Is computer interpretation of the exercise electrocardiogram a reasonable surrogate for visual reading?

Background: Interpretation of exercise tests as positive or negative is primarily based upon exercise-induced ST segment changes. Consistently accurate measurements are difficult to obtain during exercise. Hypothesis: This study compared on-line computer-generated electrocardiographic (ECG) analysis with visual interpretation. The goals were to document the extent of agreement, establish reasons for disagreements, characterize ST-segment depression (extent, onset, duration), and determine the sensitivity and ability to localize corollary artery disease for each method. Methods: Comparisons were made in 120 patients at eight Veterans Affairs Medical Centers. An exercise test was considered positive if > 1.0 mm horizontal or downsloping ST-segment depression was detected 0.08 s after the J point during exercise or recovery. The ST-segment depression had to be present on at least two successive ECG recordings 15 s apart. Computer interpretation was based on median averaged beats. Results: There was an 88% agreement of visual and computer interpretations [106/120 (both positive, n = 62; both negative, n = 44)]. The disagreements involved visual negative, computer positive in 10 cases and visual positive, computer negative in 4 cases. Correlation was excellent between methods for characterization of ST-segment depression (p < 0.0001). Sensitivity for detecting and the ability to localize coronary artery disease (greater than or equal to 70% stenosis) were similar for both methods. Conclusion: This computer algorithm using median averaged beats is a reasonable surrogate for visual interpretation of the exercise ECG, making it a valuable source of confirmation of physician readings in large research trials and in clinical settings.