Are there measurement invariance in subjective cognitive decline instruments across diverse ethnoracial groups?

Abstract

AbstractBackgroundSubjective cognitive decline (SCD) indicates a likelihood of the presence of cerebral beta‐amyloid burden associated with preclinical Alzheimer’s disease (AD), future objective cognitive decline, and clinical progression to dementia. The applicability of SCD as a preclinical AD marker in diverse ethnoracial groups has not been validated. In the few studies with diverse samples, participants from under‐represented populations such as Latinx and Black, tend to endorse differing levels of SCD than Non‐Hispanic White (NHW) individuals. Better understanding of ethnoracial differences in SCD and addressing potential cultural biases in existing measures are critical to validate its use as a preclinical AD marker in diverse samples. In this study, we examined two SCD instruments to evaluate measurement invariance across ethnoracial groups.Method807 participants aged ≥55 who reside in the United States (298 Latinx, 270 Black, 239 NHW) were recruited through Amazon’s Mechanical Turk and Qualtrics Panel. We examined the psychometric properties of the Cognitive Change Index (CCI) and Cognitive Function Index (CFI) across ethnoracial groups. First, we evaluated the factor structure of each measure using split‐half exploratory and confirmatory factor analyses (EFA/CFA). Next, we used multiple‐group CFA to evaluate configural, metric, and scalar invariance via a step‐wise approach.ResultsParticipants (mean age of 65), were mostly women (63%) and received education beyond high school (83%). Split‐half EFA/CFA on the whole sample supported a one‐factor solution for both measures. Multiple‐group CFA revealed that when race/ethnicity was considered, model fit was negatively impacted. Neither measure achieved metric invariance, indicating that factor loadings were not equivalent across groups. For example, the CFI item “memory has declined substantially” had the highest loading for NHW and Latinx, but was low for the Black group. Conversely, “trouble following the news” loaded high for Latinx and Black, but low for NHW.ConclusionMeasurement invariance was not supported for either instrument, indicating that it cannot be assumed that the same construct is being measured across NHW, Latinx and Black individuals. Scale refinement, such as abbreviating SCD measures to include only items that perform well across all groups, may minimize biases and facilitate cross‐cultural comparison in preclinical AD research.