Joint Modeling of Survival and Longitudinal Data

Abstract

In clinic trial or medical follow up studies, it has become increasing common to observe event time of interest and longitudinal covariates simultaneously. The interest of the studies is not only both survival and longitudinal processes but also their relationship. However, due to several complications, traditional approaches including partial likelihood approach for Cox model and rank based approaches for accelerated failure time model encounter difficulties in these studies. Jointly modeling the survival and longitudinal data emerges as a solution to overcome these difficulties. In this talk, a brief introduction of joint model will be given. Moreover, a general class of semi-parametric hazard regression termed “extended hazarded model” for joint model is proposed for more flexibility in survival model. The extended hazard model includes two popular survival models, Cox model and accelerated failure time model as its special cases. Under the nested structure of this general model, we may extend the joint modeling approach to facilitate model checking and model selection. An example of Taiwanese AIDS data is used to demonstrate the joint model approach.