Decision Curve Analysis

Decision curve analysis is a straightforward technique for evaluating diagnostic tests, prediction models, and molecular markers. Unlike traditional biostatistical techniques, it can provide information as to a test's clinical value, but unlike traditional decision analytic techniques, it does not require patient preferences or formal estimation of the health value of various health outcomes: Only a general clinical estimate is required. Differences between biostatistical techniques, decision-analytic techniques, and decision curve analysis are shown in Table 1.

A common clinical problem is when a physician can easily obtain information about T—the result of a diagnostic test, the level of a molecular marker, or a probability from a statistical prediction model—but wants to know D, whether or not a patient has, or will develop, a certain disease state. From ...