Attrition-Considered Statistical Characterization of 15-Year Longitudinal Cognitive Change

• Activity Number: 190
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Maria Josefsson*+
• Companies: Umeå University
• Keywords: cognitive decline ; attrition ; pattern mixture model

Abstract

Episodic memory performance declines with the passage of time. Little is known about inter-individual differences in rate of change. Many longitudinal studies suffer from attrition as the mechanism causing the missing data often is non-ignorable. Another issue to consider when analyzing data from longitudinal studies including scores from tests, is the upper and lower boundaries of the test scores which limit the rate of change to move freely. Statistical methods not taking these issues into account can cause biased estimates. In order to handle the influence of missing data, we propose a random-effect pattern mixture model which also takes the ceiling/floor effects from the test scores into consideration. The model has been applied on an ongoing longitudinal study of episodic memory with 1444 participants identifying changes in episodic memory performance over time. Based on the rate of change and initial test score and factorized into different age groups, the episodic memory curve of the participants were categorized as successful, normal, or declining. In a second step, the "successful" individuals were characterized with regard to demographic, genetic, and lifestyle factors.