Comparison of the performance of three diagnostic algorithms for regular broad complex tachycardia in practical application

The authors previously proposed a Bayesian approach to the electrocardiographic diagnosis of regular broad complex tachycardia (BCT), which can be due to VT or supraventricular tachycardia with aberrant conduction (SVTAC). They also published an account comparing the theoretical merits in the design of two of the most commonly used diagnostic algorithms for the same purpose, those of Brugada et al. and Griffith et al. In this study, a direct head-to-head comparison was performed on the practical performances of the three algorithms in this study. A set of 111 ECGs showing regular BCT (77 VT, 34 SVTAC) whose diagnoses were confirmed by electrophysiological study was shown to five internists in general medicine at a district general hospital. The observers were asked to comment on whether the ECG criteria in the three algorithms tested were fulfilled or not, and a computer program then derived the corresponding diagnoses. The sensitivity and specificity for VT achieved by the Brugada algorithm were 92% and 44%, 92% and 44% by the Griffith algorithm, and 97% and 56% by the Bayesian algorithm. The Bayesian algorithm achieved a higher sensitivity and specificity than the other two algorithms, but the differences are not statistically significant (P = 0.6583 and P = 0.5334, respectively). The Brugada, Griffith, and Bayesian algorithms show comparable performances in terms of overall sensitivity and specificity when tested in practice. Of the three algorithms, the Griffith algorithm excels in simplicity and is the easiest to implement in practice. The Bayesian algorithm achieved slightly higher values for sensitivity and specificity than the Brugada and Griffith algorithms but may be more suitable for automated computer-aided diagnosis of ECG due to its complexity.