Multisite, External Validation of an AI-Enabled ECG Algorithm for Detection of Low Ejection Fraction

Q: What is the clinical evidence from this study?

Study design: Observational. Population: Low left ventricular ejection fraction (n=13960). Intervention: Artificial intelligence-based electrocardiogram (ECG-AI) algorithm vs. Transthoracic echocardiogram (reference standard). Primary outcome: Detection of low ejection fraction (LVEF ≤40%) (AUROC 0.92, 95% CI 0.91-0.93).

Q: What does this research mean for the field?

An AI-enabled ECG algorithm accurately detects low left ventricular ejection fraction across diverse clinical populations with an AUROC of 0.92, 84.5% sensitivity, and 83.6% specificity. Novelty: ClaimNovelty.CONFIRMATORY. Consensus alignment: ConsensusAlignment.NEUTRAL.

R Carter; Patrick W. Johnson; Jordan B. Strom; Jonathan W. Waks; Andrew Krumerman; Kevin J. Ferrick; Roger DeRaad; B A Steinberg; Mikolaj A. Wieczorek; Jessica Cruz; Zachi I. Attia; Francisco Lopez-Jimenez; P. Kerim Friedman; Samir Awasthi; Mohan Krishna Ranganathan; Rakesh Barve; Heather M. Alger; Konstantinos C. Siontis; Peter A. Noseworthy

doi:10.1016/j.jacadv.2025.102537

Key result

AI-enabled ECG identifies LVEF ≤40% with an AUROC of ~0.92.

AUROC 0.92
95% CI 0.91-0.93
n=13,960

Figures

Graphical abstract published by JACC Advances — Graphical abstract · Published by JACC Advances · CC BY

Why the study?

Low left ventricular ejection fraction can progress undiagnosed, and ECG-AI screening may provide a scalable detection method.

Does an AI-enabled ECG algorithm accurately detect low left ventricular ejection fraction (≤40%) in a diverse clinical population?

Key Points

To externally validate the diagnostic performance and generalizability of an artificial intelligence-enabled electrocardiogram algorithm for detecting low left ventricular ejection fraction across diverse clinical settings.
Extracted electronic health records from 4 geographically diverse institutions in the United States for patients who received both an ECG and a transthoracic echocardiogram within 30 days.
Evaluated an analysis set of 13,960 conformant patient records (mean age 66 years; 52% male) to test the AI device's ability to detect low ejection fraction (LVEF ≤40%).
The algorithm demonstrated an AUROC of 0.92 (95% CI: 0.91-0.93), with an overall sensitivity of 84.5% (95% CI: 82.2%-86.6%) and specificity of 83.6% (95% CI: 82.9%-84.2%).
At a low ejection fraction prevalence of 7.9% (1,096/13,960), the positive predictive value was 30.5% (95% CI: 28.8%-32.1%) and the negative predictive value was 98.4% (95% CI: 98.2%-98.7%).
The tool produced negative predictions in 78% of cases, identifying low ejection fraction in 1.6% of test-negative subjects compared with 30.5% of test-positive subjects, yielding a diagnostic odds ratio of 27.7 (95% CI: 23.4-32.8).

Study Design

Type

Observational (n=13,960)

Multicenter

Yes

Structured PICO

Does an AI-enabled ECG algorithm accurately detect low left ventricular ejection fraction (≤40%) in a diverse clinical population?

P

Population

13,960 adults (mean age 66, 48% female) with paired ECG and echocardiogram within 30 days across 4 US sites, evaluated to validate an AI-ECG algorithm for detecting LVEF ≤40%.

E

Exposure

Artificial intelligence-based electrocardiogram (ECG-AI) algorithm for detection of LVEF ≤40% (software as a medical device).

C

Comparator

Transthoracic echocardiogram reporting quantitative LVEF (reference standard).

O

Outcome

Diagnostic performance (sensitivity, specificity, and AUROC) for detecting LVEF ≤40%.surrogate

Main Result

Effect estimate: AUROC 0.92 (95% CI 0.91-0.93)

External validation of an AI-enhanced ECG algorithm demonstrated high diagnostic accuracy and negative predictive value for detecting low ejection fraction, suggesting utility as a scalable screening or rule-out tool.

Limitations

Protocol written before study sites were selected, limiting prevalence estimation
Paced ECGs were excluded
LVEF measurements were not independently adjudicated, relying on real-world clinical reports
Missing clinical data due to site-specific anonymization practices
Limited number of ECG machine models evaluated (predominantly General Electric and Philips)
Protocol written before study sites selected, creating prevalence uncertainty
Paced ECGs excluded
LVEF measurements not independently adjudicated
Missing data due to anonymization practices
Limited ECG device models tested

Implication

May support ECG-AI as rule-out tool; extends validation yet leaves open outcome trials.

Abstract

BACKGROUND: Low left ventricular ejection fraction (LEF) can progress undiagnosed. Artificial intelligence-based electrocardiogram (ECG-AI) screening may provide a scalable means to detect LEF. OBJECTIVES: The purpose of this study was to validate a complete ECG-AI software as a medical device for LEF detection. METHODS: Four geographically diverse sites in the United States identified patients with both ECGs and transthoracic echocardiograms performed within 30 days of each other in clinical practice. Data were electronically extracted to specific guidelines and transmitted to the coordinating center for analysis. RESULTS: Records of 16,000 subjects were extracted, resulting in an evaluable set of 13,960 subjects (mean age 66 years; 52% male). The device demonstrated excellent discrimination (AUROC: 0.92 [95% CI: 0.91-0.93]) and was 84.5% (95% CI: 82.2%-86.6%) sensitive and 83.6% (95% CI: 82.9%-84.2%) specific for LEF. The overall prevalence of LEF in the study data set was 7.9%, with LEF among 1.6% of the ECG-AI negative and 30.5% of ECG-AI positive subjects, contributing to positive and negative predictive values of 30.5% (95% CI: 28.8%-32.1%) and 98.4% (95% CI: 98.2%-98.7%), respectively. CONCLUSIONS: External validation studies such as this one provide a rigorous framework to validate an algorithm's performance. This study demonstrated the algorithm's strong diagnostic accuracy over a geographically diverse, independent set of patients. In this generally unselected population, the algorithm produced a test negative result in 78% of the cases, suggesting potential utility as a rule-out strategy to defer echocardiography when other clinical findings are absent.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Carter et al. (2026) conducted an observational in Low left ventricular ejection fraction (n=13,960). Artificial intelligence-based electrocardiogram (ECG-AI) algorithm vs. Transthoracic echocardiogram (reference standard) was evaluated on Detection of low ejection fraction (LVEF ≤40%) (AUROC 0.92, 95% CI 0.91-0.93). An AI-enabled ECG algorithm demonstrated excellent discrimination for detecting LVEF ≤40% with an AUROC of 0.92, sensitivity of 84.5%, and specificity of 83.6%.

synapsesocial.com/papers/6a17d8413aabde875b130194 https://doi.org/10.1016/j.jacadv.2025.102537

View Full Paper

Ask AI

Bookmark

Share

Multisite, External Validation of an AI-Enabled ECG Algorithm for Detection of Low Ejection Fraction

Figures

Key Points

Study Design

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Multisite, External Validation of an AI-Enabled ECG Algorithm for Detection of Low Ejection Fraction

Figures

Key Points

Study Design

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading