Selecting and interpreting diagnostic tests

Abstract

Diagnostic tests range from the signs and symptoms obtained from the patient's history and physical examination to the sophisticated laboratory and imaging tests widely used in medical practice today. The typical diagnostic test compares clinical information gathered in a less invasive and/or less costly manner to the so-called gold standard. First, we define the basic components of diagnostic testing - i.e., the test characteristics of sensitivity and specificity and the test performance measures of positive predictive value and negative predictive value. We discuss the traditional use of these components in selecting and interpreting diagnostic test results. We then explain the Bayesian model for diagnostic testing through a discussion of pre-test probability and post-test probability and positive and negative likelihood ratios. We discuss the issue of integrating indicators of a test's characteristics and performance, highlighting the area under the ROC curve, diagnostic accuracy and the diagnostic odds ratio. The impact of prevalence on accuracy and predictive values is clarified. And lastly, the diagnostic odds ratio is presented as a measure of test performance that combines sensitivity and specificity but is independent of test prevalence. Before ordering a diagnostic test, the clinician must evaluate the potential benefits and risks of the test and how the results will alter patient management. Familiarity with the available measurement options for evaluating a test and how to interpret the results are the first steps toward making evidence-based diagnostic decisions.