Abstract:
|
Recurrent events could be stopped by a terminal event, which commonly occurs in biomedical and clinical studies. In this situation, the non-informative censoring assumption could be violated. The joint frailty model is widely used to jointly model these two processes. One assumption of this model is that recurrent events and terminal event processes are conditionally independent given the subject-level frailty. Furthermore, the correlation between the terminal event and the recurrent events is a constant over time. We propose a time-varying joint frailty-copula model to relax these two assumptions in the Bayesian framework. Conditional on the frailty, the survival functions are joint modeled by a survival copula. Also, the dynamic correlation between the terminal event and the recurrent event process is modeled by a latent Gaussian AR(1) process. The simulation results show that compared with the joint frailty model and the joint frailty-copula model, the absolute bias and mean squared error of the time-varying frailty-copula model is the smallest. Then, we applied our method to analyze the MarketScan data to identify potential risk to recurrent stroke and death.
|