Screening for Left Ventricular Dysfunction in Sepsis Using a Smartphone ECG Analysis Application: A Multicenter Validation Study.

BACKGROUND: Sepsis-induced left ventricular (LV) dysfunction complicates fluid resuscitation, increasing the risk of pulmonary edema and worsening outcomes. Echocardiographic assessment is often delayed, and natriuretic peptides such as N-terminal pro-B-type natriuretic peptide (NT-proBNP) have limited diagnostic accuracy in this setting. We evaluated a smartphone-based electrocardiogram (ECG) analysis application for screening LV dysfunction in patients with sepsis. METHODS: In this multicenter retrospective observational study, we included 425 adult sepsis patients from three university hospitals who underwent both a 12-lead ECG and transthoracic echocardiography within 72 hours of emergency department admission. ECG images displayed on a monitor were photographed and analyzed using an artificial intelligence (AI)-based smartphone application (ECG Buddy™), which generated a quantitative digital biomarker for LV dysfunction (quantitative electrocardiogram scores for LV dysfunction [QCG-LVDys]). LV dysfunction was defined as a left ventricular ejection fraction ≤ 40%. The diagnostic performance of QCG-LVDys for detecting LV dysfunction was evaluated using area under the receiver operating characteristic curve (ROC-AUC). A predefined comparison with NT-proBNP was performed in the single center where NT-proBNP was routinely measured. RESULTS: In the overall multicenter cohort, QCG-LVDys showed high diagnostic performance for LV dysfunction, with an ROC-AUC of 0.901. Using an optimized threshold (37.35), QCG-LVDys achieved a sensitivity of 89.6%, specificity of 75.1%, and negative predictive value (NPV) of 97.5%. In the NT-proBNP subset, QCG-LVDys outperformed NT-proBNP (ROC-AUC 0.911 vs. 0.637, P < 0.001). Higher QCG-LVDys values were associated with increased lactate levels (P = 0.002) and 28-day in-hospital mortality (P < 0.001). CONCLUSION: AI-driven ECG analysis via a smartphone app provided a rapid, accessible, and accurate screening tool for LV dysfunction in sepsis and outperformed NT-proBNP. Its high NPV suggests potential utility for early risk stratification and for informing cautious hemodynamic management, although prospective studies are needed to define its role in sepsis treatment protocols.