Risk Prediction of 10-Year Cardiovascular Mortality Using Automated Electrocardiographic Analysis in the National Health and Nutrition Examination Survey

Introduction: Several individual ECG parameters are predictive of cardiovascular disease (CVD) mortality, and when combined in a score, may improve risk prediction. Hypothesis: We hypothesized that an ECG risk score (based on automated measures available on many modern ECG machines) can effectively predict CVD mortality, and augment risk classification when added to traditional risk factors (TRF). Methods: We examined 6786 individuals aged 40-79 years without known CVD in NHANES III (1988-1994) followed 10 years for CVD mortality. Six pre-specified ECG variables (“ECG model”), were evaluated including P, R, T axes, QT interval, QRS width, and heart rate. Three risk scores were developed for variables containing: 1) TRF only, 2) ECG factors (including age, sex, and race), and 3) TRF+ECG combined. Validation was performed using a cohort aged 40-74 years from NHANES I (n=3773), enrolled in 1971-1974. Age stratification was performed with cutoff of 65 years because of increased competing risks in older individuals. Results: During 10 years of follow up, 384 CVD deaths occurred. Frontal QRS-T angle, T-axis, wide QRS interval (cutoff 120 ms), heart rate-corrected QT interval, and heart rate were found to be significant predictors (p<0.05) of CVD death. As per the table, the ECG score had similar performance compared to the TRF score for the subgroup < 65 years of age in the derivation and validation cohorts. The score with combined ECG + TRF had the best performance for those aged < 65 years, and resulted in a net reclassification index of 9% in the validation cohort using cutoffs of 7.5% and 20% for low, intermediate, and high risk categories. In those aged ≥ 65 years, the combined score showed improvement vs. the TRF score in the derivation, but not validation cohort. Conclusion: A risk score based on routinely reported automated ECG variables and TRF predicts risk of 10-year CVD death better than TRF alone in a cohort age < 65 years without known CVD.