Premium
[O2–12–02]: TRANSITIONS ACROSS COGNITIVE STATES AND MORTALITY AMONG OLDER ADULTS: A MULTI‐STATE SURVIVAL MODEL
Author(s) -
Robitaille Annie,
Hout Ardo,
Machado Robson Mariano,
Čukić Iva,
Deary Ian J.,
Hofer Scott M.,
Hoogendijk Emiel,
Johansson Boo,
Koval Andriy V.,
Noordt Maaike,
Piccinin Andrea M.,
Rijnhart Judith J.M.,
Skoog Johan,
Skoog Ingmar,
Vermunt Lisa,
Terrera Graciela Muniz
Publication year - 2017
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2017.07.211
Subject(s) - socioeconomic status , cognition , dementia , psychology , gerontology , educational attainment , cognitive decline , longitudinal study , cognitive impairment , cognitive skill , developmental psychology , demography , medicine , psychiatry , population , disease , sociology , pathology , economics , economic growth
that assume the trajectory form is the same for everyone in the sample. Including demographic variables as covariates in such studies only allows for the identification of mean differences. In contrast, we used structural equation model (SEM) trees to identify heterogenous subgroups that demonstrate differential trajectories of change in cognition. SEM Trees partitions the dataset into subsets (groups of participants) based on the splitting of covariates, with a latent change growth model fit to each resultant subset. Methods:9 waves from the representative samples of the Health and Retirement Study and Assets and Health Dynamics of the Oldest Old (N1⁄4 20,685), were analyzed. Covariates were years of education, race (white, black, other), Hispanic, and gender. Each identified subset was fit with SEM models on mental status scores. Results: Different trajectories were observed with covariate splits on race, Hispanic ethnicity, and education combinations. The figure shows that formental status,Whitewith 14 years of education,White, 12-13 years, and Nonwhite, 16 years had highest intercept scores, with age declines after the mid-70’s. Nonwhite andHispanics with 12-15 years andWhite, 9-11 years formed themiddle groups for intercepts, showing declines in the late 60’s. Nonwhites 11 years, and whites 6 years had the lowest intercepts and declines just after age 60. Conclusions:SEM trees indicated different patterns of cognitive aging in a population sample. These findings suggest that cognitive disparities are rooted not only by race or ethnicity but education, and that whites are more protected from cognitive decline than nonwhites even with only a few years of high school. We conclude with a discussion of the importance of examining not just mean effects, but also allowing the change parameters to vary, and by allowing interactions between covariates.