Diagnostic utility of culture/language‐specific normative data on the Version 3 of the Uniform Data Set Neuropsychological Battery for differentiating stages of cognitive impairment among Latinos in the United States

Background Accurate detection of neurocognitive disorders of aging among Latinos in the United States (U.S.) is an important step towards addressing disparities in Alzheimer’s disease and related dementias (ADRDs) and promoting health equity in this group. Federally‐funded Alzheimer’s Disease Research Centers (ADRCs) in the U.S. have utilized the National Alzheimer’s Coordinating Center (NACC) Uniform Data Set Version 3 Neurobehavioral Battery (UDS3‐NB) since 2015. Normative data on this battery adjusted for age, education and sex were published for English‐speakers in 2018 (Weintraub et al., 2018) and expanded to consider differences among non‐Hispanics White and Black people in 2020 (Sachs et al., 2020). We aimed to determine whether the use of newly developed culture/language‐specific neuropsychological normative data for Spanish‐ and English‐speaking Latinos in the U.S. improved diagnostic accuracy of the UDS3‐NB in this group. Method Participants included 1,018 Latino adults age 50+ enrolled in ADRCs. Table 1 presents participant demographics by global Clinical Dementia Rating (CDR) score, as indicative of stage of neurocognitive impairment (0 = normal, 0.5 = questionable impairment, 1 = mild impairment, 2‐3 = moderate‐to‐severe impairment). Individual test raw scores were converted to demographically‐adjusted z scores utilizing normative data on the UDS3‐NB developed for Latinos and for other culture/language groups. Individual tests z scores were converted to deficit scores and averaged to compute global deficit scores and determine global neurocognitive impairment (NCI). Rates of NCI based on culture/language specific normative data were compared across the four CDR groups via Chi‐Square tests. Sensitivity and specificity were computed for differentiating participants with CDR = 0 (normal) and CDR = 1 (mild impairment) based on different sets of normative data. Results Utilizing culture/language‐specific normative data significantly differentiated CDR groups (p