Overview
Virtua Health, a five-hospital system based in Marlton, New Jersey, has eliminated 15,000 excess hospital days. It accomplished this by building a suite of AI tools in partnership with Microsoft. The results, detailed in an April 7 Microsoft news release, highlight real, measurable outcomes across three clinical areas: patient flow, heart failure management, and sepsis detection.
For health systems still evaluating the business case for AI, Virtua’s results offer a compelling example. The organization did not simply adopt off-the-shelf tools. Instead, it built purpose-specific solutions using Microsoft’s Azure AI Foundry, Power BI, and Copilot platforms — then integrated them directly into existing clinical workflows.
Power BI Dashboards Speed Up Clinician Reviews
From 30 Minutes to 5 Minutes
One of the most immediate wins came from a shift in how clinicians access patient information. Before the AI rollout, clinicians spent roughly 30 minutes reviewing patient data before making care decisions. That process was manual, fragmented, and time-consuming.
Virtua addressed this by creating Power BI dashboards paired with Microsoft Copilot summaries. Together, these tools generate concise patient overviews that clinicians can read and act on quickly. Review time dropped from about 30 minutes to just five minutes — a reduction of more than 80%.
Why Speed Matters in Clinical Settings
Faster reviews do more than save time. They also reduce delays in discharge planning, allow care teams to see more patients per shift, and help identify patients who are ready to transition out of acute care sooner. Consequently, freeing up bed capacity becomes a natural byproduct. This direct link between faster information access and reduced excess hospital days is one of the clearest demonstrations of AI’s operational value in healthcare.
AI Model Transforms Heart Failure Care
Turning Unstructured Notes Into Actionable Predictions
Heart failure documentation presents a persistent challenge. Clinicians record observations in free-text notes, but extracting meaningful patterns from that unstructured data at scale requires significant effort. Virtua’s team tackled this problem by building a model in Azure AI Foundry.
The model converts unstructured patient notes into structured predictions. As a result, it improved heart failure patient identification by 84 percentage points — a dramatic leap in early detection capability.
Better Follow-Up, Shorter Stays
The downstream effects were equally significant. Heart failure follow-up rates climbed from 58% to 81% after the model went live. Moreover, the average hospital stay in specialized heart-failure units fell by a full day. Better identification leads to earlier intervention, and earlier intervention leads to shorter, more effective care episodes. This is a textbook example of how predictive AI can reshape clinical pathways.
Sepsis Detection Gets Sharper and Faster
The Problem With False Positives
Sepsis is one of the most time-sensitive conditions in hospital medicine. However, many commercial sepsis detection tools generate high rates of false positives — alerts that turn out to be incorrect. These false alarms erode clinician trust and create alarm fatigue, making it harder to respond effectively when a real sepsis case appears.
Virtua’s Custom XGBoost Approach
Virtua tackled this by developing a custom XGBoost machine learning model in Azure. The model trains on electronic health record data and integrates directly into Epic, the widely used EHR platform. Compared to commercial tools, Virtua’s model reduces false positives for sepsis by more than 80%. That level of precision means clinical teams can trust their alerts and act with greater confidence — without being overwhelmed by noise.
What This Means for Hospital AI Adoption
A Model for Health System Innovation
Virtua’s AI journey illustrates a broader trend in health system innovation. Rather than waiting for vendors to deliver ready-made solutions, health systems are increasingly building custom tools on foundational platforms like Azure AI Foundry. This approach allows for tighter integration with existing systems, better alignment with specific patient populations, and more meaningful performance outcomes.
Measuring ROI Beyond Cost Savings
The 15,000 excess hospital days eliminated represent real capacity freed up for other patients. Furthermore, the improvements in follow-up rates and diagnostic accuracy point to better long-term outcomes — benefits that extend well beyond simple cost reduction. Health system leaders evaluating AI investments should take note: the case for AI is no longer speculative. Virtua’s results show it clearly.
Key Takeaways
- Power BI and Copilot cut clinician patient review time from 30 minutes to 5 minutes.
- An Azure AI Foundry model improved heart failure identification by 84 percentage points and boosted follow-up rates from 58% to 81%.
- A custom XGBoost model in Azure, integrated with Epic, slashed sepsis false positives by more than 80% compared to commercial tools.
- 15,000 excess hospital days were eliminated across the five-hospital system.
- Virtua’s results demonstrate that custom-built AI tools, grounded in real clinical workflows, deliver measurable and sustainable impact.
