INS NYC 2024 Program

Poster	Poster Session 11 Program Schedule 02/17/2024 10:45 am - 12:00 pm Room: Shubert Complex (Posters 1-60) Poster Session 11: Cultural Neuropsychology \| Education/Training \| Professional Practice Issues Final Abstract #60 Examination of Measurement Invariance Across Race and Ethnicity in the Health and Retirement Study's Harmonized Cognitive Assessment Protocol Henri Lu, Simon Fraser University, Burnaby, Canada Rachel Fouladi, Simon Fraser University, Burnaby, Canada Category: Assessment/Psychometrics/Methods (Adult) Keyword 1: psychometrics Keyword 2: ethnicity Keyword 3: aging (normal) Objective: Examination of racial and ethnic disparities in dementia in older adults may include comparisons of composite cognitive domain scores. Evaluating measurement invariance is critical for determining whether latent cognitive factor scores can be meaningfully compared across racial and ethnic groups. Recently, Jones et al. (2023) proposed various factor structures for the neuropsychological battery in the Harmonized Cognitive Assessment Protocol (HCAP) sub-study of the Health and Retirement Study (HRS). It is unclear whether these models differ across race and ethnicity. Thus, the current study investigated measurement invariance of the cognitive domain scores across race and ethnicity. Participants and Methods: Data were obtained from the 2016 HRS-HCAP. Participants included 3,273 older adults (2,383 non-Hispanic White, 327 non-Hispanic Black, 363 Hispanic; M_age = 76.19 years, SD = 7.50; 39.8% male, 60.2% female) who had at least one non-missing score on the neuropsychological battery. Measurement invariance analyses across the three racial and ethnic groups (non-Hispanic White, non-Hispanic Black, and Hispanic) were conducted using the R package lavaan. The WLSMV estimator was used to accommodate categorical and non-normally distributed data. Configural (structure) invariance models were assessed with robust CFI and RMSEA, and subsequent nested models for metric (loadings), scalar (intercepts), and strict (residuals) invariance were assessed with RMSEA_D (Savalei et al., 2023). Results: Separate single factor models for immediate memory, delayed memory, attention and processing speed, and language were assessed. For immediate memory, the configural invariance model showed adequate fit (CFI = 0.978, RMSEA [90% CI] = 0.076 [0.055-0.098]). Further constraints showed good fit for metric invariance (RMSEA_D [90% CI] = 0.013 [0-0.043]), scalar invariance (RMSEA_D [90% CI] = 0.043 [0.017-0.064]), and strict invariance (RMSEA_D [90% CI] = 0.021 [0-0.048]). For delayed memory, the configural invariance model showed adequate fit (CFI = 0.976, RMSEA [90% CI] = 0.063 [0.050-0.078]). Further constraints showed good fit for metric invariance (RMSEA_D [90% CI] = 0.005 [0-0.037]), scalar invariance (RMSEA_D [90% CI] = 0.036 [0.017-0.055]), and strict invariance (RMSEA_D [90% CI] = 0.024 [0-0.047]). For language, a modified configural invariance model showed good fit (CFI = 0.986, RMSEA [90% CI] = 0.030 [0.013-0.046]). Further constraints showed good fit for metric invariance (RMSEA_D [90% CI] = 0.007 [0-0.037]), scalar invariance (RMSEA_D [90% CI] = 0.017 [0-0.046]), and strict invariance (RMSEA_D [90% CI] = 0.029 [0-0.074]). For attention and processing speed, fit of the configural invariance model was mixed, with CFI demonstrating good fit and RMSEA demonstrating poor fit (CFI = 0.987, RMSEA [90% CI] = 0.084 [0.071-0.099]). Conclusions: Results supported strict invariance across race and ethnicity (non-Hispanic White, non-Hispanic Black, and Hispanic) for immediate memory, delayed memory, and language. However, evidence of configural invariance across race and ethnicity was mixed for attention and processing speed, suggesting the need to explore other factor structures. There is evidence to support the use of latent score means for immediate memory, delayed memory, and language across racial and ethnic groups, which allows further research of racial and ethnic differences and disparities in cognitive functioning in older adults in the HRS-HCAP.

Poster

Poster Session 11 Program Schedule

02/17/2024
10:45 am - 12:00 pm
Room: Shubert Complex (Posters 1-60)

Poster Session 11: Cultural Neuropsychology | Education/Training | Professional Practice Issues

Final Abstract #60

Examination of Measurement Invariance Across Race and Ethnicity in the Health and Retirement Study's Harmonized Cognitive Assessment Protocol

Henri Lu, Simon Fraser University, Burnaby, Canada
Rachel Fouladi, Simon Fraser University, Burnaby, Canada

Category: Assessment/Psychometrics/Methods (Adult)

Keyword 1: psychometrics
Keyword 2: ethnicity
Keyword 3: aging (normal)

Objective:

Examination of racial and ethnic disparities in dementia in older adults may include comparisons of composite cognitive domain scores. Evaluating measurement invariance is critical for determining whether latent cognitive factor scores can be meaningfully compared across racial and ethnic groups. Recently, Jones et al. (2023) proposed various factor structures for the neuropsychological battery in the Harmonized Cognitive Assessment Protocol (HCAP) sub-study of the Health and Retirement Study (HRS). It is unclear whether these models differ across race and ethnicity. Thus, the current study investigated measurement invariance of the cognitive domain scores across race and ethnicity.

Participants and Methods:

Data were obtained from the 2016 HRS-HCAP. Participants included 3,273 older adults (2,383 non-Hispanic White, 327 non-Hispanic Black, 363 Hispanic; M_age = 76.19 years, SD = 7.50; 39.8% male, 60.2% female) who had at least one non-missing score on the neuropsychological battery. Measurement invariance analyses across the three racial and ethnic groups (non-Hispanic White, non-Hispanic Black, and Hispanic) were conducted using the R package lavaan. The WLSMV estimator was used to accommodate categorical and non-normally distributed data. Configural (structure) invariance models were assessed with robust CFI and RMSEA, and subsequent nested models for metric (loadings), scalar (intercepts), and strict (residuals) invariance were assessed with RMSEA_D (Savalei et al., 2023).

Results:

Separate single factor models for immediate memory, delayed memory, attention and processing speed, and language were assessed. For immediate memory, the configural invariance model showed adequate fit (CFI = 0.978, RMSEA [90% CI] = 0.076 [0.055-0.098]). Further constraints showed good fit for metric invariance (RMSEA_D [90% CI] = 0.013 [0-0.043]), scalar invariance (RMSEA_D [90% CI] = 0.043 [0.017-0.064]), and strict invariance (RMSEA_D [90% CI] = 0.021 [0-0.048]). For delayed memory, the configural invariance model showed adequate fit (CFI = 0.976, RMSEA [90% CI] = 0.063 [0.050-0.078]). Further constraints showed good fit for metric invariance (RMSEA_D [90% CI] = 0.005 [0-0.037]), scalar invariance (RMSEA_D [90% CI] = 0.036 [0.017-0.055]), and strict invariance (RMSEA_D [90% CI] = 0.024 [0-0.047]). For language, a modified configural invariance model showed good fit (CFI = 0.986, RMSEA [90% CI] = 0.030 [0.013-0.046]). Further constraints showed good fit for metric invariance (RMSEA_D [90% CI] = 0.007 [0-0.037]), scalar invariance (RMSEA_D [90% CI] = 0.017 [0-0.046]), and strict invariance (RMSEA_D [90% CI] = 0.029 [0-0.074]). For attention and processing speed, fit of the configural invariance model was mixed, with CFI demonstrating good fit and RMSEA demonstrating poor fit (CFI = 0.987, RMSEA [90% CI] = 0.084 [0.071-0.099]).

Conclusions:

Results supported strict invariance across race and ethnicity (non-Hispanic White, non-Hispanic Black, and Hispanic) for immediate memory, delayed memory, and language. However, evidence of configural invariance across race and ethnicity was mixed for attention and processing speed, suggesting the need to explore other factor structures. There is evidence to support the use of latent score means for immediate memory, delayed memory, and language across racial and ethnic groups, which allows further research of racial and ethnic differences and disparities in cognitive functioning in older adults in the HRS-HCAP.