A mathematical approach to interpretation and selection of diagnostic tests

When the presence of a specific disease is being considered, a diagnostic test can often help the physician to choose between subjecting the patient to an extensive workup (or treatment) and proceeding no further. Decisions concerning the use of a test require that three judgments be made: (1) Should the test be done? (2) Which test (if any) should be used if two or more are available? (3) If a test that can take on more than two values is performed, what is the correct cutoff point (or positivity criterion) that determines whether or not to proceed? This paper presents a mathematical technique to answer these questions, taking into account data concerning the patient (summarized as the prior, or pre-test, probability of disease), the test, and the workup (or treatment). The technique is presented in a graphical form that can be applied to any clinical situation in which the needed data are available.