ASSESSING THE DIAGNOSTIC UTILITY OF A TEST

Given the sensitivity and specificity of a binary test, Lindley and Novick's criterion of diagnostic utility is characterized by the range R of prevalence values in which opposing test results are in correct correspondence to opposing posterior odds on the disease. If the utility range of a new test includes the utility range R of an alternative test, then the new test is uniformly at least as useful as the standard, for all prevalence levels in R. A descriptive interpretation of the criterion is suggested and its statistical analysis is discussed.