The impact of an ECG computer analysis program on the cardiologist's interpretation. A cooperative study

Nine experienced electrocardiographers and the ECG computer program developed in the Veterans Administration (AVA 4.0) were evaluated against ECG-independent evidence of 180 patients' true diagnoses. A cross section of cardiac abnormalities was included. Each reader was given the 12-lead and orthogonal 3-lead ECG. The impact of ECG computer reports on the interpretations by the nine readers was evaluated by comparing their interpretations before and after the addition of a computer report. Using only high probability statements, the average accuracy of ECG diagnosis by the nine readers was 54%. It increased to 62% when the computer report was added. Computer interpretation was correct in 76%. It was shown that the Bayesian classification method together with multivariate analysis, used in the VA program, are mainly responsible for the improvement in diagnostic accuracy.