--- title: "Epic won FDA clearance for its deterioration score with rival AgileMD's algorithm as the benchmark" description: "The FDA found Epic's rebuilt patient-deterioration score substantially equivalent to AgileMD's eCARTv5, making the challenger's algorithm the benchmark the EHR giant had to match. With a mortality study behind it, Epic's bundled score resets the math for hospital AI vendors." publisher: "The Inference" section: "Health" published: 2026-11-01T03:27:51.182Z modified: 2026-11-01T03:27:51.182Z canonical: https://theinference.org/article/epic-won-fda-clearance-for-its-deterioration-score-with-rival-agilemd-s-algorithm-as-the-benchmark language: en keywords: "Health, Artificial Intelligence, FDA, Hospitals, Medical AI" --- # Epic won FDA clearance for its deterioration score with rival AgileMD's algorithm as the benchmark > The FDA found Epic's rebuilt patient-deterioration score substantially equivalent to AgileMD's eCARTv5, making the challenger's algorithm the benchmark the EHR giant had to match. With a mortality study behind it, Epic's bundled score resets the math for hospital AI vendors. In September the FDA cleared Deterioration Index Version 2, Epic's rebuilt early-warning score that flags hospitalized adults likely to need escalated care within 24 hours or to die within 72, through the 510(k) medical-device pathway [1]. To win that clearance, Epic had to prove the software substantially equivalent to eCARTv5, a competing score from [AgileMD](https://theinference.org/markets/companies/agilemd), a San Francisco company founded at the University of Chicago by physician Dana Edelson and CEO Borna Safabakhsh [1]. The regulator made the challenger's algorithm the yardstick the industry giant had to match [1]. ## The challenger drew first blood, so Epic rebuilt The history explains the clearance. [A 2024 study across seven Yale New Haven hospitals](https://www.beckershospitalreview.com/healthcare-information-technology/ehrs/epics-deterioration-index-wins-fda-clearance-after-years-of-use/) ranked Epic's earlier index near the bottom of six early-warning scores, and noted the tool had never been formally vetted by the FDA [1]. [AgileMD](https://www.agilemd.com/)'s eCART, cleared by the FDA in June 2024, was the winner of that comparison [2]. Yale New Haven adopted eCART, and by August 2025 [reported a 13% relative reduction in overall mortality](https://www.agilemd.com/news), 1.2% in absolute terms [3]. Epic submitted its V2 application on January 2, 2026, tested it on about 10.7 million score observations from three U.S. sites, and posted an area under the curve of 0.83 for the combined outcome and 0.94 for mortality alone [1]. ## The receipts: 18% fewer deaths, no ICU surge The outcome evidence came from RWJBarnabas Health and Rutgers Robert Wood Johnson Medical School, [published in NEJM AI in July 2026](https://hitconsultant.net/2026/07/29/rwjbarnabas-rutgers-nejm-ai-epic-deterioration-index-mortality-reduc/) [4]. Across 23,132 patients in 11 hospitals, in-hospital mortality among high-risk patients fell from 23.1% to 18.6%, an 18% reduction in risk-adjusted odds of death [4]. Rapid-response evaluations rose from 25.3% to 37.5% without a surge in ICU transfers, because red-alert scores pushed notifications straight to rapid-response teams' phones instead of sitting on a desktop dashboard [4]. *Chart: **At RWJBarnabas, deaths fell and rapid-response checks rose after automated alerts** Range chart comparing RWJBarnabas Health high-risk in-hospital mortality and rapid-response evaluations before and after the automated Epic deterioration alert rollout, mortality falling from 23.1% to 18.6% and evaluations rising from 25.3% to 37.5%.* *Shares among 23,132 patients across 11 RWJBarnabas Health hospitals, before and after red-alert scores were routed from Epic's Deterioration Index straight to rapid-response teams' phones: in-hospital mortality among high-risk patients and rapid-response evaluations per patient. Source: NEJM AI study, via HIT Consultant. [4]* The wrinkle: at the medium-risk threshold, V2's sensitivity was 26% for Black patients and 32% for Hispanic patients against 51% for white patients, though a site-by-site analysis narrowed the Black-white gap [1]. Every hospital still sets its own alert thresholds [1]. *Chart: **At medium risk, the rebuilt score flagged half as many Black patients as white** Column chart of the sensitivity of Epic's Deterioration Index V2 at the medium-risk alert threshold: 26% for Black patients, 32% for Hispanic patients and 51% for white patients.* *Sensitivity of Deterioration Index V2 at the medium-risk alert threshold, by race, from Epic's 510(k) submission covering about 10.7 million score observations at three U.S. sites; a site-by-site analysis narrowed the Black-white gap. Source: Becker's Hospital Review. [1]* ## Distribution versus accuracy The money math now runs through procurement. Epic's score ships inside the EHR hospitals already license, and [a Redesign Health survey of 112 executives at Epic-based health systems](https://medcitynews.com/2026/07/epic-health-system-startup/) found 71% describe their organizations as Epic-first, not one prefers external vendors by default, and 49% demand significantly higher ROI from outside vendors before they buy [5]. AgileMD's remaining edge is accuracy plus reach: [400-plus hospitals, 4.4 million-plus patient encounters and 4,800-plus clinical pathways](https://www.agilemd.com/) the score feeds [2]. Neither company is listed, so the pressure lands on standalone early-warning vendors' pricing and on hospital AI budgets rather than on any ticker [2][5]. ## The renewal question decides which advantage wins Yale New Haven is the tell: the system that panned Epic's first score and adopted the rival's. Whether it renews with eCART or drifts back to the bundled tool decides which advantage, accuracy or distribution, actually wins [1][3]. Watch too that AgileMD's newsroom has posted nothing since February 2, 2026, while Epic's clearance is a month old [3]. A clearance after years of unregulated use is not the finish line; it is the starting gun. ## Takeaway The FDA cleared Epic's rebuilt deterioration score only against rival AgileMD's eCARTv5 benchmark; with an 18% mortality study behind it, the bundled score resets the math for standalone hospital AI vendors. ## Sources 1. [Becker's Hospital Review, Epic's Deterioration Index wins FDA clearance after years of use, 8 October 2026](https://www.beckershospitalreview.com/healthcare-information-technology/ehrs/epics-deterioration-index-wins-fda-clearance-after-years-of-use/) 2. [AgileMD, company homepage, accessed 1 November 2026](https://www.agilemd.com/) 3. [AgileMD, News page, accessed 1 November 2026](https://www.agilemd.com/news) 4. [HIT Consultant, RWJBarnabas Links Real-Time Epic EDI Alerts Directly to Rapid Response Teams, 29 July 2026](https://hitconsultant.net/2026/07/29/rwjbarnabas-rutgers-nejm-ai-epic-deterioration-index-mortality-reduc/) 5. [MedCity News, Epic-First Doesn't Mean Epic-Only: Where Hospitals Are Still Buying from Startups, 29 July 2026](https://medcitynews.com/2026/07/epic-health-system-startup/) --- Epic won FDA clearance for its deterioration score with rival AgileMD's algorithm as the benchmark — The Inference. Canonical: https://theinference.org/article/epic-won-fda-clearance-for-its-deterioration-score-with-rival-agilemd-s-algorithm-as-the-benchmark