Digital Stethoscope-Derived Single-Lead Electrocardiogram and Artificial Intelligence to Detect Low Ejection...

Identifying patients likely to benefit from an echocardiogram before surgery is prudent in resource-limited settings. Recently, artificial intelligence has facilitated the diagnosis of low ejection fraction (EF) from electrocardiograms. This multicenter study aimed to evaluate the diagnostic accuracy of a single-lead electrocardiogram-based algorithm obtained by an electronic stethoscope (Eko CORE 500) for identifying individuals with reduced left ventricular ejection fraction (LVEF) (≤40%) during preoperative evaluation.

Multicenter, prospective, observational diagnostic accuracy study.

Ten tertiary care-level hospitals across India.

Nine hundred eighty-one preoperative patients.

None.

Patients referred for a 2-dimensional echocardiogram as part of a preoperative workup were eligible. An Eko CORE 500 stethoscope was used to collect paired single-lead electrocardiograms and heart sounds at the pulmonic auscultation position. After data collection concluded, all recordings were analyzed using the Eko application. If a recording passed an automated electrocardiogram signal quality check, the algorithm categorized it as either LVEF ≤ 40% or LVEF > 40%. Performance was assessed against an echocardiogram-derived LVEF ≤ 40%. A total of 981 subjects undergoing various surgical procedures were included. After 114 recordings deemed to have poor electrocardiogram signal were excluded, the final cohort consisted of 867 subjects. One hundred thirty-eight patients had EF ≤ 40% as determined by echocardiography. The algorithm detected LVEF ≤ 40% with an area under the receiver operating characteristic curve (AUC-ROC) of 0.73 (95% confidence interval [CI], 0.67-0.78), sensitivity of 60.1% (95% CI, 51.8%-67.9%), and specificity of 81.3% (95% CI, 78.4%-84.0%). When patients were stratified by type of surgery, performance was superior in the noncardiac group (AUC-ROC of 0.75 [95% CI, 0.59-0.92] in noncardiac group v 0.70 [95% CI, 0.64-0.75] in cardiac group for LVEF ≤ 40%).

The Eko digital stethoscope shows promise as a clinical decision-making tool for preoperative assessment. The ability to accurately detect reduced EF, especially in the setting of noncardiac surgery, suggests the algorithm's potential utility as a rule-out test to optimize resource allocation for preoperative screening. These findings suggest that its use could be considered in specific patient populations, particularly in resource-limited settings.