Dispersion on Spanish-language cognitive testing: Sociodemographic and sociocultural group comparisons of...

Objective: This study assessed whether sociodemographic/sociocultural variables were related to dispersion (i.e. intraindividual variability [IIV]) on Spanish-language cognitive testing. Method: Adults residing in the USA completed the Spanish-language NIH Toolbox Cognition Battery (n = 250; M = 38.8 ± 13.7 years-old; 72.0% women; 100% Hispanic), with dispersion calculated using the intraindividual coefficient of variation (ICV) and intraindividual standard deviation (ISD) for crystallized tests, fluid tests, and the total battery using age- and demographic-adjusted scores. Dispersion was compared based on age, gender, education (years), birthplace, education (country), bilingualism, income, and crystallized composite. Results: Fluid test dispersion did not differ based on sociodemographic/sociocultural variables. For the crystallized ICV, participants educated domestically had higher dispersion of age- and demographic-adjusted scores; and participants born domestically and with lower crystalized composites had higher dispersion of demographic-adjusted scores. There were no group differences for the crystallized ISD. For the total ICV and ISD, men had higher dispersion of age-adjusted scores, but not demographic-adjusted scores; and participants born domestically and with lower crystallized composites had higher dispersion of age- and demographic-adjusted scores. For the total ISD, but not the ICV, age-adjusted score dispersion was higher for participants educated domestically and with greater income; and demographic-adjusted score dispersion was higher for participants with greater income. Demographic-adjusted scores showed higher dispersion than age-adjusted scores, with large differences based on score calculation. Conclusions: Some sociodemographic/sociocultural variables were related to dispersion, corresponding with mostly small effect sizes. Demographic adjustments to test scores appeared to substantially increase within-person variability on cognitive testing, which may artificially increase dispersion estimates.