# Measurement invariance of a general cognitive performance measure across 27 European countries and Israel

## TL;DR

This study examines whether a general cognitive performance measure is consistent across 27 European countries and Israel using data from the SHARE survey.

## Contribution

The study provides evidence on the measurement invariance of a global cognitive performance measure across multiple countries.

## Key findings

- A unidimensional model of GCP fit the data well in most countries.
- Approximately 31.85% of factor loadings and 54.81% of item intercepts showed noninvariance.
- Researchers should avoid cross-country comparisons of GCP due to noninvariance.

## Abstract

The use of global or composite cognitive measures is extended in both clinical and academic settings. In this line, several population-based surveys include measures of cognition that have frequently been combined into a single score. However, some methodological aspects of this practice have gone unnoticed. One of such aspects has been to provide evidence of the measurement invariance of the combined measure across countries involved in the surveys. Using data from the Survey of Health, Ageing and Retirement in Europe (SHARE), this study aims at providing evidence of the factor structure of a Global Cognitive Performance (GCP) measure and testing whether this structure remains invariant across 27 European countries and Israel. The sample was composed of 55,569 adults aged between 60 and 102 years old (M = 72.07, SD = 7.97). 56.58% were female. Confirmatory Factor Analysis was used to establish the measurement model of GCP in the general sample and within each country. Afterwards, measurement invariance across countries was evaluated using the traditional as well as the alignment approach. The unidimensional model of GCP deemed an adequate fit to the data in the general sample as well as within each country, except for Malta, which was excluded from further analyses. After dismissing full measurement invariance, we studied approximate measurement invariance using alignment. 31.85% of factor loading estimates were noninvariant, while 54.81% of item intercept estimates showed deviations from invariance. Given evidence of items’ intercepts and factor loadings noninvariance, researchers working with SHARE data should abstain from making cross-country comparisons of GCP. Some plausible explanations for noninvariance of items’ intercepts are further discussed.

The online version contains supplementary material available at 10.1007/s10433-025-00872-y.