Local versus Global Measures of Accuracy

Abstract

Unlike many tests, diagnostic imaging provides information about the location of disease, in addition to its presence or absence. To account for this ability of imaging, the authors make the distinction between local and global measures of accuracy. Local pertains to each anatomic site of potential disease in the patient and the corresponding image, while global refers to all such sites, collectively. After making this distinction, the authors demonstrate two major problems with global sensitivity and specificity, which are more commonly reported than their local counterparts. First, global sensitivity, unlike local sensitivities, varies with the anatomic distribution of disease in the reference population. Second, the conventional global measures cannot be integrated with pretest knowledge about the anatomic distribution of disease. Consequently, predictive values based on global sensitivity and specificity under estimate the probability of disease when imaging and clinical findings correlate anatomically and overestimate the probability when the findings do not correlate. The local versus global distinction supports the commonsense notion that information pertaining to the anatomic distribution of disease is crucial for test interpretation.