Diagnostic tests in obstetrics: a method for improved evaluation.

Abstract

With the proliferation of diagnostic tests in obstetrics, several recurrent questions arise. How does one determine whether one diagnostic test is superior to another available test? What test cutoff value best separates diseased from nondiseased patients? How much does performance of additional tests assist in arriving at a correct diagnosis? This article reviews a simple yet sophisticated analytic technique, the receiver operating characteristic curve, and demonstrates its application to several obstetric diagnostic tests. Receiver operating characteristic curve analysis is used to select a cutoff value for the 1-hour glucose tolerance test, to compare amniotic fluid tests of fetal lung maturity, and to determine the optimal combinations of factors and overall performance of the fetal biophysical profile. The value of receiver operating characteristic curve analysis lies in providing a clear graphic analysis of the performance of diagnostic tests over their entire range of values. It also provides the starting point for evaluating the costs and benefits of alternative cutoff points in differing clinical settings.