Semi-Markov Models for Transitions from Cognitively Intact to Clinically Impaired

• Activity Number: 526
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Richard Kryscio*+ and Yushun Lin and Erin Abner
• Companies: University of Kentucky and Citi Bank and Center on Aging
• Keywords: semi-Markov ; dementia ; panel data ; clinical mild cognitive impairment ; transitions

Abstract

In dementia research, cognitively intact (CI) elderly subjects enroll in cohorts where each participant volunteers for annual cognitive assessments. While all subjects are CI at baseline, future assessments may result in a diagnosis of clinical Mild Cognitive Impairment (MCI) or dementia. We consider modeling the flow of subjects through CI, clinical MCI, dementia, and the competing state of death, and investigate risk factors for these transitions using semi-Markov models. In our application to 543 subjects enrolled in the Biologically Resilient Adults in Neurological Studies (BRAiNS) cohort at the University of Kentucky's Center on Aging, we found considerable competition for transitions to clinical MCI and dementia due to death. Results show that the holding time distributions for the states CI and clinical MCI are well-approximated by Weibull distributions with parameters depending on the final state, while transition probabilities among the states expressed as multinomial logistic models are functions of the risk factors under study.