TrialsgrambetaRCTs are beautiful

Lin AI-ECG alert

2024

AI-Enabled Electrocardiography Alert Intervention and All-Cause Mortality: a Pragmatic Randomized Clinical Trial

In 15,965 hospitalised patients, an AI-ECG alert flagging those predicted to be at high risk of death reduced all-cause mortality within 90 days compared with usual care without the alert.

Result

Efficacy: All-cause mortality within 90 daysHR 0.83 (95% CI 0.70-0.99)

Below 1 favours an AI-ECG alert flagging those predicted to be at high risk of death, above 1 favours usual care without the alert.

Event rates
Absolute difference
0.7 points
NNT
143

Absolute difference and NNT are derived from the event rates.

Also reported: In patients with high-risk ECGs HR 0.69 (95% CI 0.53–0.90) and cardiac death 0.2% vs 2.4% (HR 0.07, 95% CI 0.01–0.56), with more intensive care in the intervention arm.

Original publication
Subtopic
Acute diagnosis & risk alerts
Design
Pragmatic multisite RCT (39 physicians)
Relevance
Important: one of several pillars.
Source
Lin CS, et al. Nat Med. 2024;30(5):1461-1470.doi 10.1038/s41591-024-02961-4

Discussion & critique

The first randomised trial in which an AI-ECG alert reduced a hard outcome, mortality, rather than a diagnostic yield, and the strongest outcome evidence for AI-ECG to date. The authors concede the mechanism is not fully elucidated (more intensive care for flagged patients is the likeliest pathway), the upper confidence bound approaches 1, the trial was open-label, and randomisation involved only 39 physicians, which limits generalisability.