Effects of exercise test results on physician perceptions of coronary artery disease probability using a threshold analysis.
The incremental diagnostic value of exercise electrocardiographic testing compared with clinical data alone in the diagnosis of coronary artery disease (CAD) was found to be of limited value in a previous study. That study used a computer algorithm for diagnosing disease. Thus, the strict use of the results apply only to computerized disease diagnosis rather than physicians' diagnoses. The aim of the present study was to determine whether exercise test data improve a physician's ability to detect the presence or absence of CAD and 3-vessel or left main CAD, and whether the data effect the decision to perform angiography. The study sample comprised a data base of 312 patients whose clinical data, exercise test results and coronary anatomy were known. Individual cases were presented to 8 cardiologists with and without exercise data. The cardiologists provided estimates of disease probability and were asked whether they would request coronary angiography. Receiver-operating characteristic curves and goodness-of-fit analysis showed better discrimination and calibration of the estimates for the presence of CAD and 3-vessel/left main CAD after exercise test results were known. Individualized probability thresholds for deciding whether to request angiography were determined for each physician. Forty-three percent of patients crossed greater than or equal to 1 threshold, with 64% of the crossings in the correct direction. For estimating the presence of any CAD, 84% of crossings from below to above a threshold were correct, whereas only 36% from above to below were correct. For 3-vessel/left main CAD, 36% of crossings from below to above a threshold were correct, whereas 88% from above to below were correct.(ABSTRACT TRUNCATED AT 250 WORDS)