Demographics and brain models variably predict cognition in an ethnoracially diverse sample by cohort

Abstract Background Considerable heterogeneity in cognitive trajectories has been well documented in studies of older individuals from diverse samples, including minority communities. Such studies often have not included brain factors in their models. Modeling demographic factors and neuroimaging brain data as predictors of cognition in a diverse sample set is crucial to further delineate differences in associations to cognitive outcomes by ethnoracial cohort. Method Four ethnoracial cohorts of older participants with variable baseline cognitive statuses were sampled from the UC Davis Aging and Diversity Cohort (Table 1). We modeled cognitive baseline and change outcomes for episodic verbal memory and executive function, using predictors consisting of demographic and brain variables in an ethnoracially diverse sample. Brain measures were brain gray and white matter signatures of cognition (Fig. 1a, b) and white matter hyperintensity burden (WMH) at baseline. For longitudinal, we retained the WM measures but used signature atrophy regions of cognitive change. We hypothesized that distinct profiles of predictor variables might emerge by ethnoracial group. Result Table 2 presents an overview of our model fit by R 2 metric across all four ethnoracial groups. Baseline fits were higher for non‐white ethnoracial groups than for Whites. Longitudinal fits were similarly high for all cohorts having sufficient data. Table 3a, b reveals varying explanatory profiles for baseline predictors. For non‐brain variables, education was most consistently significant, with others also significant in several models. Brain GM was consistently significant except in Asians; meanwhile WM signatures were significant only in Asians. In longitudinal models (Table 4a, b) brain tissue atrophy was most consistently significant, along with WM measures among Whites. No non‐brain variables were significant. Conclusion Our findings support baseline ethnoracial variability expressed in our initial hypothesis, while also suggesting that brain GM and tissue atrophy measures may be strong predictors regardless of cohort. Education is frequently significant across cohorts at baseline, consistent with existing research. GM baseline and tissue atrophy are consistent predictors except in Asians, contrasting our hypothesis. WM factors vary by cohortgrouping, exhibiting significance in Asians and Whites, suggesting ethnoracial profile differences in support of our hypothesis and requiring further research.