A series of 4 new tests (A, B, C, and D) are developed to diagnose hemachromatosis. The operating characteristics for a variety of different cut points for each of the four tests are plotted on a receiver operating characteristic (ROC) curve.
Test A has the best overall performance. Many test outcomes are continuous variables, which are arbitrarily divided at some point into normal and abnormal values by using a cut point. One way to visually compare the operating characteristics of different tests is to plot the cut points of each test on a receiver operating characteristic (ROC) curve. The ROC curve is a visual representation of the true positive rate (sensitivity) plotted as a function of the false positive rate (1.0-specificity) for different cut points. A test with the best sensitivity and specificity for each of its cut points will have a curve that “crowds” the upper left margins of the ROC curve. This concept is particularly valuable when comparing two or more tests. The test with the greatest overall accuracy will have the largest area under the ROC curve and will be located closest to the upper left corner.
- In a comparison of two or more tests on a receiver operating characteristic (ROC) curve, the test with the best overall accuracy for each of its cut points will have the largest area under the ROC curve.