In dementia, remote, internet-based cognitive assessment allows increased access for individuals from diverse backgrounds and geographic regions, repeated assessment, and cost efficiency. However, normative data for ethnoculturally-diverse populations in this context are limited. The Brain Health Registry (BHR) is a large-scale, longitudinal, internet-based study in which the Cogstate Brief Battery (CBB) is used to measure cognition, and which includes data from ethnoculturally-diverse adults. The development of robust normative data leverages longitudinal data to exclude participants who eventually develop cognitive impairment, despite being unimpaired at their baseline timepoint. This produces higher, less variable scores compared to conventional normative data, thereby increasing their sensitivity to cognitive dysfunction. Existing CBB normative data are not robust, only age adjusted, and based entirely on supervised, in-clinic assessment. The aims of this study were to examine demographic effects on performance data derived from unsupervised CBB performance and to subsequently develop robust, demographically-adjusted CBB norms within four separate ethnocultural groups from the BHR study.

Latinx (any race; n=641), non-Latinx Asian (n=514), non-Latinx Black/African American (n=463), and non-Latinx white (n=13,302), cognitively-unimpaired BHR participants (N=14,920; M_age=56.6, SD=13.4, range=18–90+ years; M_education=16.4, SD=2.3, range=12–20 years) across the US met robust normative screening criteria. At baseline, participants completed the unsupervised, self-administered CBB, comprising four subtests: Detection (psychomotor speed), Identification (attention), One Card Learning (learning), and One Back (working memory). Robust, demographically-adjusted CBB T-scores were generated from multivariable fractional polynomial modeling of demographic predictors (age, gender, education) of standardized CBB subtest scores within each ethnocultural group.

ANCOVAs showed age, gender, education, and ethnocultural group were significantly independently associated with raw scores from CBB subtests (η_p² range=.001-.155, all ps<.001), except for gender’s effect on One Card Learning (η_p²<.001, p=.207). For each subtest, older age and fewer years of education were associated with worse performance. Female gender was associated with slower speeded outcomes and better working memory. Ethnocultural group effects were variable. Following development and application of robust, demographically-adjusted normative equations, demographics showed nonsignificant effects on all robust CBB T-scores (all ps>.05), demonstrating complete attenuation of confounding demographic effects on each CBB subtest.

It is imperative that users of demographically-adjusted norms apply them in an ethical, evidence-based manner appropriate to the purpose of the evaluation and context of the individual. Users are advised to carefully consider the sociocultural factors (e.g., educational quality, acculturation, language proficiency, social adversity), for which ethnocultural status serves as a proxy, that influence cognitive performance in our patients, participants, and the normative samples we compare them to. This study is the first to present robust, demographically-adjusted norms for four ethnocultural groups, increasing the research and clinical utility of a brief, remotely-administered, cognitive assessment tool for use in ethnoculturally and geographically diverse populations. However, the generalizability of these norms to the broader US population is limited given the high level of education obtained by the present sample. As such, users are advised to apply these norms cautiously, only when an individual’s demographics are represented by the sample, and appropriate to the evaluation context.