Differential Item Functioning of Informant Measures of Cognitive Functioning in the U.S. and Mexico

TL;DR

This study compares how informants in the U.S. and Mexico report cognitive impairment, finding differences that may affect dementia prevalence estimates.

Contribution

The study introduces a harmonized measure of informant-reported cognitive functioning across the U.S. and Mexico using structural equation modeling.

Findings

A harmonized CSID measure shows fixed factor loadings but different item thresholds between the U.S. and Mexico.

Informants in Mexico report less cognitive impairment when controlling for direct cognitive assessments.

Results suggest potential underestimation of cognitive decline in Mexico via either direct assessments or informant reports.

Abstract

Population studies of cognitive functioning rely on survey questionnaires to estimate the prevalence of dementia in nationally representative samples. In addition to direct tests of cognitive functioning, the international studies of the Harmonized Cognitive Assessment Protocol also include informant questionnaires, in which family members or other persons close to the research participant are asked about the participant’s cognitive functioning. While international harmonized measures of directly assessed cognitive have been developed, there are no harmonizations of informant assessments. This study investigates differences in how informants respond to questions about cognitive functioning in the US and Mexico. Using data from the Harmonized Cognitive Assessment Protocol (HCAP) of the Health and Retirement Study (HRS) (n = 3,183) in the United States and the Mexican Cognitive Aging Study (MexCOG) (n = 1,259) in Mexico, we examined informant reports of cognitive functioning of the Community Screening Instrument for Dementia (CSID). We test for differential item functioning (DIF) using structural equation modeling (SEM) to develop a confirmatory factor analysis (CFA) model of the CSID informant questionnaire in Mexico and the U.S. We find that a harmonized measure of the CSID has fixed factor loadings between countries but different item thresholds. Our harmonized measure of informant reports of cognitive functioning suggests that informants report less cognitive impairment for older adults in Mexico when controlling for directly assessed cognitive functioning. Our results suggest that either the harmonized direct assessment of cognition may underestimate cognitive functioning in Mexico or that informants in Mexico are systematically underreporting symptoms of cognitive decline.