Table 2 Diagnostic accuracy studies

In diagnostic accuracy studies, a series of patients suspected of having a target condition undergo both an index test (i.e., the test that is being evaluated) and the clinical reference standard (i.e., the best available method for establishing if a patient does or does not have the target condition) [6].

Assuming that the results of the index test and reference standard are dichotomous—either positive or negative—we can present the results of the study in a contingency table (or “2 × 2 table”), which shows the extent to which both tests agree (Fig. 1). Discrepancies between the results of the index test and the reference standard are considered to be false-positive and false-negative index test results.

Although it is possible to generate a single estimate of the index test’s accuracy, such as the diagnostic odds ratio [7], it is usually more meaningful to report two statistics: one for patients with the target condition, or sensitivity, and one for patients without the target condition, or specificity (Fig. 1). One reason is that the clinical consequences of misclassifications from false-positive and false-negative test results usually differ. As a visual aid, we can picture a test’s sensitivity and specificity as a point in the receiver operating characteristic (ROC) space, which has these two dimensions: sensitivity (y-axis) and 1-specificity (x-axis) (Fig. 2).