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 coronary 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 (≥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.