UNC Health modernizes data analytics for care, operations, and research with Fabric

UNC Health is a sprawling academic medical system with a footprint spanning 52 counties and 20 hospitals across North Carolina. It includes UNC Hospitals and its provider network, along with the clinical programs of the UNC School of Medicine. UNC Health operates at the intersection of two equally demanding missions: delivering high‑quality patient care at scale and advancing research that improves how that care is delivered. Data sits at the center of both missions, and must be reliable, compliant, and scalable.

“It's a lifeblood of an academic institution to do research,” says Rachini Moosavi, chief analytics officer for UNC Health, “And when you have an academic medical center, it's even more critical, because that's how data and analytics help us solve some of the biggest hurdles in healthcare delivery.”

For years, the infrastructure underneath that data struggled to keep pace as data storage grew by roughly 40 percent year over year. The tipping point came in December 2023, when a CPU failure took the production data warehouse offline.

When compliance and possibility converged

Healthcare data is governed by HIPAA, and early cloud platforms often placed significant responsibility on customers to implement and manage security and compliance controls.

Rather than positioning cloud migration as a defensive response to aging hardware, UNC Health began framing the shift as a strategic opportunity, as generative AI started reshaping what was possible.

For UNC Health, the question wasn’t whether to move to the cloud, but what kind of platform could support healthcare‑grade governance without recreating on‑premises complexity.

"Having an environment that's more SaaS, where a lot of that was built into the system itself, gave us confidence that we could move in that direction rather than infrastructure in the cloud that we had to maintain ourselves," says Shaun McDonald, Manager of the Enterprise Data Warehouse team.

"[Generative AI] allowed us to paint a picture,” Moosavi says. “If you want conversational analytics, your data needs to live in the cloud, so you can layer on all these great capabilities on top of it. It allowed us to have a different level of conversation."

UNC Health selected Microsoft Fabric as its unified analytics platform, standardizing its entire data estate in a single, governed environment. That decision was reinforced by its existing Microsoft ecosystem, allowing teams to build on familiar tools and integrations rather than introducing a fragmented new platform. Analytics teams also use Copilot in Microsoft Fabric to assist with pipeline development and notebook workflows, accelerating day‑to‑day engineering tasks. As a result, UNC Health now runs its enterprise analytics workloads on Microsoft Fabric within a single governed data environment.

A foundation that frees teams to focus on what matters

With Fabric carrying all production workloads, UNC Health could focus on building rather than fixing. Community analysts now access centrally managed data models to drive business insights across the system, without competing for fixed server resources.

By consolidating contract analysis and vendor evaluation across multiple datasets within the Fabric environment, the supply chain team estimates they were able to save approximately two to three weeks of preparation across the documents and analyses they've produced.

Beyond operational efficiencies, Fabric’s impact is especially visible in UNC Health Alliance's care gap closure work. Its population health mission depends on knowing what care patients have received, including care delivered outside UNC's own walls. Much of that information arrives as unstructured documents, making it difficult to surface reliably within clinical workflows.

For years, trained abstractors manually reviewed unstructured clinical documents to determine whether patients had received appropriate screenings. For diabetic eye exams, the manual process included navigating multiple systems, reading full reports, classifying results, and entering findings into the EHR.

“When you're getting into manual chart reviews, that data is often free text, like colonoscopy reports, mammograms, eye exams, which can be very difficult," explains Dr. Isha Mehta, Medical Director of Population Health Informatics.

To support this work, UNC Health developed EASI (Extract, Analyze, Synthesize, and Integrate), a framework that scans available documents, to identify whether a relevant exam is present and extracts source information into a standardized format that can be integrated into Epic. While EASI helps prioritize document review, trained abstractors remain the final checkpoint before any finding enters the patient record.

“The nice thing about this framework is that we still have humans in the loop,” says Dr. Mehta. “We never want to put something in our record system or be marked as not being fully correct.”

By February 2026, UNC Health’s Care Gap Closure team reviewed approximately 9,700 diabetic eye exam care gaps, as well as nearly 180,000 additional care gaps across other quality screening measures. The initial diabetic eye exam pilot demonstrated improvements in documentation detection accuracy within manual abstraction flows. After refinement, abstraction detection accuracy improved from 83% to 93% during the diabetic eye exam pilot, with 40 of 43 exams fully accurate. Following the pilot, UNC Health expanded use of the solution to support broader care gap closure efforts.

"With EASI, we actually cut down almost 50% of our time spent for each patient, about three minutes on average per chart," says Patrizia Dowdell, Program Manager at UNC Health Alliance. "That has been a huge step for my team. We are able to focus on other care gaps as well."

Beyond efficiency gains, the clinical significance becomes clearer. When EASI surfaces an abnormal finding that would otherwise have sat unread in a media file, it can help inform clinical prioritization.

“If we are surfacing more abnormal results, that means our patients are more likely to receive timely follow-up, because otherwise it could just be lost,” Dr. Mehta says. “Just knowing that a patient has an abnormal retina finding that may warrant follow-up can influence clinical prioritization and attention in their care."

EASI is now expanding beyond diabetic eye exams to additional screening use cases, including colorectal, cervical, and breast cancer, extending its impact across population health initiatives. Fabric has also made previously difficult‑to‑access data reusable across the organization, including ECG and signal data that can now be governed once and applied across both operational and research use cases.

Innovation with guardrails

Research is a core pillar of UNC Health's academic mission, and it requires an environment distinct from clinical operations, one that is designed to safeguard sensitive patient data while providing researchers with controlled access to data and tools that support exploratory and innovative work.

Until recently, sensitive EHR data was stored on password-protected network drives with limited programmatic controls to prevent unauthorized movement.

"The vast majority of researchers always want to do the right thing, but it's really easy to drag a file onto your desktop or save something in your temp folder without intending to,” says Dr. Emily Pfaff, Associate Professor of Medicine and Director of the Informatics and Data Science Core at NC TraCS, UNC's NIH-funded Clinical and Translational Science Institute. “If your laptop is compromised, all of a sudden we have a huge issue.”

The answer was SHIRE: the Secure Health Informatics Research Environment, built on Microsoft Azure infrastructure within a trusted research enclave framework. Researchers log in to a familiar Windows desktop, find their data already provisioned by a team of “honest brokers,” and access advanced analytics and AI tools within an environment designed so that data remains within a defined secure perimeter.

Since launching in November 2025, SHIRE supports 25 active studies, with large language models being explored within the environment. The team aims to scale the platform to support additional studies through 2026, expanding access to security analytics and AI capabilities for researchers.

“Before we started building SHIRE, we were starting to be in a position where we had to weigh certain research requests, which is a real bummer for the investigator,” Dr. Pfaff says. “But it's also bad for the institution because the institution wants to be able to say we're doing the most novel informatics and data science research around, and we literally were not able to support that.”

Fabric serves as the governed data foundation that enables SHIRE when research requires access to clinical or operational data.

A shared foundation for real-time insight across healthcare

The near-term vision at UNC Health points to conversational analytics. The goal: enable leaders, clinicians, and analysts to query data directly, in real time, without routing a request through a ticketing queue, while continuing to expand secure, scalable research innovation through SHIRE.

“It’s going to be a game changer,” says Moosavi. “It's no longer a ticketing queue with a 30-to-45-day response time. Leaders and analysts can access insights in real time.”

Dr. Pfaff agrees that this accessibility is vital. “Unleashing really smart people on clinical data can help us make critical discoveries that make demonstrable impacts on patients' lives,” she says. “Achieving a balance between not so locked down that you can't do any work, but not so open that patients' data are at risk: [that’s] what will enable us to keep making progress on data-driven clinical research.”

Fabric provides a governed foundation that allows clinical insight and research integrity to coexist, creating a scalable blueprint that extends beyond the walls of a single institution.

“You often hear things like, ‘A healthcare organization is competing on analytics,’" Moosavi concludes. “But a lot of our organizations are finally leaning in and saying, ‘We're going to help each other. Let's stop re-creating the wheel and start doing it together.’ And if we can solve this for healthcare, it's going to take all of us leaps and bounds beyond where we could have gone alone."

Discover more about UNC Health on Facebook, Instagram, LinkedIn, and X/Twitter.