Multistage Random Effects Model for Disease Progression in Longitudinal Aging Studies

• Activity Number: 243
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Kumar Rajan*+ and Denis Evans
• Companies: and Rush University Medical Center
• Keywords: Random-effects models ; numerical integration ; disease states ; maximum likelihood integration

Abstract

Random-effects models have been commonly used to study degenerative disease processes in older adults. However, understanding disease mechanism involves understanding the transitions among several disease states. In this presentation, we will focus on three disease states within the disability framework, namely, onset of disability, progression after onset, recovery after onset or progression. We propose to use a logistic regression model with random effects for onset and recovery, and a count regression model with random effects for progression after onset. The random effects from onset, recovery and progression states will be correlated and assumed to be from a jointly normal distribution, since a subject can go through all three phases of disease states in a longitudinal study of 10 years with annual follow-up. The marginal likelihood after integrating the random effects does not have a closed form; therefore we will Gaussian-Hermit quadrature as a numerical approximation. The results showed that risk factors were different for onset, progression and recovery states and the covariance of the random effects was fairly strong between the three processes.