abstract
- Medical diagnosis involves generating a set of hypotheses and obtaining information that modifies these hypotheses. Sources of this information include the history, physical examination, and laboratory investigations, all of which function as diagnostic tests. Studies of diagnostic tests are useful when a) the population under study is representative of those to whom we would like to apply the results; b) when an independent, blind comparison is made of the test results with a reference standard; and c) when the reference standard is performed on all patients, rather than restricted to those patients with particular test results. Clinicians can use the data from such high quality studies in the form of sensitivity and specificity, as well as likelihood ratios, which indicate the direction and magnitude of the change in probability of a target condition from pretest to posttest. Study results will be more easily applicable to practice when the performance and interpretation of the test is similar in study and clinical settings. We conduct diagnostic tests primarily to improve the process of patient care and patient outcome, and test ordering behavior ideally reflects these goals.