Navigating the landscape of cognitive screening in multicultural memory clinics

Europe, like much of the world, faces a profound demographic shift with those aged ≥65 projected to rise to approximately 33% by 2050. ¹ The prevalence of dementia, including Alzheimer's disease (AD), will increase commensurate with this, challenging healthcare systems. ² This aging population is becoming more diverse, with a growing share with an immigrant background. ³ Consequently, memory clinics in high-income countries now serve increasingly heterogeneous populations spanning multiple languages, ethnicities, and cultures. ⁴ These changes demand critical reassessment of cognitive screening and diagnostic processes to ensure they are fit for purpose in more multi-cultural populations.^5,6

Cognitive screening instruments (CSIs), typically administered within 30 min, ⁷ are widely used in busy clinics to triage patients, inform diagnostic pathways, and determine who may require more in-depth neuropsychological assessment.^8,9 Despite their efficiency, the sheer number and variety of CSIs makes standardization difficult. Notably, tools like the Mini-Mental State Examination (MMSE) ¹⁰ and Montreal Cognitive Assessment (MoCA) ¹¹ were validated in relatively homogenous populations, raising concerns about cultural bias.^12,13 For example, the MMSE and MoCA contain culturally-specific items (e.g., language-dependent tasks), which may disadvantage individuals from different educational or cultural backgrounds. In addition, instruments such as the MMSE are less reliable and accurate in patients with low literacy levels. ¹² Clinicians increasingly recognize that using a single, harmonized CSI is inappropriate in all settings and particularly across diverse populations, potentially underestimating cognitive function in individuals from non-dominant cultural or linguistic backgrounds.^6,14,15 Nielsen et al.'s paper is timely in this regard. ¹⁶

Nielsen and colleagues assessed five brief CSIs, the Multicultural Cognitive Examination (MCE), Brief Assessment of Impaired Cognition (BASIC), BASIC Questionnaire (BASIC-Q), Rowland Universal Dementia Assessment Scale (RUDAS), and Category Cued Memory Test (CCMT) in six clinics across five European countries. In a pooled sample of native-born and participants with immigrant backgrounds (56%), specifically those who had migrated to Europe (predominantly from outside of the continent, except seven individuals who had moved internally), the authors showed that all tools showed fair to very good accuracy separating controls from mild cognitive impairment (MCI) and very good to excellent accuracy for dementia. The MCE, BASIC, and BASIC-Q were statistically more accurate overall. Diagnostic accuracy did not differ significantly by immigrant status or education, suggesting these CSIs are appropriate for use in multicultural clinical settings. ¹⁶

Caution is warranted, however. Despite the study's impressive breadth (sampling immigrants from 40 countries, speaking 33 distinct languages), subgroup sizes were small and no statistical power calculation to assess sample size was provided, raising concerns about selection bias, limited representativeness and reduced generalizability. Half of the 187 participants were recruited from a single site in Copenhagen, Denmark, while other sites, such as London, in the United Kingdom (n = 1) and Paris, France (n = 4), contributed minimally and included no native-born participants. Further, native-born participants were from either Denmark or Spain. This all skews the data toward a Danish experience. Further, only six were from sub-Saharan Africa, and the sample from South America was concentrated in one country, Spain. Accordingly, Spanish was the most common first language (36%). While the authors provide demographic details, this distribution raises questions about whether the sample reflects broader immigrant populations across Europe or if patients who attended these memory clinics were a more health-literate, socioeconomically advantaged, or otherwise selective subset, which could further reduce generalizability. Additional details, made available by the authors following publication, suggest that approximately 10% of the sample with immigrant backgrounds had no formal education, while 20% worked in a profession (skilled occupation). Understanding the representativeness of the sample is still challenging as few studies have specifically examined the socioeconomic characteristics of middle-aged and older adults from such diverse backgrounds and especially in those attending memory clinics. One study using data from the Survey of Health and Ageing in Europe (SHARE), predominantly including native born Europeans who moved to another European country, found that 10.4% of these self-report having no formal education, albeit that study did not provide a breakdown by cognitive status. ¹⁷ Similarly, the proportion previously employed in skilled professions is broadly similar to European data that suggests approximately one-third of those with an immigrant background work in such roles. ¹⁸ The absence of information on age at immigration in the study by Nielsen and colleagues is a limitation, ¹⁶ as earlier arrival often correlates with higher educational and occupational integration, and studies indicate that those immigrating as adults are more likely to experience underemployment or unemployment.^19–21

Language proficiency and interpreter use are another factor in cross-cultural assessment. ²² In this study, most participants with an immigrant background (75%) required an interpreter or researcher-facilitated translation. While this underscores the importance of language-sensitive testing, the potential impact of interpreter-mediated assessment on diagnostic accuracy was not explored. Untrained interpreters can undermine comprehension and response validity, especially in cognitively impaired individuals.^6,23,24 Given the cost and limited availability of trained interpreters, this is a practical concern. More research is required to evaluate this, although for some cross-cultural CSIs, no significant influence of interpreter use on diagnostic accuracy has been found. ²⁵ Irrespective, the requirement for an interpreter and the emotional overlay associated with this interaction can influence the results of testing, and as with any social interaction it can act a cognitive disrupter, potentially adversely impacting the testing environment.^23,26 Another limitation is that the authors did not provide specific details on whether participants were bilingual or the degree of fluency achieved in their adopted or second language, only whether they were assessed in that second language, which appears to have been the case for only a minority and was often necessitated when interpreters were unavailable. It is important to understand the effects of bilingualism and its complex relationship with biculturalism as this can also affect cognitive testing reducing accuracy.^27,28 The psychometric equivalence of different language versions of CSIs, particularly for use with balanced bilinguals (who are fluent in two languages) is uncertain and requires research. ²⁸

Although translations of the MMSE and MoCA exist for many of the 33 languages spoken by study participants, (for example the MMSE is available in ≥70 languages, and the MoCA in ≥100 languages and dialects), ²⁹ they were not directly compared to the CSIs in this study. Further, it is unclear how many valid translations exist for the MCE, BASIC, BASIC-Q, RUDAS, and CCMT to facilitate their use without trained translators. Few comparative studies exist to date, and those that do are often impeded by small, heterogeneous, and likely underpowered samples. Reflecting this limitation, comparisons between the MMSE and RUDAS frequently show no statistically significant differences in diagnostic accuracy.^30,31 Before this study, the MCE had only been validated against the RUDAS, ³² where it demonstrated superior accuracy for dementia, and then more recently in a Swedish cohort, where it performed comparably to the MMSE and RUDAS, but outperformed the Clock Drawing Test. ³³ A study of the BASIC and BASIC-Q found these tools to be significantly more accurate than the MMSE in distinguishing MCI from controls, though their ability to differentiate MCI from dementia was similar, likely reflecting the known limitations of the MMSE in detecting MCI. ³⁴ As with the present study, small sample sizes in these studies have limited efforts to assess the performance of these CSIs across different dementia subtypes (most had AD). ¹⁶ This underscores the need for comprehensive, head-to-head comparisons of translated versions of both traditional and culturally-attuned instruments, conducted in larger, diverse samples and across various dementia subtypes. This is especially critical as these tools gain wider validation across settings and languages. However, each translation, modification and cultural adaption of brief CSIs can affect test quality and fidelity with the original language version, potentially reducing accuracy and limiting comparisons across cultures and languages, ³⁵ suggesting that such studies should be conducted carefully.

The European Consortium on Cross-Cultural Neuropsychology (ECCroN) stresses the need for tools that validly assess core cognitive domains across diverse groups. ¹⁴ Many traditional CSIs rely heavily on language-based and executive function tasks, domains linked to the prefrontal cortex that are influenced by education and cultural experience.^26,36 It would have been interesting to understand which components or items of these culturally adapted instruments most influence their accuracy. While the authors did not explore the effects of specific cognitive domains and individual subtests on the diagnostic accuracy of these instruments, future research could address this, particularly as culturally adapted instruments can in theory mitigate these effects, reducing linguistic demands by incorporating more nonverbal or contextual tasks. Without a deeper understanding of the cognitive constructs actually being measured and how these manifest across cultural and educational contexts, direct comparison of CSIs risks misclassification or underdiagnosis. Further, research is also needed to understand the neurobiological origin of cultural bias in cognitive testing, including how life experiences, which are intertwined with culture, physically shape the neural circuits involved. ³⁷ Indeed, the authors did not provide details on the life experiences of participants, particularly whether those with an immigrant background were first generation immigrants or whether this included an application for asylum or other form of international protection.

Finally, the assumption that immigrant groups have low education or literacy is increasingly being challenged, particularly among younger cohorts. ³⁸ Indeed, Nielsen et al. show similar educational attainment among native and immigrant groups, with the latter being younger on average. ¹⁶ This suggests that cohort effects, as younger generations achieve higher education, may positively influence cognition, ³⁹ acknowledging that relative differences between non-immigrants may be accounted for by an immigrant health paradox, where patients from immigrant backgrounds are less likely to receive a diagnosis. ⁴⁰ Global education expansion suggests this cohort effect may be real and universal, though other factors like health behaviors may also contribute. ⁴¹ In this study, mean schooling was 8.1 years for dementia and 10.5 for participants with MCI, regardless of background. The proportion with <7 years schooling or illiteracy was not provided in the publication, but the authors have subsequently indicated it was 38%, which is similar to that of the sample of 95,940 native-born Europeans aged ≥50 years reported in SHARE (30.3%). ¹⁷ This indicates that low education may not have been a defining feature among these immigrants. However, without more detailed data on education quality or literacy, interpreting education's confounding effects is difficult. These nuances matter because much of the rationale for culturally adapted tools rests on assumptions about limited educational exposure, which may not hold true for all immigrant groups or for second-generation populations or equally between countries. ⁴² Receiver Operating Characteristic curve analyses by education or cultural background were not available, likely due to small samples, which could have clarified subgroup performance.

Clinically, how should these results guide CSI selection in diverse memory clinics? Should clinicians use instruments like BASIC or BASIC-Q for all, or tailor by background? Clinicians often select CSIs based on diagnostic probability or clinical factors (education, literacy, native language, or sensory impairment etc.). The need to tailor by cultural background is plausible but uncertain, especially among similar cultures such as native Spanish speakers from Latin America living in Spain. For example, the World Health Organization-Composite International Diagnostic Interview required modification for only 7% of 5000 questions for persons in Latin America, and although some were deemed important, most were minor. ⁴³ Profiling patients at referral for an appropriate CSI is challenging. Default cross-cultural CSIs might be a solution, but broader, adequately-powered comparative studies with validated translations are needed before changing practice. Importantly, CSIs are only one piece of a comprehensive diagnostic process. Integrating culturally sensitive assessment with biomarkers, neuroimaging, detailed informant-supported clinical histories and an assessment of functional abilities, especially in cross-cultural contexts will improve dementia diagnosis in multi-ethnic populations. ⁴⁴ Finally, the use of digitalized CSIs, which predominantly used non-verbal stimuli, should be encouraged as these can reduce the dependence on language in cognitive testing. ⁴⁵

In summary, the study by Nielsen et al. ¹⁶ adds valuable evidence to a growing area of research, emphasizing cultural evaluation of CSIs and the complexity of achieving cultural fairness in cognition assessment. ¹⁴ Given that access to memory clinics is often inequitable,^46–48 such research is essential to inform policy and promote inclusive care in an ageing, multicultural world. The use of CSIs must evolve to meet diverse needs without compromising accuracy. Clinicians, researchers, and policymakers must commit to the development of services that provide equitable, rigorous cognitive assessment for every patient. However, without large, adequately-powered, representative samples and exploration of interpreter effects, premature or narrow conclusions risk being drawn. Further research with diverse cohorts using cross-cultural translated instruments is essential. Ultimately, truly cross-cultural CSIs must address language bias as well as cohort effects, education quality, literacy, and acculturation, factors which are often intertwined and challenging to isolate. ⁴⁹