Source: Healthcare IT News — 2026-08-03
Summary
The American Heart Association's AI Assessment Lab published its inaugural Impact Report, an independent evaluation of Ultromics' EchoGo Heart Failure — an FDA-cleared algorithm that reads standard echocardiogram videos to flag heart failure with preserved ejection fraction (HFpEF), a form of heart failure whose symptoms are easily mistaken for other conditions and routinely diagnosed late. Using an independent dataset curated by Dandelion Health rather than Ultromics' own data, the AHA found the tool could identify HFpEF an average of 263 days earlier than standard clinical care, translating to modeled reductions in hospital admissions, readmissions, and ED visits, plus real cost savings for health systems.
Key Takeaways
- EchoGo Heart Failure identified HFpEF an average of 263 days (about 8.6 months) earlier than standard clinical care, among patients who would otherwise have faced a delayed diagnosis.
- Modeled outcomes over five years: 406 fewer hospital admissions, 501 fewer readmissions, 564 fewer emergency department visits, and 477 lives saved per 10,000 patients.
- Projected cost savings: up to $1.9 million per health system over five years, roughly $1,800 per patient from both the health-system and payer perspective.
- The evaluation used an independent dataset (Dandelion Health), not Ultromics' own training/validation data — a meaningfully more credible design than a vendor self-reporting its own tool's performance.
- This is the AHA AI Assessment Lab's first Impact Report, positioning the AHA as a third-party validator for cardiovascular AI claims — a template that matters more broadly given how much health-AI marketing runs on vendor-reported numbers alone.
Reel Script
Hook (~17s, 38 words): A heart condition that's frequently missed until it's already causing hospitalizations can now be caught eight and a half months earlier — using an AI reading of a scan most patients already get. And this time, it's independently verified, not vendor-reported.
Core Concept (~65s, 150 words): HFpEF — heart failure with preserved ejection fraction — is a form of heart failure that's notoriously easy to miss because its symptoms, like fatigue and shortness of breath, look like a dozen other conditions. Doctors already order echocardiograms — ultrasound videos of the heart — for lots of reasons, but reading subtle HFpEF signals out of that video by eye is hard. EchoGo Heart Failure is an FDA-cleared algorithm that re-analyzes those same routine scans and flags HFpEF patterns a human read might miss. What makes this report different from typical AI-health marketing is who ran the evaluation: the American Heart Association's AI Assessment Lab tested the tool against an independent dataset it didn't build itself — Dandelion Health's data, not Ultromics' own — which is the difference between a vendor grading its own homework and a third party checking it.
Hands-On (~50s, 115 words): The actual numbers, worth putting on screen as a simple before/after: average time-to-diagnosis improved by 263 days. Modeled over five years and 10,000 patients: 406 fewer hospital admissions, 501 fewer readmissions, 564 fewer ER visits, and 477 lives saved. On the financial side: up to $1.9 million saved per health system over five years, about $1,800 per patient. That's not a benchmark score or a lab demo — it's a projected clinical and financial outcome from an independently audited algorithm already cleared by the FDA and already reading real patient scans.
Takeaway (~24s, 53 words): This is what "AI in medicine done right" looks like — an independent verifier, a real clinical endpoint, and dollar figures attached, not just an accuracy percentage. If you're evaluating any health-AI claim, ask exactly this question: who validated it, and against whose data?