A nonparametric maximum likelihood estimator for the receiver operating characteristic curve area in the presence of verification bias.

The efficacy of a diagnostic test can be represented by the area under the receiver operating characteristic (ROC) curve. In estimating the ROC curve area, a common problem is verification bias resulting from selectively verifying a subset of patients initially tested. This paper proposes a simple verification bias correction procedure for estimating the ROC curve area and its corresponding variance.