Affiliations: [a] Department of Family and Community Medicine, The University of Texas Health Science Center, San Antonio, TX, USA | [b] Department of Epidemiology and Biostatistics, The University of Texas Health Science Center, San Antonio, TX, USA | [c] Department of Psychiatry, The University of Texas Health Science Center, San Antonio, TX, USA | [d] Department of Medicine, The University of Texas Health Science Center, San Antonio, TX, USA | [e] South Texas Veterans’ Health System, Audie L. Murphy Division GRECC, San Antonio, TX, USA
Correspondence:
[*]
Correspondence to: Raymond F. Palmer, PhD, Department of Family and Community Medicine, The University of Texas Health Science Center at San Antonio, 7703 Floyd Curl Drive, San Antonio, 78284-7792 TX, USA. Tel.: +1 210 358 3200; Fax: +1 210 567 5507; E-mail: palmerr@uthscsa.edu
Abstract: Background:Structural equation models (SEM) can explicitly distinguish dementia-relevant variance in cognitive task performance. The resulting latent construct “δ” (for dementia) provides a relatively “error free” continuously varying dementia-specific phenotype. Objective:To estimate δ’s change over time (Δδ) and determine Δδ’s predictive validity using future dementia status as an outcome. Methods:Data from n = 2,191 participants of the Texas Alzheimer’s Research and Care Consortium (TARCC) were used to construct a latent growth curve model of longitudinal change over four years using five cognitive measures and one measure of Instrumental Activities of Daily Living. Four final latent factors, including baseline δ and Δδ, were simultaneously entered as predictors of wave 4 dementia severity, as estimated by the Clinical Dementia Rating Scale “sum of boxes” (CDR). Results:All observed measures exhibited significant change [χ2 = 1,152 (df = 229); CFI = 0.968; RMSEA = 0.043]. The final model demonstrated excellent fit to the data [χ2 = 543 (df = 245); CFI = 0.991; RMSEA = 0.023]. All latent indicator loadings were significant, yielding four distinct factors. After adjustment for demographic covariates and baseline CDR scores, d and Δd were significantly independently associated with CDR4, explaining 25% and 49% of its variance, respectively. The latent variable g’ significantly explained 3% of CDR4 variance independently of d and Δd. Δg’ was not significantly associated with CDR4. Baseline CDR explained 16% of CDR4 variance. Conclusions:Future dementia severity is almost entirely explained by the latent construct δ’s intercept and slope.
