A collapsing net benefit approach for comparing diagnostic tests of multistage clinical conditions

Using accuracy measures alone to compare diagnostics tests may be unconvincing to clinicians. Diagnostic tests accuracy is commonly evaluated in a clinical performance as its classification accuracy (specificity, sensitivity, negative and positive likelihood ratio) or its predictive values (negative and positive predictive value). However, the limitation of those measures is that one test may have a better sensitivity and worse specificity than another test. Comparing tests based on the net benefit approach by using benefit-risk measures is another approach where benefits and risks are put on the same scale to determine whether a test has better, worse, or the same when assessing a test’s clinical consequences. Consequently, evaluating diagnostic tests based on benefit-risk involves both the tests\u27 accuracy and the clinical implications of the diagnostic errors. Diagnostic tests are commonly classified into two stages: either positive or negative for a clinical condition (diseased or non-diseased). However, some diseases have more than two stages, such as Alzheimer\u27s. The benefits and risks of the clinical consequences could differ from stage to stage. This study extends the net benefit approach of evaluating diagnostic tests in binary disease cases to multistage clinical conditions. Consequently, we extend the diagnostic yield table to multistage clinical conditions. We develop a decision process based on net benefit for evaluating diagnostics tests. It provides additional interpretation for rule-in or rule-out clinical conditions and their adverse consequences from unnecessary workup in multistage diseases. Numerical examples, as well as real data, are provided to illustrate the proposed measures