A multilevel analysis of elders’ cognitive performance using data from the longitudinal study of generations

This study is a secondary data analysis focusing on describing elders’ intra-individual variability in their initial level of cognitive performance and their rate of cognitive change, as well as elders’ inter-individual variability in their initial level of cognitive performance and their rate of cognitive change in relation to the covariates of gender, college education, self-reported memory decline, and functional impairment. A multilevel model was fit into two waves of supplementary data in years 2000 and 2004 from the larger Longitudinal Studies of Generations (LSOG) using HLM software. Cases with missing data were deleted, leaving 180 (N) elders aged 55 and over for 2 sets of 2-level analysis with outcome variables of the clock drawing performance score, and the cube drawing performance score. All the covariates in the level-2 equations were dichotomized. In the analysis with the clock score as outcome variable, no significant effect was found for any of the covariates. In the analysis with the cube score as outcome variable, functional impairment was found to be significantly related to both elders’ cognitive function at age 70, and predicted elders’ rate of cognitive change in a 4-year period. Specifically, other things being equal, for people who had some functional impairment, their cube drawing performance was .034 logits lower than those who did not have any functional impairment at age 70. However, having at least some functional impairment was associated with a .004 logits increment to the growth rate of cube drawing performance. Possible reasons for both non-significant and significant effects’ of covariates on clock and cube drawing performance were discussed in light of findings from prior studies, characteristics of the present sample, and limitations of clock and cube drawing tasks as valid and reliable measures of cognitive performance. Directions for future research were suggested.