WITHDRAWN: Probit Latent Class Models for Evaluating Accuracy of Diagnostic Tests with Indeterminate Results

• Author(s): Huiping Xu and Bruce A. Craig
• Keywords: Diagnostic tests ; indeterminate results ; sensitivity ; specificity ; latent class analysis

Abstract

Indeterminate results often occur with diagnostic tests. When assessing diagnostic accuracy, it is important to properly report and account for these results. In the literature, these results are commonly discarded or treated as either a positive or negative result, which limits the usefulness of the results. Estimation of diagnostic accuracy is further complicated when a gold standard is not available. Traditional latent class models can be applied to analyze these data. However, these models assume independence of tests conditional on the disease status, which is rarely valid in practice. We propose a probit latent class model that allows arbitrarily structured correlations among tests, while taking into account the indeterminate results. The maximum likelihood estimates are obtained using the MCEM algorithm. We demonstrate our model using a simulation study and a medical study.