Measurement Invariance of Neuropsychological Tests Across Different Sociodemographic Backgrounds in the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil)

Abstract

Objectives: Neuropsychological tests are often used in aging studies to compare distinct groups regarding diagnosis and prognosis. Because sociodemographic characteristics can influence cognitive scores, measurement of invariance is useful to assess the tests' ability to estimate the underlying constructs equally across groups and thus prevent biased interpretation. This study aims to verify measurement invariance in a large and socioculturally diverse sample. Method: Middle-aged and older adults (N = 13,395) from the ELSA-Brasil study were evaluated regarding health status and cognition. Mean age was 51.5 +/- 8.9 years. 54% were female, 52% were white, and 10% had lower education. Measurement invariance was analyzed using multiple-group confimiatory factor analysis for a two-factor model (memory and executive function). Configural, metric, and scalar invariance were tested across age, sex, race, and education groups. Results: The 2-factor model had a good lit to the entire dataset and for each group. Configural, metric. and scalar invariance confirmed that cognitive measures had the same structure and the items had similar relationships with the latent factors. Strict invariance was not achieved in most group comparison. Conclusions: The 2-factor model was appropriate for estimating cognitive performance in participants without measurement error by age, sex, race, and education. These results suggest that differences in cognitive outcomes are likely to indicate true differences.