Diagnostic accuracy in cytopathology

Cytopathologic diagnoses play an important role in clinical decision making. As a result, the tests that produce material for cytopathologic interpretation must be evaluated critically and carefully. One way to evaluate these tests is by the assessment of diagnostic accuracy. The commonly used sensitivity and specificity format dichotomizes the data set and often loses or misinterprets important diagnostic information, especially the data which is semiquantitative. In addition, sensitivity and specificity do not allow the clinician to revise the probability of disease depending upon the test result. The LR and ROC analysis are two methods that systematically measure the diagnostic accuracy of tests, including those that produce semiquantitative data. LR and ROC measurements incorporate the concepts of sensitivity and specificity, but, in addition, provide a way to assess the clinical utility of a test by estimating the post-test probability of disease. These methods are related to other techniques in medical decision making such as decision analysis and information theory, and are straightforward and easily utilized by clinicians and pathologists.