Copula Models for Multivariate Multistate Markov Processes Observed Subject to Right Censoring

• Activity Number: 253
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 2:00 PM to 3:50 PM
• Sponsor: SSC
• Author(s): Liqun Diao*+ and Richard Cook
• Companies: University of Waterloo and University of Waterloo
• Keywords: Copula model ; Markov process ; Multiplicative intensity ; Multistate model ; Right censoring

Abstract

A novel model is described for the joint analysis of multiple multistate chronic disease processes subject to right censoring. Rather than using the commonly adopted intensity-based or frailty-based frameworks, we used a formulation based on copula models. With the argument of density decomposition and conditional independence assumptions, an appealing joint model is obtained by assuming the joint distribution of absorption transition times is governed by a multivariate copula function. The copula formulation herein ensures that 1) a wide range of marginal processes can be specified, 2) the joint model will retain these marginal features to provide simple estimates and straightforward interpretation of transition rates and covariate effects for each component process, and 3) the scientific understanding regarding the relation between processes is facilitated. The copula formulation also allows a variety of approaches for estimation and inference including both parametric and semiparametric analysis. We focus on processes with Markov margins which are often suitable when disease processes are progressive in nature. A simulation study and an application provide further illustration.