Stress-Echocentric Diagnostic Algorithms

Abstract

The bayesian analysis of diagnostic testing combines the information on sensitivity and specificity with the prevalence of the disease in the population under study [1]. In this analysis the probability of having the disease before the test is considered the a priori (pretest) likelihood since it can be estimated by retrospective observations. After the test, the new value of probability of the patient having the disease will be the post-test likelihood. For instance, a positive exercise electrocardiography test indicates a probability of coronary artery disease of 90% in a patient with typical angina, of 80% in a patient with atypical chest pain, and of 35% in an asymptomatic person. With a useful diagnostic test the probability of the disease will be very high after a positive test and very low after a negative test. The bayesian approach has been widely criticized because of the lack of independence between clinical data and test results and because of the intrinsic inability of any probability theory to take the place of clinical reasoning [2]. However, a probabilistic approach to the diagnosis of coronary artery disease has the merit of offering a quantitative substrate to any attempt to rationalize the diagnostic process [3]. In the bayesian analysis the area within the post-test curves of probability will graphically display the overall diagnostic usefulness of the test.

What is simple is always false; what is complicated, is impossible to use

Paul Valéry