Design: Systematic review and meta-analysis of three randomised trials
Population: 37,467 adults without known heart failure having routine ECGs
Primary outcome: New echocardiography-confirmed ejection fraction of 50% or less
Effect: RR 1.36 (95% CI 1.12–1.64); number needed to screen 204; echo referral RR 1.02

A meta-analysis pooled three randomised trials, two individually randomised and one cluster trial, of 37,467 adults without known heart failure having ECGs as part of routine care. In the intervention arms, an AI algorithm flagged ECGs suggesting low ejection fraction to the clinician.

New echocardiography-confirmed low ejection fraction (50% or less) was diagnosed more often with AI-ECG screening (RR 1.36, 95% CI 1.12–1.64; I² 0%), a number needed to screen of 204. Echo referral rates did not rise overall (RR 1.02). The tool's positive predictive value was about 20–34%, so most flagged patients did not have low ejection fraction.

Left ventricular systolic dysfunction is common, treatable and often silent. If an ECG is already being done, AI reading turns it into a screen at almost no extra cost. The catch is platform and access: the trials used specific validated algorithms, these tools are not widely available in Indian primary care, and an AI flag needs an echo pathway behind it.

Where an AI-ECG tool is available, act on a low-EF flag with an echocardiogram
Expect most flags to be false positives — the positive predictive value was about 20–34%
Do not use AI-ECG to rule out heart failure in a symptomatic patient
Make sure an echo pathway exists before adopting AI-ECG screening
Check that any tool used has been validated in a population like yours

Why it matters

It turns an ECG already being done into a screen for a treatable, silent condition.

The statistics, in plain English

A risk ratio of 1.36 means about a third more new diagnoses. A number needed to screen of 204 means 204 people screened for one extra case found. An I² of 0% means the three trials agreed.