THE RELATIONSHIPS AMONG PERFORMANCE-MEASURES IN THE SELECTION OF DIAGNOSTIC-TESTS

Diagnostic tests are typically evaluated using performance measures, s uch as sensitivity, specificity, positive and negative predictive valu es, accuracy, and likelihood ratios. Numerous criteria have been sugge sted as to the types of diagnostic scenarios for which each of these m easures is most important. Reports in the medical literature will ofte n characterize a test by the values of some, but not all, of these per formance measures. At times the unreported measures can be of interest in evaluating whether to use a test. A set of formulas is presented w hich, in many cases, enables one to determine unreported measures from those provided, for example, accuracy from sensitivity, specificity, and positive predictive value. The relationships between the measures of diagnostic test effectiveness and the prevalence of disease are dis cussed. An application to the diagnosis of acute myocardial infarction with new biochemical markers is used to illustrate these relationship s.