Abstract
Objective: There are diagnostic disparities in dementia across ethno-racial communities. The potential contribution of measurement bias in cognitive assessments was investigated by examining the factor structure and measurement invariance of the Uniform Data Set neuropsychological battery (UDS3-NB) across ethno-racial groups as well as Spanish and English speakers along a cognitive continuum. Methods: Data were from the National Alzheimer's Coordinating Center (NACC) and the 1Florida Alzheimer's Disease Research Center (1FLADRC). Confirmatory factor analyses were conducted in NACC (n = 29,462; M age = 71.0, SD = 10.4; 58.5% female, M education = 15.8, SD = 3.0; 68.6% non-Hispanic White, 5.3% -primary Spanish speaking) to determine the UDS3-NB factor structure. Measurement invariance was tested in NACC and separately within 1FLADRC (n = 829; M age = 69.9, SD = 8.5; 56.5% female; M education = 14.6, SD = 3.5 years; 33.3% non-Hispanic White, 33.8% primary Spanish speaking), across ethno-racial (non-Hispanic White, Hispanic White, Black/African American), primary language (English, Spanish), and cognitive status (cognitively normal, mild cognitive impairment, dementia). Invariance was assessed at configural, metric, scalar, and strict levels. Results: A 4-factor model (memory, processing speed/executive functioning, language, attention) demonstrated acceptable to good fit in NACC (CFI = 0.97, TLI = 0.96, RMSEA = 0.07, SRMR = 0.03) and 1FLADRC (CFI = 0.97, TLI = 0.96, RMSEA = 0.05, SRMR = 0.04). Standardized factor loadings ranged from 0.43-0.87. Metric invariance was supported across all contrasts in both cohorts. Higher levels of invariance were cohort-dependent: CN vs. MCI achieved strict invariance in both cohorts, whereas non-Hispanic White vs. Hispanic White and English vs. Spanish only achieved strict invariance in NACC. Conclusions: The UDS3-NB demonstrates structural validity and metric equivalence across ethno-racial, linguistic, and cognitive status groups.