ABSTRACT
This study examined the feasibility of using indices obtained from a long term Holter ECG record for sudden cardiac arrest (SCA) risk stratification. The ndices tested were the QT-RR interval co-variability and the alternans ratio percentile (ARP(θ)) which is defined as the θ(th) percentile of alternans ratios over a 24 hour period. The QT-RR interval co-variabilities are evaluated by the serial correlation coefficient between QT and RR trend sequences (QTRC). Previously reported Kalman filter technique and a simple smoothing spline method for the trend estimation are compared. Parameter θ in the alternans ratio percentile index was optimized to achieve the best classification accuracy. These indices were estimated from 26 cardiovascular outpatients for Holter ECG record. Patients were classified into high and low risk groups according to their clinical diagnosis, and the obtained indices were compared with those of 25 control subjects. A risk stratification using the two indices QTRC and ARP(θ) yielded an average sensitivity of 0.812 and a specificity of 0.925. The sensitivities and specificities of all three categories exceeded 0.8 except for the sensitivity to detect the high-risk patient group. Other short-term ECG parameters may need to be incorporated in order to improve the sensitivity.
