Quantitative information of specific diagnostic tests

The goal of diagnostic testing is to maximize information (I) of a specific disease of interest (D) resulting from the performance of a specific diagnostic test procedure (T). However, all tests suffer from errors which result in incomplete information and inaccurate diagnostic conclusions. The methods of Information Theory have successfully solved a range of signal transmission problems involving physical systems operating under conditions of noise. Medical testing procedures were found to be analogous to noisy systems; hence, Information Theory methods were applied to minimize errors in diagnostic testing. Prior to performing a diagnostic test, the quantity of information, Apriori Information (ID), regarding the presence or absence of the disease was only a function of the prevalence (P) of disease in the population. After performing a diagnostic test, the quantity of information, Aposteriori Information (IDIT) was a function of not only P but also the test sensitivity (A) and test specificity (B). The quantity of information gained by test performance was computed from the difference between aposteriori and apriori information. delta I = IDIT - ID To illustrate the relationship between pretest, apriori, and post-test, aposteriori, information the values of ID and IDIT were computed for five common cardiovascular tests applied to populations with different coronary artery disease prevalence. This cross-sectional analysis studied the quantitative information obtained from Electrocardiography (ECG), Bicycle Ergometer Stress Tests (BEST), Stress-Echo Ergometer Tests (SEET), Thallium 201 Stress Tests (ThST), and Coronary Arteriography (CorA). Apriori information ranged from a minimum of 0 (for prevalence = 0.5) to a maximum of 1.0 (for prevalence = 0 or 1.0) The aposteriori information was computed for all apriori information and occupied the range between that of a "perfect" test (A = 1.0 and B = 1.0) and that of a "worthless" test (A = 0.5 and B = 0.5). All tests demonstrated greater information gain when apriori information was minimal; Little additional information could be gained when the apriori information was close to certainty (i.e., for prevalence near 0 or 1.0). Electrocardiography demonstrated little significant IDIT at any value of ID. Thallium 201 stress tests provided similar aposteriori information values to those of Stress-Echo Ergometer Stress and either demonstrated greater information gain than Bicycle Ergometer Stress Tests. Coronary Arteriography provided the maximum values of aposteriori information. Information Theory methods provided an effective quantitative method to compare the effectiveness of diagnostic tests over a wide range of disease prevalence.