Markov Chains and Continuous Time Multi-State Markov Models Comparisons in Longitudinal Clinical Analysis

• Activity Number: 270
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 2:45 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Lijie Wan* and Richard J. Kryscio
• Companies: University of Kentucky and University of Kentucky
• Keywords: Markov Chains ; Continuous-time multi-state Markov models ; interval censoring ; competing risk

Abstract

Multi-state Markov models are widely used to analyze longitudinal data describing the progression of a chronic disease or condition, like dementia. Several studies have focused on modeling true disease progression as a discrete time Markov chain, which requires certain assumptions. Recently, continuous-time multi-state Markov models have also become very popular. In this paper, we discuss the relationship as well as differences between these two modeling techniques. Our simulation study shows that when longitudinal data are arise from equally spaced intervals and with no unobserved transition between two contiguous assessment time points, these two types of models work equally well. When the data are not equally spaced, or there are possible unobserved transitions between two contiguous assessment time points (e.g., a patient dies without being observed first passing through severe clinical disease), the continuous-time multi-state Markov model is preferred. We also apply our model to a real dataset, the Nun Study, a cohort of 461 participants who were cognitively normal at study baseline and followed to autopsy.