Intraindividual variability on the NIH Toolbox Cognition Battery: Sociodemographic comparisons and test-retest reliability of dispersion in cognitive test scores.

Abstract

Objective: This study examined intraindividual variability for fluid cognition test scores on the NIH Toolbox Cognition Battery (NIHTB-CB) by (a) comparing dispersion across sociodemographic characteristics and (b) estimating test-retest reliability of dispersion scores. Method: Participants completed the NIHTB-CB as part of the norming study (n = 730; M = 47.4 ± 17.6 years-old; 64.4% women; 63.2% White), with a subsample completing retest within 1-2 wk (n = 138; M = 54.8 ± 20.0 years-old; 48.6% women; 68.1% White). Raw NIHTB-CB scores were converted to age-adjusted standard scores and demographic-adjusted T-scores. IIV was estimated via the intraindividual standard deviation (ISD) and intraindividual coefficient of variation (ICV) for the five fluid cognition tests. Participants were compared on the ISD and ICV based on age, gender, race/ethnicity, education, and crystallized composite score. Test-retest reliability was estimated using the intraclass correlation coefficient (ICC) with a 95% confidence interval. Results: Group comparisons indicated significant ICV differences based on race/ethnicity and crystallized composite for standard scores and significant ISD and ICV differences based on gender and crystallized composite for T-scores, but notably, all effect sizes were small in magnitude. The test-retest reliability of the ISD was poor for standard scores (ICC = 0.33 [0.01, 0.55]) and T-scores (ICC = 0.49 [0.23, 0.66]). The test-retest reliability of the ICV was poor for standard scores (ICC = 0.43 [0.16, 0.62]) and moderate for T-scores (ICC = 0.63 [0.43, 0.75]). Conclusions: The ISD and ICV did not substantially differ based on sociodemographic characteristics and both scores lacked stability over a brief test-retest interval, indicating a need for additional psychometric research on dispersion scores to inform research and practice.