Diagnosing diastolic dysfunction using explainable advanced electrocardiography

BACKGROUND AND AIMSDiastolic dysfunction is a precursor to heart failure with preserved ejection fraction (HFpEF), and early detection by electrocardiography (ECG) would be valuable. We hypothesised that an explainable advanced ECG (A-ECG) score could accurately detect diastolic dysfunction with clinically meaningful diagnostic and prognostic performance. METHODSA derivation cohort was included after standard 12-lead ECG and echocardiography demonstrating normal systolic function, and either the presence or absence of diastolic dysfunction ([&ge;]grade 2, 2025 guidelines). A multivariable machine learning A-ECG diastolic dysfunction score (0-100%, positive score defined as >50%) was derived using elastic net with nested resampling. Performance for identifying diastolic dysfunction was assessed in a separate external validation cohort. Prognostic performance for predicting cardiovascular events including mortality was evaluated in the UK Biobank. RESULTSA 7-measure A-ECG diastolic dysfunction score derived in the derivation cohort (n=414) performed excellently in the validation cohort (n=3,418, area under the receiver operating characteristics curve [95% confidence interval] 0.87 [0.86-0.88]). Survival analysis using spline modelling adjusted for age, sex, and cardiovascular risk factors in the UK Biobank (n=27,239, 966 events, follow-up 1.9 [0.7-4.4] years, age 66{+/-}8 years, 50% female) showed that over the range of the score (0-100%), the hazard ratio for events increased linearly to 4 (p<0.001) CONCLUSIONSExplainable A-ECG analysis of the standard 12-lead ECG can be used to accurately identify diastolic dysfunction. The score reflecting the continuous probability (0-100%) of having diastolic dysfunction has a linearly increasing and strong adjusted prognostic association with cardiovascular events.