The Electrocardiographic Exercise Test in a Population with Reduced Workup Bias: Diagnostic Performance, Computerized Interpretation, and Multivariable Prediction

Abstract

Empirical scores, computerized ST-segment measurements, and equations have been proposed as tools for improving the diagnostic performance of the exercise test. To compare the diagnostic utility of these scores, measurements, and equations with that of visual ST-segment measurements in patients with reduced workup bias. Prospective analysis. 12 university-affiliated Veterans Affairs Medical Centers. 814 consecutive patients who presented with angina pectoris and agreed to undergo both exercise testing and coronary angiography. Digital electrocardiographic recorders and angiographic calipers were used for testing at each site, and test results were sent to core laboratories. Although 25% of patients had previously had testing, workup bias was reduced, as shown by comparison with a pilot study group. This reduction resulted in a sensitivity of 45% and a specificity of 85% for visual analysis. Computerized measurements and visual analysis had similar diagnostic power. Equations incorporating nonelectrocardiographic variables and either visual or computerized ST-segment measurement had similar discrimination and were superior to single ST-segment measurements. These equations correctly classified 5 more patients of every 100 tested (areas under the receiver-operating characteristic curve, 0.80 for equations and 0.68 for visual analysis; P < 0.001) in this population with a 50% prevalence of disease. Standard exercise tests had lower sensitivity but higher specificity in this population with reduced work-up bias than in previous studies. Computerized ST-segment measurements were similar to visual ST-segment measurements made by cardiologists. Considering more than ST-segment measurements can enhance the diagnostic power of the exercise test. *For members of the Veterans Affairs Cooperative Study in Health Services #016 (QUEXTA) Study Group, see Appendix 2.