Mixed effects models with censored covariates, with applications in HIV/AIDS studies (Q1658195)

Summary: Mixed effects models are widely used for modelling clustered data when there are large variations between clusters, since mixed effects models allow for cluster-specific inference. In some longitudinal studies such as HIV/AIDS studies, it is common that some time-varying covariates may be left or right censored due to detection limits, may be missing at times of interest, or may be measured with errors. To address these ``incomplete data'' problems, a common approach is to model the time-varying covariates based on observed covariate data and then use the fitted model to ``predict'' the censored or missing or mismeasured covariates. In this article, we provide a review of the common approaches for censored covariates in longitudinal and survival response models and advocate nonlinear mechanistic covariate models if such models are available.