Verification Bias
Year: 2022
Type: article
Abstract: Sometimes it is not feasible to obtain disease status verification for all study subjects. Analysis of only those with disease ascertainment can result in biased estimates of the accuracy (sensitivity, specificity, ROC curve) of a diagnostic test, screening test, or biomarker if the estimation method does not properly account for the missing disease ascertainment. This paper discusses the impact of this bias, verification bias, when estimating the accuracy of dichotomous and continuous diagnostic tests. In addition, methods to correct for verification bias are described. Areas that require additional attention are also highlighted.
Author: A. Alonzo, Todd
Institution: University of Southern California
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4384921909