M Health Fairview

During the COVID-19 pandemic, emergency departments faced slow diagnostic turnaround times due to supply chain issues with nasopharyngeal swabs and PCR tests. Clinicians needed a faster way to identify likely COVID-19 cases to initiate treatment sooner and prevent unintentional exposure to staff and other patients.

The University of Minnesota developed a computer vision algorithm trained on 118,000 chest X-rays (100,000 non-COVID, 18,000 COVID) to automatically evaluate X-rays for COVID-19 patterns within seconds of image capture. The algorithm was integrated directly into Epic's EHR via the Cognitive Computing platform, surfacing a risk score to care teams in real time.

All 12 M Health Fairview hospitals adopted the algorithm upon launch. The algorithm provides a COVID-19 likelihood score within seconds, enabling earlier treatment decisions and infection control measures. The solution was made available at no cost to over 450 health systems worldwide through Epic's App Orchard.

• Integrating AI directly into existing EHR workflows (Epic) dramatically lowers the barrier to clinical adoption.
• Training on large, institution-specific datasets (118,000 X-rays from M Health Fairview) produced a validated, deployable model.
• Making the algorithm free and distributable through Epic's App Orchard amplified impact beyond a single health system.