Neurocognitive testing in a diverse linguistic setting: Measurement invariance of ICMR‐Neuro Cognitive Tool Box across five languages

Background Recently there have been ongoing efforts to harmonise research across dementia cohorts from different centres and countries. Establishing comparability of cognitive test performance across languages and educational levels therefore becomes crucial. Differences due to educational disparity or linguistic variability will need to be differentiated from more genuine differences due to disease. This is especially important in countries like India that are characterised by a socioculturally heterogeneous population. To address this, the Indian Council of Medical Research‐Neuro Cognitive Test Battery (ICMR‐NCTB) was developed and validated to diagnose dementia in five Indian languages. In this study, we aimed to investigate the cognitive construct and comparability of the ICMR‐NCTB across five Indian languages using confirmatory factor analysis and measurement invariance. Methods 991 healthy literate controls with normal cognition (Telugu = 302; Hindi = 191, Malayalam = 195; Bengali = 135 and Kannada = 168) were assessed with the ICMR NCTB which includes tests of attention/executive function, verbal and visual episodic memory, language and visuospatial Functions. The dimensional structure of ICMR‐NCTB was studied with regard to its cognitive constructs, and measurement invariance across five different Indian languages was evaluated. Results Confirmatory factor analysis of the data demonstrated that a five factor model in which the tests were divided into the cognitive domains: attention/executive function, verbal memory, non‐verbal memory, language and visuospatial functions provided the best model fit both within and across all five languages. Analysis of factorial invariance across groups indicated that configural and metric invariance were achieved, thereby suggesting that the tests of the ICMR‐NCTB measure the same cognitive construct across different languages. On analyzing scalar invariance, non‐invariance of intercept was demonstrated, indicating a possible measurement bias between the different language groups. Conclusions The study results indicate that cognitive constructs measured by the tests in the ICMR‐NCTB across the five different Indian languages are the same. However, there are differences that influence the performance of subjects across languages groups. The study findings have significant implications for harmonising cognitive test data in dementia research that involves linguistically diverse cohorts.