Abstract 3619: Added Value of the Resting ECG Neural Network (RENN) Score
Background: Cardiovascular mortality (CVM) remains the most common cause of death in the United States. A Resting ECG Neural Network (RENN) score was recently created that uses 132 computerized ECG characteristics and 4 baseline demographic variables in an artificial neural network to predict CVM.
Objectives: To risk stratify patients referred for exercise testing with the RENN score and to demonstrate added value of this scoring system.
Methods: A total of 2250 patients referred to the Palo Alto VA between 1997 and 2004 for treadmill testing were studied. RENN scores were calculated from the baseline resting ECGs using the previously created neural network. Subjects underwent treadmill exercise testing using an individualized ramp protocol and were followed for CVM. The Kaplan-Meier method to estimate cumulative survival and multivariate Cox proportional hazard regression analysis to estimate hazard ratios were used.
Results: Over a mean of 8.6 years, there were 163 cardiovascular deaths. The patients were on average 56 years old, 11% diabetic, 13% hyperlipidemic and 49% hypertensive. Kaplan-Meier analysis showed a 10-year mortality of 2%, 8% and 27% for low (RENN <0.4), moderate (RENN 0.4 – 0.8) and high (RENN >0.8) -risk RENN groups. Cox regression models adjusted for age, sex, diabetes, hypertension, hyperlipidemia and smoking status revealed HR of 1.9 (95% CI 1.2 – 3.3) for moderate risk and 6.8 (95% CI 4.0 – 11.7) for high risk groups relative to the low risk RENN group. Among the 735 patients classified into the moderate risk Duke Treadmill Score (DTS) tertile, 304 (41%) of the patients were classified into lower risk and 88 (12%) into higher risk categories by the RENN score. The predictive accuracy of the RENN score in the mid DTS tertile by C index was 0.71.
Conclusion: The RENN score is able to accurately risk stratify patients referred for exercise testing and to successfully reclassify approximately half of the patients with an intermediate DTS risk. The computer-generated RENN score, if incorporated into existing ECG recorders, could be used clinically to complement exercise testing and direct further medical care.