Effects of Selection Bias and Competing Risks on Factors Determining Conversions to Cognitive Impairments and Mixed Dementias

• Activity Number: 158
• Type: Topic Contributed
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Richard J. Kryscio*+ and Yushun Lin and Liou Xu and Erin Abner
• Companies: University of Kentucky and University of Kentucky and University of Kentucky and University of Kentucky
• Keywords: transition models ; Markov ; joint models ; competing risks ; selection bias ; mixed dementias

Abstract

We present an overview of research involving data from serial cognitive assessments of elderly individuals followed to dementia or death. Assessments are categorized into four states: intact cognition, mild cognitive impairment, global impairments, and clinical dementia, all of which are interval censored events. Clinical dementia comprises heterogeneous pathologies; underlying disease(s) may not be clarified until autopsy. The analytic goal is to define relationships among risk factors, transitions between cognitive states, and probability of dementia before death. Statistical issues include selection bias, competing risk of death, and the effect of mixed dementias. Mixed dementia analysis will focus on hippocampal sclerosis, which occurs frequently in the very old. Results based on transition models using either a nonstationary Markov chain with absorbing states including interval censored deaths or joint models for the actual residual lifetime and the cognitive states will be summarized. Applications to two mature cohorts will illustrate the results: Biologically Resilient Adults in Neurological Studies (BRAINS), an observational cohort, and the Nun Study, a population cohort.