to Build Your Healthcare Data Foundation?
Whether you're a healthcare network trying to turn digitized data into trusted decision-making, or any organization weighing a centralized versus federated approach to governance, Keyrus can help assess where you stand and build a roadmap to get there. Contact us to get started, or explore our healthcare industry work.