PROC Analysis: Effect of Threshold Value for Diagnostic Test

Abstract

Binary test outcomes are typically considered as dichotomized versions of a continuous test variable, observable or latent. The effect of the threshold for test positivity on test sensitivity and specificity has been studied extensively in ROC analysis. Considerably less attention has been given to the study of the effect of the positivity threshold on the predictive value of a test. In this talk I will present statistical methods for the joint study of the positive and negative predictive value of diagnostic tests. We define the Predictive Receiver Operating Characteristic (PROC) curve, which summarizes the pairs of positive and negative predictive value as the threshold for test positivity varies. The properties of the PROC curve provide understanding of the effect of the decision threshold for test positivity, as well as the interplay between the positive and negative predictive value as this threshold changes. We continue the analysis of the predictive accuracy with the presentation of the predictive model, which is used to estimate the positive and negative predictive value jointly. Then we propose summary measures to quantify and compare predictive performance of diagnostic tests using the predictive model and the PROC curve.