Choice of Time Scale in Analyzing Longitudinal Data on Cognition and Its Effect on Inference

• 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): Lei Yu*+ and Eisuke Segawa and Sue Leurgans and David A. Bennett
• Companies: Rush Alzheimer's Disease Center and Rush Alzheimer's Disease Center and Rush Alzheimer's Disease Center and Rush Alzheimer's Disease Center
• Keywords: Longitudinal data ; Cognitive function ; Time scale

Abstract

In longitudinal studies of cognition, investigators examine the long term trajectories of cognitive functions and explore risk factors for change. Similar to time-to-event analysis, one practical issue of modeling continuous cognitive outcomes is the choice of an appropriate time scale. On one hand, time varying age is by far the most important risk factor and potential confounding variable in the study of aging, and it emerges as a natural conceptual time scale; on the other hand, time-on-study with adjustment for age at baseline allows the modeling of heterogeneity within the cohort, and therefore is often chosen when analyzing longitudinal cohort data. We aim to investigate whether these time scales present comparable characterizations of change in cognition. Using simulated data and examples from the Religious Order Study, an ongoing longitudinal study of aging and dementia, we will compare the statistical inference from the two approaches by assuming three different scenarios (1) linear trajectories; (2) linear trajectories with cohort effects; and (3) terminal decline. This study is supported by grants P30AG10161 and R01AG15819 from the National Institute on Aging.