Measures of gain in certainty from a diagnostic test.

The authors propose several measures for quantifying the change in the clinical estimate of a patient's chances of having a disease that occurs as a result of diagnostic testing. Under most circumstances, the gain in clinical certainty from a positive test result is more affected by the specificity (T) of the test, while the gain from a negative test result is more affected by sensitivity (S). The prevalence of the disease in the tested population is also an important determinant of the magnitude of gain in certainty. Measures of the expected gain in certainty can be calculated by weighting the gains from a positive or negative result by the likelihood of the respective test outcome. Indices of expected gain depend directly on the quantity S + T, implying that sensitivity and specificity have equal importance in determining expected gain. When S + T = 1, the test provides no information; when S + T is greatest, the expected gain is maximized. Expected gain is also related to the receiver operating characteristic curve for a diagnostic test: the point on the receiver operating characteristic curve at which S + T is greatest corresponds to the point at which the distance from the major diagonal is greatest at which the slope of the receiver operating characteristic curve equals 1.