Latent Factor Structure and Measurement Invariance of the NIH Toolbox Cognition Battery in an Alzheimer’s Disease Research Sample

Abstract

Objective: This study investigated the latent factor structure of the NIH Toolbox Cognition Battery (NIHTB-CB) and its measurement invariance across clinical diagnosis and key demographic variables including sex, race/ethnicity, age, and education for a typical Alzheimer’s disease (AD) research sample. Method: The NIHTB-CB iPad English version, consisting of 7 tests, was administered to 411 participants aged 45–94 with clinical diagnosis of cognitively unimpaired, dementia, mild cognitive impairment (MCI), or impaired not MCI. The factor structure of the whole sample was first examined with exploratory factor analysis (EFA) and further refined using confirmatory factor analysis (CFA). Two groups were classified for each variable (diagnosis or demographic factors). The confirmed factor model was next tested for each group with CFA. If the factor structure was the same between the groups, measurement invariance was then tested using a hierarchical series of nested two-group CFA models. Results: A two-factor model capturing fluid cognition (executive function, processing speed, and memory) versus crystalized cognition (language) fit well for the whole sample and each group except for those with age < 65. This model generally had measurement invariance across sex, race/ethnicity, and education, and partial invariance across diagnosis. For individuals with age < 65, the language factor remained intact while the fluid cognition was separated into two factors: (1) executive function/processing speed and (2) memory. Conclusions: The findings mostly supported the utility of the battery in AD research, yet revealed challenges in measuring memory for AD participants and longitudinal change in fluid cognition.

Harmonization of the English and Spanish versions of the NIH Toolbox Cognition Battery crystallized and fluid composite scores.

“…There were relatively lower loadings on the fluid cognition factor by the episodic memory test (i.e., Picture Sequence Memory; λ range: .35 to .47) and the working memory test (i.e., List Sorting; λ range: .40 to .51) compared to tests of executive functions and processing speed (i.e., Dimensional Change Card Sort, Flanker, and Pattern Comparison: λ range: .59 to .80). In prior factor analyses, the episodic and working memory tests loaded on a third memory factor among participants 45-64 years old (Ma et al, 2021) and more specific construct factors when paired with traditional neuropsychological tests of the same constructs (i.e., episodic memory and working memory), which led to greater model fit (Mungas et al, 2014). In future versions of the NIHTB-CB, the test developers may consider basing composite scores on latent constructs that may better reflect the constructs measured by the individual tests, but such an approach would likely require a longer battery.…”

“…The norming of the NIHTB-CB followed a stratified sampling strategy with predetermined target sample sizes for specific age ranges among English- and Spanish-speaking adults recruited across the United States, with efforts to recruit a sample of balanced gender representation and a distribution of race, ethnicity, and education consistent with the national population (Beaumont et al, 2013). Prior research has established measurement invariance for the English-language NIHTB-CB across age, gender, race/ethnicity, and diagnosis of cognitive impairment (Ma et al, 2021; Mungas et al, 2014); however, no studies have evaluated measurement invariance based on language of administration. An examination of measurement invariance between the English- and Spanish-language NIHTB-CB could clarify whether scores from these test batteries could be combined and meaningfully compared.…”

Racial Differences in Associations of Cognitive Health Status With Happiness, Helplessness, and Hopelessness Among Older Adults: An Exploratory Study

“…Wisconsin ADRC longitudinally follows participants; However, for the current cross-sectional study, only participants’ baseline data were used. Participants were recruited from community and memory clinics ( Ma et al, 2021 ). After provided informed consent, participants completed clinical and cognitive assessments.…”

“…Total remaining sample included 755 participants. ADRC’ detailed clinical diagnosis procedure is described elsewhere ( Blazel et al, 2020 ; Ma et al, 2021 ).…”

Neuropsychological Tests in Post-operative Cognitive Dysfunction: Methods and Applications

“…The battery consists of tests of seven constructs, including five fluid cognition (executive function, attention, episodic memory, working memory, and processing speed) and two crystallized cognition (vocabulary and word reading). This test group can maintain its stability among people of different age groups, races, genders, and education levels (Ma et al, 2020). It also has good reliability and validity in the evaluation of the healthy elderly (Scott et al, 2019).…”