Clinical Manifestations.

Abstract

The California Cognitive Assessment Battery (CCAB; ccabstudy.com) is a remotely monitored, at-home, computerized behavioral test and questionnaire suite. Here we evaluate the interpretation of 25 tests within the US English version of the CCAB using both an exploratory factor analysis (EFA) of 155 healthy adults and a confirmatory factor analysis (CFA) of 312 healthy older adults participating in a longitudinal study. We also estimated cross-sectional age gradients for best fitting latent factors [1]. First, one set of "classic" latent factors (from 1 factor to 8 factors) was extracted from the literature [2][3][4] for assignment of CCAB tests into traditional categories. Second, 155 adult participants (age 22-86 y.o., mn 54, 57% male, education 8-20 yrs., md 14) participated in a prefinal version of the CCAB. EFA was performed on their primary CCAB results using the 'psych' (v. 1.9.12.31-1) toolbox in R (v. 3.6.3) with 11 different rotations to extract a second set of CFA latent factors. Third, above two sets of latent factors were used in the 'lavaan' CFA toolbox (v. 0.6.5-1) in R and applied to the first timepoint (enrollment and session 1) of CCAB longitudinal data from an older adult group (age 56-89 y.o., mn 71, 58% male, education 0-20 yrs., mn 14). The EFA and CFA model fit characteristics (Fig 1) show moderately good fits for the EFA and best fitting CFA models. The EFA produced reasonable models from 2 factors to 6 factors. The best fitting "classic" CFA along with its latent factor age regressions is shown in Figure 2. The best fitting EFA-derived CFA along with its latent factor age regressions is shown in Figure 3. The best-fit classic CFA results showed moderately good fit, and visual tests of both speed and memory showed steeper cross-sectional age gradients than did respective verbal response tests. The EFA-derived latent factors were better fitting than were the classic factors as they were derived from the CCAB itself. In both cases, however, CCAB test speed measures showed steeper cross-sectional age gradients than did memory measures. Ongoing work will characterize longitudinal latent factor age gradients.