Decision making with clinical diagnostic tests in terms of statistical decision theory and information theory

The value of a diagnostic test for an individual patient is expressed in terms of the test's ability to determine whether the patient does or does not have a disease. For diagnostic tests with a continuous diagnostic variable, various decision thresholds can be selected to adjust the sensitivity and specificity of the diagnostic test to conform with the clinician's concerns (e.g., health costs, financial costs, and diagnostic information of the test). Three methods of statistical decision theory and information theory are presented for critically evaluating the value of a clinical diagnostic test. To provide an example of the application of the three methods we critically evaluate the value of using intraocular pressure (IOP) data from three populations: open angle glaucoma, open angle glaucoma suspect, and no open angle glaucoma. Based on these methods, IOP data are shown to provide a means of indicating the higher probability of the presence of open angle glaucoma (whereupon the clinician would need to conduct further diagnostic tests), but on average IOP data are shown to have very little value as a diagnostic variable.