The hazards of bedside Bayes.

Those who advocate the use of Bayesian or decision-analysis approaches to solve clinical problems often assume that data on the sensitivities and specificities of diagnostic tests are readily obtainable and reliable. For the most part, however, there is little information currently available on the sensitivities and specificities of common tests. A critical review of the literature was performed for seven commonly used tests. This study showed that there is significant variability among the reported results of five of these tests. Such variability was not expected but is one more factor that any quantitative approach must consider. Two tests, the rapid-sequence excretory urogram and the thallous chloride TI 201 cardiac stress test, did not, however, show significant variation among the reported sensitivities and specificities. These findings have relevance to clinicians attempting to diagnose disease, to advocates of quantitative decision making, and to researchers seeking to clarify the nature and role of diagnostic tests.