Patterns of Cognitive Decline in Alzheimer's Dementia: An Application of Hierarchical Bayesian Multiphase Models

Abstract

As the Alzheimer's disease process progresses in time measurements of cognitive functioning exhibit nonlinearity. Multiphase models were used to quantify this nonlinearity for thirty-six well characterized individuals(~12 observations per individual over ~15 years in the study) by partitioning each into a healthy aging phase and a diseased phase. This enabled us to detail both the magnitude and timing that Alzheimer's disease alters different aspects of cognitive function. Estimation of these models was done using Bayesian methods. Eight different outcomes representing three areas of memory functioning(visual, verbal, working) were used to define a pattern of cognitive decline. The earliest phase change was found to be visual memory(~6 years before diagnosis) and was followed by changes in verbal and working memory beginning roughly four years later.