Assessing cognitive decline in Vietnamese older adults using the Montreal Cognitive Assessment-Basic (MoCA-B) and Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) during the COVID-19 pandemic: A feasibility study

Abstract Objectives: The lack of cognitive assessment tools suitable for people with minimal formal education is a barrier to identify cognitive impairment in Vietnam. Our aims were to (i) evaluate the feasibility of conducting the Montreal Cognitive Assessment-Basic (MoCA-B) and Informant Questionnaire On Cognitive Decline in the Elderly (IQCODE) remotely on the Vietnamese older adults, (ii) examine the association between the two tests, (iii) identify demographic factors correlated with these tools. Methods: The MoCA-B was adapted from the original English version, and a remote testing procedure was conducted. One hundred seventy-three participants aged 60 and above living in the Vietnamese southern provinces were recruited via an online platform during the COVID-19 pandemic. Results: IQCODE results showed that the proportions of rural participants classified as having mild cognitive impairment and dementia were substantially higher than those in urban areas. Levels of education and living areas were associated with IQCODE scores. Education attainment was also the main predictor of MoCA-B scores (30% of variance explained), with an average of 10.5 points difference between those with no formal education and those who attended university. Conclusions: It is feasible to administer the IQCODE and MoCA-B remotely in the Vietnamese older population. Education attainment played a stronger role in predicting MoCA-B scores than IQCODE, suggesting the influence of this factor on MoCA-B scores. Further study is needed to develop socio-culturally appropriate cognitive screening tests for the Vietnamese population.

Vietnamese older population. Education attainment played a stronger role in predicting MoCA-B scores than IQCODE, suggesting the influence of this factor on MoCA-B scores. Further study is needed to develop socio-culturally appropriate cognitive screening tests for the Vietnamese population.
Assessing cognitive decline in Vietnamese Older Adults using the Montreal Cognitive Assessment-Basic (MoCA-B) and Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) during the COVID-19 Pandemic: A Feasibility Study Appropriate cognitive assessment tools should be considered in terms of the culture and education level of the participants. However, most cognitive tests have been developed for use among Western populations. The lack of cognitive measurement tools for Asian countries, including Vietnam, is a limitation for Vietnamese people to access culturally appropriate assessment tools, especially for those with minimal formal education and individuals who are illiterate, accounting for 18% of the older population (Hoi et al., 2010).
Currently, there is a wide range of cognitive assessment tools being used to detect mild cognitive impairment (MCI) and dementia in populations with low education and literacy levels, such as the Mini-Mental State Examination (MMSE) (Pellicer-Espinosa & Díaz-Orueta, 2022). However, other studies revealed that the MMSE is unreliable for individuals with low literacy (Pellicer-Espinosa & Díaz-Orueta, 2022). The Montreal Cognitive Assessment (MoCA) was developed as an alternative method for the MMSE, with comparable performance in detecting MCI. It was initially validated among highly educated people. The MoCA-Basic (MoCA-B) was specifically developed for older individuals with low literacy levels by replacing literacy-dependent tasks with literacy-independent ones that measure the same cognitive function (Julayanont et al., 2015). Evidence for the validity of the MoCA-B has been demonstrated in Thailand and China, two countries that have cultural and socioeconomic aspects similar to Vietnam (Chen et al., 2016;Julayanont et al., 2015). It may be an appropriate cognitive assessment instrument for older people with low levels of education. Hence, the first main objective of this study is to examine the performance of MoCA-B in the older Vietnamese population and explore the different demographic factors that may affect MoCA-B results.
Vietnamese older adults with low cognitive and education levels often find completing these assessment tools by themselves difficult. The presence of an informant who is knowledgeable about the older person is a potentially useful source of data. The Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) may reduce the influence on educational attainment (Fuh et al., 1995;Jorm & Jacomb, 1989;Perroco et al., 2008) through a structured interview method with a collateral (Harrison et al., 2016;Tokuhara et al., 2006). Proponents of IQCODE suggest several potentially beneficial properties of when compared to standard assessments (such as MMSE or MoCA): (1) IQCODE may be less likely to deviate from cultural norms and educational levels because its interview items are mainly related to daily living activities and memory (Meyer et al., 2020); (2) the scale has good reliability among raters, with high uniform internal consistency and Cronbach's alpha in the range 0.93 to 0.97 (Jorm & Jacomb, 1989).The study of Meyer and colleagues demonstrated the feasibility and acceptability of the 16-items version of IQCODE among Vietnamese Americans in the US (Meyer et al., 2020). To our knowledge, the full version with 26 items of IQCODE has not been standardized as a cognitive assessment tool in native Vietnamese people. The second goal of our study is to examine the implementation of IQCODE on the Vietnamese older population through collateral reporting, while also exploring the demographic factors that influence IQCODE scores.
The IQCODE and MoCA-B are two measures that are potentially appropriate for use in low and middle-income countries like Vietnam. However, the relationship between IQCODE and MoCA-B is not well established. There was a moderate, inverse association between MoCA and IQCODE in a study between two ethnicities: Mexican Americans and non-Hispanic White (R 2 = 0.29) (Briceño et al., 2022). Therefore, the third objective of this study is to identify the correlation between these two commonly used screening tools for detecting potential cognitive impairment in Vietnamese older adults.

The procedure and participants
Participants were recruited by local volunteers and came from seven Vietnamese provinces from urban and rural locations. Data were collected between January 15, 2022, and May 31, 2022. The inclusion criteria that the informants had to have known the participants for a minimum of 10 years, leading to reasonable assumption that the responses were based on well-established relationships across the sample. Of the 87 participants being excluded out of 260 initially recruited were due to the history of 1) severe brain traumatic injury (n = 10), 2) stroke (n = 15), 3) severe to very severe depression (n = 31), 4) change of intention to take part in the survey by the participants or their family members (n = 25), 5) innumeracy (n = 1), 6) other reasons (n = 5).
Demographic information of participants recorded in the study included age, gender (male and female), and educational level (coded as "illiteracy" (no education or has attended some school but cannot read), "primary school," "high school," and "college/ university"). The occupation was coded as "not currently in employment" (including those retired, or unemployed with no part-time work) or as currently employed in predominantly "mental" or "manual" labor. Lifestyle and current comorbidities (diabetes, hypertension, hypotension, brain injuries, stroke, dyslipidemia, cardiovascular diseases, hyperacusis, and others) were also ascertained by self-report.
The convenience sample was identified by local volunteers and recruited to obtain a balance of male:female (1:1) and urban:rural (1:1) ratio. The study adopted a convenience sampling method. The local volunteers were recruited online. Then, local volunteers invited their older neighbors or members of local clubs for older people. The study did not include people from nursing homes and social protection centers. Demographic characteristics are provided in Table 1.
The study was conducted via an online platform due to the COVID-19 pandemic and restrictions on traveling. The procedure was designed following the instructions of the Inter Organizational Practice Committee (IOPC) recommendations for tele-neuropsychology (Bilder et al., 2020). It has been demonstrated to be feasible and is considered a standard for improving the accessibility of quality healthcare in rural areas (Lara et al., 2020;Wadsworth et al., 2016). A screening session to assess eligibility and to get informed consent was conducted face-to-face by a local volunteer before the interview. The IQCODE and MoCA-B were administered via online platforms (ZOOM app and meet.google.com). The local volunteers were present with the participants during the interview to assist them and to ensure that the testing conditions were not affected by environmental interference. Network technical issues (transmission lines), such as the clarity of the sound, and the comprehensibility of the questionnaire, were also checked and maintained during the remote testing process. There were a few cases of technical problems which have caused interruption during the MoCA-B interview. In these situations, we considered the severity of the problem to decide whether we should continue while taking notes at the time of the problems or cancel the assessment results. No case was canceled in the study. The study was approved by the School of Biomedical Engineering Ethics Committee, International University, Vietnam National University-Ho Chi Minh City, Ho Chi Minh City, Vietnam.

The Montreal cognitive Assessment-Basic (MoCA-B) Vietnamese version
The MoCA-B is a 30-point test that evaluates six cognitive domains: executive functioning, visual perception, language, attention, delayed recall memory, and orientation (Julayanont et al., 2015). There is evidence demonstrating that the MoCA-B is a reliable screening tool for older adults from illiteracy to more than 12 years of education (Chen et al., 2016;Julayanont et al., 2015;Pellicer-Espinosa & Díaz-Orueta, 2022;Saleh et al., 2019). Other studies have demonstrated that demographic factors may influence the performance of MoCA-B, such as age and years of education (Hayek et al., 2020;Pellicer-Espinosa & Díaz-Orueta, 2022). To incorporate the existing adjustment for illiteracy, if the MoCA-B score was less than 30, one additional point was added to the score of participants who were considered illiterate, regardless of their education level (Janelidze et al., 2017).

Adaptation and translation of MoCA-B
The Vietnamese version of MoCA-B was translated and adapted based on the original English version on the MoCA official website: https://www.mocatest.org. The MoCA-B was adapted following the instructions of previously published guidelines (Guillemin et al., 1993) (Sousa and Rojjanasrirat, 2011). This procedure has been used successfully to adapt neuropsychological assessment tools and questionnaires for the Vietnamese Tran et al., 2013). The MoCA-B translation was authorized by Dr. Ziad Nasreddine who created the test.

Forward translation.
Two graduate Vietnamese researchers who majored in psychology and were fluent in English, but had no previous understanding of the MoCA-B, independently conducted the forward translation. After that, the two members developed a unified version of the forward translation.
Backward translation. One Vietnamese neuropsychologist and one Vietnamese researcher who were knowledgeable in psychology and neuroscience and fluent in English, but not familiar with the MoCA-B, independently carried out a reverse translation from the previous unified translation. They then met to finalize a reverse translation.

Modification and improvement.
Any disagreements between the translated and back-translated versions were revised and resolved by translators and an expert neurologist. Several revisions were made to make the MoCA-B more culturally appropriate: Immediate recall. The word "ghế," "sáo," or "tàu" were changed to "cái ghế," "cây sáo," or "tàu lửa," to ensure consistency in the number of two sounds in each word so that the words could be as easily comprehended as possible.
Calculation. The Vietnamese older adults may not be familiar with the foreign currency "dollar," so the Vietnamese Dong currency was used. They also may not be familiar with the concept of "bill" -as cash is still a popular exchange choice in Vietnam -thus it was substituted with "item with a value of …" Abstraction. In Vietnamese, "banana" means "quả chuối" and "orange" means "quả cam." The word "quả" indicates that this pair is part of the fruit category, so "quả" was removed from the instruction.
Twenty external auditors were recruited to read a pre-final draft. They were volunteers from 24 to 80 years old and lived in both urban and rural areas. Their educational backgrounds ranged from 0 to 16 years of education. Follow-up interviews focused on the readability and cultural appropriateness of the scale. Any comments raised were considered for the final version.
Remote testing. The remote assessment process was piloted including 15 subjects (6 men and 9 women) to check the feasibility of this procedure. Their ages ranged from 61 to 92 years old. All lived in the rural areas of Vietnam. We consulted recommendations and instructions for implementing the MoCA-B https://mocatet. org/remote-moca-testing/. The MoCA-B was administered by certified investigators in the research team, which included 5 members: 3 psychologists, 1 psychiatrist, and 1 neurologist. Some technical problems occurred during the pilot testing and had been recorded to search for solutions.

The informant questionnaire on cognitive decline in the Elderly (IQCODE)
The IQCODE is a 26-item informant questionnaire that retrospectively examines the consistent change in learning, recall abilities, recognition, comprehension, and other aspects of intelligence (Jorm & Jacomb, 1989). A trained rater asks questions to an informant -who has known the participant for at least 10 years. This was to assess an individual's change in cognition over the last 10 years. If the informant did not know the participant for 10 years, another informant was recruited who met the 10 year requirement. The questionnaire uses a 1-5 Likert scale, with 1 being "much improved" and 5 being "much worse. " Items cover the assessment of working, short-term and long-term memories, as well as daily activities. To score the IQCODE, add up the score for each question and divide by the number of questions. For the long IQCODE, divide by 26. The result is a score that ranges from 1 to 5. A score of 3 means that the subject is rated on average as "no change. " A score of 4 means an average of "a bit worse, " and 5 means an average of "much worse" (Jorm, 2004). If the long IQCODE is used for screening for dementia, a cutting point of 3.27/3.30 balances sensitivity and specificity. The questions are prompt, and the assessment time is as short as seven minutes. The cut-off score for subjects with mild cognitive impairment (MCI) is from 3.3 to 4.0 and for dementia is from 4.0 to 5.0, as previously conducted (Jorm, 1994(Jorm, , 2004Jorm et al., 1991). The Vietnamese version of the IQCODE used in our study was referred from the one in Meyer and colleagues' study (Meyer et al., 2020). The Vietnamese version of IQCODE was administered to 20 volunteers. The participants consisted of individuals ranging in age from 24 to 80 years old, and they resided in both urban and rural areas. They had varying educational backgrounds, from 0 to 16 years of education, and follow-up interviews were conducted to assess the readability and cultural appropriateness of the scale

Statistical analyses
Principal components analysis was conducted using SPSS 22.0, all other analyses were conducted using R version 4.1.0. Code (R Core Team, 2021).

Demographic characteristics and descriptive analysis
IQCODE and MoCA-B scores and participant ages were described using medians and interquartile ranges. Categorical demographic variables were described by frequency. Chi-squared tests were used to make pairwise comparisons between categorical variables with p < .05. Scale reliability was estimated using Cronbach's alpha.

Predictors of MoCA-B and IQCODE scores
Linear regression models were used to identify the predictors of MoCA-B and IQCODE scores within this sample. The first models included only demographic factors, then a model was estimated to determine the association between educational attainment and MoCA-B scores conditional on subjective cognitive function measured by IQCODE. Statistical model validity was tested by visual inspection of diagnostic plots. Age was entered as a continuous variable in the models, but conclusions are unaffected if a categorical variable was used.

Contribution of demographic factors (age, education) and IQCODE to individual MOCA-B components
As a post-hoc analysis, to better understand how age, education, and informant ratings of cognitive impairment contributed to MoCA-B scores, the relative contributions of age, education, and IQCODE score to each MoCA-B domain were calculated using the semi-partial R-squared. For each predictor, this is the increase in variance explained by that predictor when added to a model already including each of the other predictors. A linear regression model for each MoCA-B item was estimated independently, and the semi-partial R-squared was calculated from each model using the relaimpo package for R (Groemping, 2007).

Factor structure of the IQCODE
Principal component analysis (PCA) of IQCODE was performed using the promax rotation with Kaiser normalisation, to identify the structure of the cognitive domains assessed by the questionnaire. The number of components was fixed to three.
Finally, the association between components of IQCODE (measured by scores from the PCA) and the corresponding cognitive domains of MoCA-B was then examined, using Spearman's correlation coefficient statistical significance called at p < .05. Correlation coefficients between .10 and .29 represent a small association, coefficients between .30 and .49 represent a medium association, and coefficients of .50 and above represent a large association or relationship.

Demographic characteristics
Two hundred and sixty participants with family member informants from seven southern provinces initially agreed to participate in the study. Of those, 173 participants were eligible and are included in the final analysis. The demographic characteristics of 173 participants are shown in Table 1. The sample has an approximately equal number of males versus females in both rural (41 and 41) and urban (42 and 47) settings. Twelve participants were illiterate. The number of illiterate people is six times as many among the rural participants (10/82 = 12%) compared to urban participants (2/89 = 2.2%). Conversely, more urban participants had completed high school and college than rural participants, with only one college-educated rural participant compared to 36/89 (~ 40%) among urban participants. Regarding employment status, there were noticeable differences (chi-squared test for the association between employment status and urban/ rural living p < .001): (i) the rate of participants who were not currently employed (including retired people) in urban areas (58/89 = 65%) was higher than in rural areas (33/84 = 39%); (ii) manual labor was more common among rural participants (49/84 = 58%) compared to urban (14/89 = 16%), while (iii) mental labor was prevalent among urban participants (17/89 = 19%) compared to only (2/84 = 2%) of rural participants.

IQCODE scores and their correlation with demographic characteristics
The Vietnamese version of IQCODE demonstrated high internal consistency reliability for a Cronbach's alpha value of 0.947. The median and interquartile range of IQCODE was 3.27 (3.08, 3.58). Based on standard IQCODE cut-offs, 52.6% of participants would be considered to have "normal" cognition (IQCODE score ≤ 3.3); 40.5% with symptoms of MCI, whereas only 6.9% had reported IQCODE scores suggesting dementia (4.0 < IQCODE score ≤ 5.0). The specific response for each IQCODE item is depicted in Supplement Figure 1. Of the informants in our study, 46.8% of the informants were adult children, 42.2% were their spouses. The proportion of household members/ informants by urban versus rural areas corresponds to the proportion of elderly people in urban and rural areas because they live with the elderly.
Based on IQCODE scores, the proportions of rural participants who would be classified as having MCI (54% male and 44% female) and dementia (7.3% male and 17% female) were significantly higher compared to urban ones (36% male and 32% female for MCI; 0 male and 2.1% female for dementia) (Chi-square test, p < .001). The IQCODE total scores' median was also lower for rural compared to urban participants (Chi-square test, p < .001), as shown in Table 2.
To identify demographic predictors of IQCODE scores, a multivariable regression model was applied. Regression coefficients are shown in Table 3 and data is visualized in Figure  1. The associations between age or gender and IQCODE were not statistically significant (Table 3, Figure 1A and B). IQCODE scores were higher among those who were illiterate or had only completed primary school compared to those with high school or college education ( Figure 1D). According to the regression model, IQCODE scores were correlated with living areas and educational level, with education explaining 12% of the variance in IQCODE, and a 0.34 point difference between the groups with the highest and least education (p = .03) (Table 3). Similarly, when classified into cognitive function groups, IQCODE results were correlated with educational attainment and living areas (χ2 (6, 173) = 19.58, p = .03 and χ2 (2, 173) = 16.49, p < .001, respectively) Supplement Table S5). There was little evidence for any association between physical comorbidities and cognitive function in our dataset either in sum or individually (Supplement Table S5), although numbers for specific comorbidities are low and their ascertainment may have been affected by recall bias among more cognitively impaired groups.

MoCA-B and its correlation with demographic characteristics
The median and interquartile range of MoCA-B were 24 and (20, 26). The medians and interquartile ranges of MoCA-B results according to demographic factors, lifestyle, as well as comorbidities, were described in Supplement Table S2. The mean time to perform the test was 18.3 ± 5.6 min (maximum = 39 min; minimum = 9 min). The Vietnamese version of MoCA-B demonstrated moderately high internal consistency reliability for a Cronbach's alpha value of 0.826.
To identify demographic predictors of MoCA-B scores, a multivariable regression model was applied and the coefficients are shown in Table 4 and data is visualized in Figure 2. Within this sample, no correlation between gender and MoCA-B scores was identified (Table 4, Figure 2A). In contrast to IQCODE, there was a strong association between age and MoCA-B scores (an average of 0.3 MoCA-B points per year of age,   Figure 2B). In the multivariable model, the MoCA-B scores in rural participants was lower by 1.4 points (95% CI= −0.15 − 2.8) compared to urban participants (Table 4, Figure 2C), with most of this explained by the difference in educational attainment between rural and urban groups. The regression model (including age, sex, education, and living areas) for MoCA-B explains a substantial proportion of the variation in MoCA-B scores within this sample (R-squared = 43%) with 27.4% of the variance in MoCA-B explained by educational attainment alone, despite the extra points for illiteracy already present in the MoCA-B calculation (Table 4). After adjusting for age, sex, and urbanicity, those with college-level education scored on Note. n = 173. Multivariable regression model was applied to explore the associations between demographic factors (sex, age, living area and educational attainment) and iQCoDe scores. average 6.9 (95% CI = 3.6 − 9.7) points higher than those who were illiterate, and 3.2 (95% CI = 1.1 − 5.4) points higher than those with primary level education only ( Figure 2D).

Association between IQCODE scores and MoCA-B scores
A multivariate regression of MoCA-B scores on IQCODE and educational attainment reveals a strong effect of education on MoCA-B after adjusting for IQCODE, while there is no evidence for any relationship between IQCODE and MoCA-B after stratifying for education (Table 5, Figure 3). To better understand the contributors to the MoCA-B score we calculated the percentage of each item score explained by age, educational attainment, and IQCODE scores (semi-partial R-squared shown in Table 5, unadjusted R-squared shown in Supplement Table S4). Age and education were consistently more important determinants of MoCA-B item scores than IQCODE. In particular, executive function, orientation, abstraction, and visuoperceptual were the cognitive functions explained most substantially by years of education (Table 5).

Correlation between cognitive components of IQCODE and MoCA-B
Principal component analysis (PCA) of the IQCODE revealed a single dominant component reflecting overall cognitive function and explaining 44% of variance. A three-factor solution was then requested to further examine the correlation structure within the measure and Note. n = 173. Multivariable regression model was applied to explore the associations between demographic factors (sex, age, living area and educational attainment) and MoCa-B scores. revealed three correlated components: 1) executive function/calculation; 2) short-term memory/orientation; 3) long-term memory (Supplement Table S3). There was a small negative correlation between IQCODE memory/orientation component and the sum of MoCA-B memory and orientation questions (Spearman's rho = −0.164, p = 0.03), but little evidence for a correlation between IQCODE executive function/calculation component and MOCA-B executive function/calculation (Spearman's rho = −0.119, p = 0.118).

Quality of remote assessment sample
Our main purpose was to search for cognitive assessment tools that minimize cultural differences and were suitable for the illiterate, and those with formal education. Due  to the Covid-19 pandemic, the evaluation procedures were completed remotely. The MoCA-B has been considered as a potential candidate, as the MoCA website offers instructions for remote assessment (Phillips et al., 2020). A study on older veterans showed good accuracy and inter-rater reliability when comparing in-person and video administration of MoCA (DeYoung & Shenal, 2019). There were also other study findings demonstrating that this approach to data collection is possible when mediated with local support. To ensure that the remote assessment was conducted as efficiently and accurately as the offline assessment, we had to consider the factors of test space, sound, and transmission. Therefore, local volunteers have an important role in controlling these objective factors in the evaluation process. As a result, they had to be well-trained in working with MoCA-B remote assessment. Online platforms such as Zoom and meet.google.com have allowed investigators to connect with elderly people in remote areas.
The demographic characteristics of the collected sample were broadly representative of the Vietnamese population (General Statistics Office, 2019). There is a risk of bias using this approach to selection and recruitment as there was considerable variation in the educational attainment of cognitively impaired participants in our study. The training of the investigators, emphasis on objectivity, sufficient allocation of time for implementation, and other factors helped to minimize potential influence in test results. The study also demonstrated the potential for local volunteers to support the data collection and have a key role in providing technical support during their research participation.

Education is the main predictor of IQCODE and MoCA-B scores in older Vietnamese
In this study, we found a slight correlation between IQCODE scores and educational attainment. These findings reflect those of studies using the IQCODE questionnaire in Thailand and China (Fuh et al., 1995), and may reflect genuine differences in cognitive decline between these groups. On the other hand, there was a very strong effect of educational attainment on the MoCA-B score. 30% of the variance in MoCA-B was explained by educational attainment alone, despite the extra points for low education already present in the MoCA-B calculation (Table 4, Figure 2). Notably, none of the participants who were reported to be illiterate scored more than 21 points out of 30. Executive function, orientation, abstraction, and visuoperceptual are the cognitive function scores most substantially affected by educational level. These results are consistent with previous studies on Thai, Chinese, Arabic, and Ecuadorian versions of the MoCA-B which demonstrated how educational level of achievement affected final scores (Chen et al., 2016).
There are several factors that may explain the strong influence of education on the performance of MoCA-B across many studies, including ours. Firstly, our study considered years of schooling as the main criterion to define education. However, the quality of education might not be equivalent to years of education. In minority groups or in rural areas, the actual neuropsychological performance might be lower than expected when taking into account only years of education. Educational attainment is also likely to be associated with early life circumstances, income, lifestyles, and access to health care, which may in turn affect the results of cognitive assessment and the risk of late-life cognitive decline of older adults (Zahodne et al., 2015).
Another explanation is the impact of culture on education and neuropsychological test performance. The MoCA-B was modified to adapt to people with minimal formal education, however, its cultural aspects have not been assessed. A meta-analysis investigating the cross-cultural applicability of the MoCA (the original version of MoCA-B) showed its heterogeneity across many different cultures (O'Driscoll & Shaikh, 2017). Therefore, when using the MoCA-B on a specific population, it may be necessary to consider both the effect of acculturation and regional variation. Nowadays, most of the older people in Vietnam continue to live in local realities and are familiar with their traditional culture, not with patterns from foreign ones. The less acculturated people might perform lower than expected on many domains of cognitive tests (Manly et al., 2004).
An important psychosocial factor that is country-specific is the historical context of Vietnam and the effect of warfare exposure on educational attainment and cognitive function. Most of the population over the age of 60 spent their childhood experiencing poverty, adversities, and stressful living situations during the American War (the 1950s-1970s), and could not go to school. Early life trauma exposure (before five years old) impacts brain areas associated with cognition, memory, and learning therefore may affect not only learning ability but also neuropsychological performance in the elderly (Bremner, 2006;Op den Kelder et al., 2022). Additionally, traumatic events may be a risk factor for cognitive decline and Alzheimer's Disease in older people (Burnes & Burnette, 2013).
While the low educational level is a risk factor for cognitive impairment (Angevaare et al., 2022), the very large differences in MoCA-B scores between the groups seen in our sample are unlikely to be attributable to the genuine difference in cognitive function, particularly since they persist after stratifying for IQCODE scores. Cultural adaptations and the definition of appropriate cut-offs for the Vietnamese older population are likely to be necessary. Further, attention should also be paid to the development of other non-verbal cognitive assessment tests for the illiterate and those with minimal formal education.

Weak correlation between the IQCODE and MoCA-B
The lack of association between these two tests might be explained by the difference in their approaches. The MoCA-B is a performance-based tool, which provides an objective overview of the cognitive functions of participants at the time of evaluation. Whereas informant-based instruments like the IQCODE rely on proxy assessment to understand changes in cognition in the 10 years prior to the interview. The informants had known participants for at least 10 years, it is therefore, reasonable to assume responses were based on a well-established relationship. However, the characteristics of informants can impact the outcomes of IQCODE scores based on how they interpret the questions, for example, the informant's age (same versus younger generation), their emotional state, care-related stress and ability to assess the person's cognitive functioning (Nygaard et al., 2009). Whether informant judgments are more likely to minimize cognitive deficits is under-researched. Despite that, evidence from research conducted in Vietnam suggested that neurologically healthy informants (who are immediate family members of main subjects) have appropriate insights and are generally a reliable source to obtain information about neuropsychological features of both older healthy adults  and those with clinical conditions affecting cognition, such as traumatic brain injury .
There have been several studies reporting the principal components analysis of the IQCODE items, which indicated that IQCODE measured primarily one general factor of cognitive decline (Jorm & Jacomb, 1989). In our study, IQCODE items may be categorized into three main cognitive components (Supplement Table S3). As we compared each component of IQCODE with the corresponding items of MoCA-B, there was only a small negative association in memory orientation between the two cognitive tests. These results implied that IQCODE and MoCA-B might be addressing different aspects of cognitive domains, which might explain the weak correlation between the two scores.

Limitations and future implications
This study was based on a convenience sample of older Vietnamese and so associations may not be representative of the general Vietnamese community, however, we do not believe the sampling could be solely responsible for the large effect of educational attainment on cognitive scores that we observed. Further studies are needed to explore the role of cultural stigma on willingness to endorse symptoms on the IQCODE. Future works could look into examining the MoCA-B in other clinical groups or explore other cognitive screening tools that can detect cognitive impairment without having educational attainment and/or living areas impacting its results in Vietnamese elder populations.

Conclusion
We found that it was feasible to implement the IQCODE and MoCA-B remotely for cognitive assessment in the Vietnam order population. However MOCA-B scores are considerably lower among illiterate people and those with few years of education, independently of IQCODE score. Larger studies based on randomly selected participant samples with concurrent validated clinical assessments of cognitive function are necessary to better understand the performance of cognitive tests and to define the cut-off score of MoCA-B in the Vietnamese population and potentially, modify MoCA-B and IQCODE so that they are less dependent on education level and cultural factors in this group.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This work was funded by the Global Challenges Research Fund Networking Grant (GCRF) project number GCRFNGR6\1403. Claire Goodman is part-funded by the NIHR Applied Research Collaborations for East of England (ARC-EoE) and is an NIHR Senior Investigator. The views expressed in this publication are those of the authors and not necessarily those of the NIHR. Huong Ha, Toi Vo, and Quynh Nguyen are funded by Vietnam National University Ho Chi Minh City (VNU-HCM) under grant number NCM2020-28-01.