The evaluation of diagnostic tests: principles, problems, and new developments.

This article describes the use of probability in reasoning about diagnostic test results and the importance of accurate measures of test performance. Most studies of test performance are retrospective analyses of unplanned observations. Retrospective studies are likely to result in biased estimates of test performance. This article shows how to adjust for biases in the results of retrospective studies and how to design a prospective study that will accurately measure the sensitivity and specificity of a test.