Bayesian model for the risk of tuberculosis infection for studies with individuals lost to follow-up

Abstract

OBJECTIVE: To develop a statistical model based on Bayesian methods to estimate the risk of tuberculosis infection in studies including individuals lost to follow-up, and to compare it with a classic deterministic model. METHODS: The proposed stochastic model is based on a Gibbs sampling algorithm that uses information of lost to follow-up at the end of a longitudinal study. For simulating the unknown number of reactors at the end of the study and lost to follow-up, but not reactors at time 0, a latent variable was introduced in the new model. An exercise of application of both models in the comparison of the estimates of interest was presented. RESULTS: The point estimates obtained from both models are near identical; however, the Bayesian model allowed the estimation of credible intervals as measures of precision of the estimated parameters. CONCLUSIONS: The Bayesian model can be valuable in longitudinal studies with low adherence to follow-up.