[Measures for the performance of diagnostic tests]

Systematic reviews have revealed that the methodological quality of studies on diagnostic accuracy is mediocre. In addition, clinicians may experience difficulties in interpreting and using measures describing the accuracy of a test. Using the Ottawa Ankle Rule as an example, we review the measures commonly used to describe the accuracy of diagnostic tests. The performance of tests is often given in terms of sensitivity and specificity. However, these measures have no relevance to clinicians unless they can be converted into predictive values. We describe how to calculate the predictive values and how they can be determined using likelihood ratios and Fagan's nomogram. The reader is introduced to the critical appraisal of results based on studies of the accuracy of tests. We describe how both the clinical spectrum and the methodological quality can influence estimates of diagnostic accuracy and stress the inevitable uncertainty involved in extrapolating results from the literature to clinical practice. Methods of calculating the confidence intervals for estimates of accuracy are presented.