A semi-parametric joint latent class model with longitudinal and survival data

In many longitudinal studies, we are interested in both repeated measures of a biomarker and time to an event. When there exist heterogeneous patterns of the longitudinal and survival profiles, we propose a latent class joint model to identify subgroups of subjects and study the association between longitudinal and survival outcomes. The model is estimated by maximizing the full likelihood function. We use B-splines to approximate the baseline hazard function which involves a diverging number of parameters. Asymptotic properties of the estimator for the joint latent class model are investigated. We conduct simulation studies to assess the performance of the developed method. A real data example, Mayo Clinic Primary Biliary Cirrhosis Data, is analyzed using the joint modeling approach.