Abstract 15139: Applying an Artificial Intelligence-Enabled Electrocardiographic System for Reducing Mortality: A Pragmatic Randomized Clinical Trial

Introduction: The early identification of high-risk patients for sequent intensive care is a significant challenge.

Hypothesis: We aimed to evaluate the outcome of an artificial intelligence (AI)-enabled electrocardiogram (ECG) in identifying patients at high risk of mortality.

Methods: In this single-blind, patient-level randomized controlled trial (NCT05118035), we recruited 39 attending physicians and their patients from emergency and inpatient departments. The AI-ECG intervention involved an AI report in electronic health records and an active warning message to physicians for high-risk patients. Primary endpoint was 90-day all-cause mortality. Secondary analyses included medical behavior changes and causes of death.

Results: Data from 15,965 patients (8,001 intervention; 7,964 control) with a mean age of 61±18 years were included. The intervention group had a significantly reduced cumulative proportion of death (HR: 0.83, 95% CI: 0.70-0.99) compared to the control group (3.6% vs. 4.3%). High-risk cases (709 intervention; 688 control) identified by AI-ECG showed a 31% reduction in mortality (16.0% in intervention arm vs. 23.0% in control arm, HR: 0.69, 95% CI: 0.53-0.90; p for interaction = 0.026). The intervention group with high-risk ECGs received more intensive care, arrhythmia interventions, echocardiographic and electrolyte examinations, contributing to a significant reduction in cardiac death (0.2% in intervention arm vs. 2.4% in control arm, HR: 0.07, 95% CI: 0.01-0.56).

Conclusions: AI-ECG effectively identifies high-risk patients, facilitating intensive care and resulting in reduced all-cause mortality.