Linear and Nonlinear Effect of Years of Schooling, Sex, and Age on the CERAD-MX and Complementary Tasks in a Mexican Sample: A Cross-Sectional Study

Abstract

Objective Since evidence of adults’ cognition decline is based on standardized testing, we developed regression-based continuous norms by linear regression (LR) and nonlinear quantile regression (NQR) with years of schooling (YoS), age, and sex as covariates on the Mexican adaptation of the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD-MX) and complementary tasks. Methods 392 healthy, Spanish-speaking Mexican adults (50.25% women) aged 18–59 completed the 15 CERAD-MX cognitive tasks and complementary tasks. We used raw scores and examined YoS-related effects considering sex and age as covariates. For the NQR, we used calibrated scores for sex and age. While LR represents one line across the performance, NQR differentiated several nonlinear performance bands by quantiles. Results LR showed positive relationships between YoS and cognitive performance with a funnel variance pattern. Therefore, this relationship is better represented with NQR than LR. A small, but significant, negative effect of age was found for this age range (18–59 years). The band with fewer years of schooling (1–6) showed greater variability in the cognitive measures than those with more years of schooling (16–22). Conclusion This study shows that NQR is useful for accurately positioning participants’ performance relative to their peers. NQR accounts more than LR for the inconsistent variability of cognitive performance as a function of YoS by identifying the variability according to YoS (low, medium, high). Thus, NQR represents an appropriate way to construct norms for the cognitive performance of adults.

Longitudinal normative standards for cognitive tests and composites using harmonized data from two Wisconsin AD‐risk‐enriched cohorts

“…More importantly, the restricted regression quantile approach used with this dataset guarantees that model fits are free of crossed quantile estimates. Although interest in conditional norms models is growing, and longitudinal norms for AVLT have recently been published alongside a ShinyApp, 59 use of quantile regression in this setting remains relatively uncommon 60,61 …”

“…Although interest in conditional norms models is growing, and longitudinal norms for AVLT have recently been published alongside a ShinyApp, 59 use of quantile regression in this setting remains relatively uncommon. 60 , 61 …”