An Information Theory Approach to the Diagnosis of the Presence and Severity of Coronary Artery Disease

Abstract

SummaryNon-invasive testing for the diagnosis of coronary artery disease (CAD) is usually based on the exercise ECG alone or combined with thallium scintigraphy, while coronary angiography generally represents the final work-up.In this paper, we evaluate the relative merits of each diagnostic strategy in terms of its financial cost and of the reduction in diagnostic uncertainty. The uncertainty level before and after test is measured with the help of concepts used in information theory. The basic inputs of the program are:1) the prior probability of CAD (p) and of its severity, obtained from a large data bank (CADENZA);2) conditional probabilities of the presence and severity of CAD according to exercise ECG and thallium scintigraphy results obtained from the Louvain data bank; and3) the financial cost of each technique as well as society’s willingness to pay (WTP) for a reduction in diagnostic uncertainty.We present a graph showing the optimal strategy in terms of p and WTP. When information is sought only about the presence of CAD, arteriography is hardly ever the preferred test. However, once the severity of CAD also matters, it is the best procedure for values of p exceeding .75. Below this level, thallium scintigraphy is the most effective approach.