Pragmatic randomised trial of an AI-ECG for low ejection fraction in primary care (EAGLE)
Giving primary care teams the output of an AI that reads routine 12-lead ECGs for signs of a weak heart raised new diagnoses of low ejection fraction within 90 days from 1.6% to 2.1% of 22,641 patients, in a cluster-randomised trial across 120 teams at 45 Mayo Clinic clinics and hospitals, without increasing echocardiogram use overall.
- Lab
- Mayo Clinic
- Model
- AI-ECG convolutional neural network for left ventricular systolic dysfunction
- Field
- Medicine
- Date
- 2021-05-06
- Human collaborators
- Xiaoxi Yao, Paul A. Friedman, Peter A. Noseworthy
What was found
120 primary care teams (358 clinicians) were randomised. ECGs taken in routine care for adults with no prior heart failure were run through the algorithm, and clinicians in the intervention arm had access to its results; 6% of patients were flagged as likely to have low ejection fraction. The primary outcome, a new diagnosis of ejection fraction of 50% or less within 90 days, rose from 1.6% to 2.1% overall (odds ratio 1.32) and from 14.5% to 19.5% among AI-positive patients. Echocardiogram use was similar overall (18.2% against 19.2%) but higher for AI-positive patients (38.1% against 49.6%). The model is the one Mayo published retrospectively in 2019. Anumana, a company formed by Mayo Clinic and nference, holds an FDA clearance for a low ejection fraction AI-ECG algorithm developed from this work.
Novelty check
Europe PMC, titles (artificial intelligence OR deep learning OR neural network) AND (electrocardiogram OR ECG) AND (ejection fraction OR ventricular dysfunction), before 6 May 2021: seven hits, all retrospective development or external validation studies, plus this trial's own design paper (Data Brief, 2019); the model itself was published by Attia et al. (Nature Medicine, 2019). No earlier randomised trial of an AI-ECG.
Caveats and known objections
One health system that both developed the model and ran the trial. The outcome is a diagnosis, a process measure; the trial did not measure heart failure events or survival. Clinicians acted on only about half of positive results, so the effect depends on uptake as much as on the model. Autonomy is graded ai-assisted, the weaker reading: the AI detects a pattern clinicians cannot see on an ECG, but the measured outcome needed a clinician to act on the alert and an echocardiogram to confirm it. The FDA-cleared Anumana algorithm is a commercial descendant of the Mayo model, not necessarily the version tested here.
Trial registration
That a study was registered, with the outcomes it declared and the date it declared them, so a published endpoint can be checked against the one planned. It does not certify the results, their analysis, or any clinical benefit; registry records are submitted by the sponsor and are not peer reviewed.
Regulatory authorization
FDA AI-enabled devicesLow Ejection Fraction AI-ECG Algorithm (Anumana)
K232699
That the FDA authorized the device for marketing in the US for a stated indication, on the evidence the manufacturer submitted. Most AI devices reach the market by 510(k), which requires substantial equivalence to an earlier device rather than new clinical evidence, so an authorization is not evidence of clinical benefit.
Anumana's commercial algorithm, cleared in September 2023 and developed from the Mayo Clinic AI-ECG; the FDA record does not identify it with the trial version.
Nobody outside the lab has checked this yet.
Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.
Or on GitHub: submit a check challenge the grade send a correction or send a pull request
Flag this for triage
Signals order the review queue and nothing else. They are never published, and they never move a grade: that takes a citation.
Entry history (1 event)
- AddedEntered the registry graded Peer reviewed and AI-assisted.
Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.
Graded peer reviewed for verification and ai-assisted for autonomy. What these mean.
Cite this entry
whataifound.org. (2021). Pragmatic randomised trial of an AI-ECG for low ejection fraction in primary care (EAGLE). whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2021-05-06-eagle-ai-ecg-low-ejection-fractionBibTeX
@misc{whataifound-mayoclinic-2021-fraction,
title = {Pragmatic randomised trial of an AI-ECG for low ejection fraction in primary care (EAGLE)},
author = {{whataifound.org}},
year = {2021},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by Mayo Clinic. Verification: Peer reviewed. Autonomy: AI-assisted.},
url = {https://whataifound.org/finding/2021-05-06-eagle-ai-ecg-low-ejection-fraction}
}