Interpreting diagnostic accuracy studies for patient care.

plot of sensitivity against 1−specificity (or false positive rate). Figure 2 shows an ROC plot of test accuracy of a single test at different thresholds. ROC plots are also used within studies to compare different tests, to compare different groups of patients, and to investigate variability between different test observers (readers). ROC plots are useful in systematic reviews to present results from m ultiple studies. Several concepts need to be considered carefully in the interpretation of data from a diagnostic accuracy study: • How does accuracy change with different diagnostic thresholds? • If paired outcomes (such as sensitivity and specificity) are compared for different scenarios, they often change in opposite directions. For example, sensitivity is often higher in one test and specificity higher in the other. Which is more important? • What are the clinical consequences of a missed (false negative) diagnosis or a false positive diagnosis? Can these risks be presented together—for example, as a relative benefit? • What is the best way to include disease prevalence in the summary of clinical benefit? • Are results presented in terms of what happens to individual patients, which are often the easiest for clinicians (and their patients) to understand?