Reliability and stability of cognitive intraindividual variability indices among cognitively unimpaired older adults

Objective: Cognitive intraindividual variability (IIV) dispersion measures within-person variability in performance across a neuropsychological test battery and is prognostic of structural and functional decline. Although evidence for the construct validity of cognitive dispersion has grown considerably, its utility in applied settings relies on its stability and reliability. Thus, we examined mean-level stability, test-retest reliability, and reliable change of several cognitive dispersion indices among cognitively unimpaired older adults. Method: Participants were 2224 robustly cognitively unimpaired older adults from the National Alzheimer's Coordinating Center (NACC) Uniform Data Set 3.0. Intraindividual standard deviation (ISD) and coefficient of variation (CoV) dispersion indices were calculated for raw, scaled, and demographically-adjusted normed scores for the neuropsychological test battery at baseline and initial follow-up visit (M_days = 424.99). Analyses included paired samples t-tests and generation of effect sizes for mean-level stability, correlations for test-retest reliability, and practice-adjusted reliable change indices (RCIs) for classifying individuals. Results: The mean-level change effect sizes for ISD and CoV scores were very small (ds < .06); only CoV showed significant improvements. Test-retest reliability was poor (M_r = .41) for cognitive dispersion indices compared to global composite scores (r = .82). RCIs suggested that normed score changes greater than 1.75 standard deviations were significant. Conclusions: Findings showed stable mean-level cognitive dispersion across 424.99 days; however, consistent with process-based neuropsychological test scores literature, test-retest reliability was poor, perhaps reflecting random measurement error among cognitively unimpaired individuals. Change in cognitive dispersion may inform clinical prognosis and serve as supplementary information within broader neuropsychological evaluations given reliability limitations.