Two dependent diagnostic tests: Use of copula functions in the estimation of the prevalence and performance test parameters

Abstract

In this paper, we introduce a Bayesian analysis to estimate the prevalence and performance test parameters of two diagnostic tests. We concentrated our interest in studies where the individuals with negative outcomes in both tests are not verified by a gold standard. Given that the screening tests are applied in the same individual we assume dependence between test results. Generally, to capture the possible existing dependence between test outcomes, it is assumed a binary covariance structure, but in this paper, as an alternative for this modeling, we consider the use of copula function structures. The posterior summaries of interest are obtained using standard MCMC (Markov Chain Monte Carlo) methods. We compare the results obtained with our approach with those obtained using binary covariance and assuming independence. We considerate two published medical data sets to illustrate the approach.