Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Aug 25.
Published in final edited form as: Lancet Healthy Longev. 2026 Jul 12;7(6):100867. doi: 10.1016/j.lanhl.2026.100867

Differences in the prevalence and patterns of dementia risk factors across 14 countries and regions: a harmonised cross-national analysis

Emma Nichols 1,*, Zachary J Kunicki 1,*, Michael Markot 1, Drystan Phillips 1, Jenny Wilkens 1, Alden L Gross 1, Jinkook Lee 1
PMCID: PMC13502513  NIHMSID: NIHMS2198887  PMID: 42437564

Summary

Background

The design of evidence-based interventions to reduce the burden of dementia requires knowledge of the prevalence and patterns of modifiable risk factors. However, most existing evidence comes from high-income countries (HICs). Hence, we aimed to quantify differences in the prevalence and patterns of dementia risk factors across diverse contexts.

Methods

We conducted a comparative cross-sectional study using harmonised data from 14 countries and regions (Ireland, the USA, England, Northern Ireland, Eastern Europe, Western Europe, Northern Europe, Southern Europe, South Korea, Mexico, China, Malaysia, Brazil, and India), including HICs and low-income and middle-income countries (LMICs). We included individuals aged 50 years or older from 11 nationally representative ageing studies, using data from the most recent available study waves with refresher samples collected between 2009 and 2023. Respondents were excluded from specific analyses if data on the risk factors of interest were missing. We estimated the prevalence of 12 established binary dementia risk factors (low education, hearing loss, high LDL cholesterol, depression, physical inactivity, diabetes, smoking, hypertension, obesity, excessive alcohol consumption, social isolation, and vision loss) using descriptive statistics and examined patterns by age group (70 years and older and 50–69 years), gender, and education using Poisson models with robust variance estimation. We also compared the rank order of the prevalence of risk factors across countries and assessed risk factor co-occurrence and clustering.

Findings

Data on 214 251 respondents were included in the study. We observed some variation in the prevalence and patterns (by age, gender, and education) of risk factors between HICs and LMICs. For example, low education had higher prevalence in many LMICs (85·6% [95% CI 84·8–86·5] in China vs 12·0% [95% CI 11·3–12·7] in the US), whereas obesity was more prevalent in HICs than in LMICs (44·9% [95% CI 43·3–46·5] in the US vs 13·3% [95% CI 12·9–13·7] in India). Risk factor distributions differed by age group, gender, and education, although patterns were not consistent across all settings. Risk factors commonly co-occurred across settings, with more than 50% of individuals having at least two risk factors across all countries and regions. Moreover, broadly similar clusters of risk factors—related to cardiovascular diseases, risky behaviours, and social or sensory factors—were observed across settings.

Interpretation

Differences in the prevalence and patterns of dementia risk factors highlight the need to tailor prevention strategies to specific contexts. However, findings also reveal consistent patterns in risk factor co-occurrence and clustering, which could guide the design of multidomain interventions and policy approaches to reduce dementia risk across settings. Overall, these findings support the use of both context-specific and shared approaches to reduce the burden of dementia.

Funding

National Institutes of Health.

Introduction

Dementia prevalence is expected to triple by 2050, largely because of population ageing.1 However, there is increasing recognition that a large proportion of dementia burden is attributable to modifiable life course risk factors.2 According to the 2024 update of the Lancet Commission report, 45% of dementia burden is attributable to 14 modifiable risk factors—namely, low education, hearing loss, high LDL cholesterol, depression, traumatic brain injury, physical inactivity, diabetes, smoking, hypertension, obesity, excessive alcohol use, social isolation, air pollution, and vision loss.3 Understanding the prevalence and patterns of these risk factors is key for designing effective evidence-based interventions to reduce dementia burden.

Most evidence underlying the recommendations in the Lancet Commission report comes from high-income countries (HICs),3 but the need for data from low-income and middle-income countries (LMICs) is increasingly acknowledged, given potential differences across settings and contexts (ie, countries or regions with differing demographic, socioeconomic, and health system characteristics).4,5 For example, by integrating evidence on differences in the prevalence of risk factors across contexts, the proportion of dementia that was potentially modifiable was estimated to range from 40% in China to 56% in Latin America.6 More detailed country-specific analyses further evaluate the effect of common assumptions (eg, that the effects of risk factors do not differ across contexts) on estimates of modifiable dementia burden, with one study suggesting that up to 70% of dementia might be modifiable in India.7

Existing research on dementia risk factors in diverse settings has largely focused on population attributable fractions (PAFs) in specific countries (eg, Chile8 and Brazil9) without explicitly examining how risk factors coexist or how their prevalence and patterns vary. However, risk factors commonly co-occur and cluster together,10,11 and this co-occurrence contributes to dementia risk.12 Although this evidence on co-occurring risk factors has motivated the development of multidomain intervention trials, such as the Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER) and worldwide FINGER trials,13,14 comparative data on risk factor prevalence and patterns across contexts to guide the development or scale-up of policies or interventions are scarce.

This study aimed to fill this evidence gap by quantifying differences in the prevalence and patterns of 12 established dementia risk factors across 14 countries and regions in diverse global contexts to inform cross-country comparisons. To maximise comparability between contexts, we leveraged nationally representative harmonised data collected between 2009 and 2023. Findings can inform how policies and interventions targeting dementia risk factors should be tailored to varying country contexts.

Methods

Study design

In this comparative cross-sectional study, we used data from 11 nationally representative studies on ageing from the Health and Retirement Study International Network of Studies (HRS INS), a network of global aging studies designed to facilitate cross-national comparisons. To maximise comparability across datasets, we used harmonised HRS INS data from the Gateway to Global Ageing Data.15 This study adhered to the STROBE guidelines.

Data from The Irish Longitudinal Study on Ageing (TILDA), the Health and Retirement Study (HRS), the English Longitudinal Study of Ageing (ELSA), the Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA), the Survey of Health, Ageing and Retirement in Europe (SHARE), the Korean Longitudinal Study of Ageing (KLoSA), the Mexican Health and Aging Study (MHAS), the China Health and Retirement Longitudinal Study (CHARLS), the Malaysia Ageing and Retirement Survey (MARS), the Brazilian Longitudinal Study of Ageing (ELSI), and the Longitudinal Ageing Study in India (LASI) were included (appendix p 3). Study-specific details are available in the appendix (pp 4-5). To account for cultural and contextual differences across Europe, we divided SHARE into four European regions (north [hereafter SHARE N], south [SHARE S], east [SHARE E], and west [SHARE W]) based on UN Statistics Division classification.16 Thus, we analysed 14 countries or regions—namely, Ireland, the USA, England, Northern Ireland, Eastern Europe, Western Europe, Northern Europe, Southern Europe, South Korea, Mexico, China, Malaysia, Brazil, and India. We excluded respondents younger than 50 years to standardise the age inclusion criteria.

We used data from the most recent study wave with a refresher sample (hereafter referred to as the primary wave) for most source datasets to maximise policy relevance while ensuring representativeness. The primary study waves were conducted between 2009 and 2023 (appendix p 4). For some studies, data from older waves were used for the analyses of specific risk factors (details in appendix pp 6-16). All studies used multistage probability sampling to ensure population representativeness. Each study received approval from local ethics committees, and informed consent was obtained from all respondents. No ethics approval was required for this study.

Risk factors and covariates

We included 12 of 14 dementia risk factors identified in the Lancet Commission report and commonly available in the HRS INS. These risk factors were low education, hearing loss, high LDL cholesterol, depression, physical inactivity, diabetes, smoking, hypertension, obesity, excessive alcohol consumption, social isolation, and vision loss. Traumatic brain injury and air pollution were excluded because the data were unavailable.

Respondents self-reported details on their education, smoking status, physical activity levels, and alcohol consumption. Education categories were standardised based on International Standard Classification of Education (ISCED) guidelines.17 For this study, we considered education below the level of upper secondary education as low education. For primary analyses, we defined smoking as current or former smoking. We used established equations to estimate the Leisure Score Index (LSI) based on vigorous, moderate, and light activity and defined low physical activity as activity with a LSI of less than 14 (appendix pp 9-10).18 We defined excessive alcohol use as the consumption of 15 or more drinks per week for men and the consumption of eight or more drinks per week for women (appendix pp 13-15).19 High cholesterol, diabetes, and hypertension data were based on self-reported physician diagnosis. Hearing and vision loss were self-reported using survey questions with binary or Likert-scale response options (appendix pp 6-7, 16-17); ratings of fair or poor were classified as hearing or vision loss in all cases except for KLoSA and CHARLS. In KLoSA and CHARLS, differences in the distribution of responses suggested that respondents interpreted fair as adequate, rather than poor, and this was classified accordingly; the term fair when directly translated to Korean or Mandarin is commonly interpreted as adequate or reasonable, aligning with our descriptive observations and supporting our dichotomisation strategy.

Symptoms of depression were assessed using the Center for Epidemiologic Studies Depression Scale in most studies, with context-specific cutoffs based on existing literature.20 In SHARE, the EURO-D and its corresponding cutpoint were used instead (appendix pp 7-9).21 Because of the differences in the questions and response scales used to assess social isolation across studies, we used a simplified definition for social isolation based on living arrangements and classified individuals who live alone as socially isolated. Obesity was defined using two cutpoints. Based on the evidence indicating the need for alternative BMI cutpoints in Asian populations,22 we used a threshold of 27·5 for CHARLS, KLoSA, MARS, and LASI data and a threshold of 30 for other datasets. BMI was based on measured height and weight when these parameters were available, owing to the low risk of bias, and on self-reported height and weight in KLoSA. In MHAS, measured BMI from a previous wave was used in analyses examining individual risk factors, whereas BMI from self-reported data in the primary wave was used in analyses examining the joint distribution of risk factors. Additional details are available in the appendix (pp 6-16).

We used self-reported age and sex (hereafter gender) as covariates.

Statistical analysis

We used descriptive statistics to examine the prevalence of each binary risk factor. We also calculated the prevalence and means across risk factor definitions in sensitivity analyses. We used sorted prevalence estimates to calculate ranks and visualised the rank order of risk factor prevalences using a heatmap. To examine patterns among risk factors by age, gender, and education, we used study-specific and risk-specific Poisson models with Horvitz–Thompson-type robust standard errors (SE) to estimate risk ratios. Models were mutually adjusted for all three variables. To evaluate whether education patterns differed by gender, we ran follow-up models with interaction terms between education and gender.

In analyses of the joint distribution of risk factors, we excluded risk factors that were not available in the primary wave. We first calculated the weighted proportion of individuals with more than two, three, or four risk factors. Given differences in the number of risk factors available by study, we also estimated the mean number of co-occurring risk factors as a proportion of available risk factors for each study. Additionally, we identified risk factor clusters. We used parallel analysis to select the number of risk factor clusters in each study. Based on principal component analysis with varimax rotation, we assigned each risk factor to the component with the highest loading and named the clusters based on the included risk factors. We used a heatmap to visualise findings and compare patterns of clustering across countries and regions.

We conducted six sensitivity analyses intended to evaluate the effect of decisions related to the measurement of risk factors on prevalence estimates. First, we examined the prevalence of secondary and tertiary education and of low education. Second, we reported the prevalence of three smoking categories (never, former smoker, and current smoker). Third, we evaluated an alternative strategy to convert categorical data on physical activity to the LSI (appendix pp 9-10), while also reporting vigorous, moderate, and light activity separately. Fourth, because LASI did not include questions on the quantity of alcohol consumption comparable to those in other studies, we additionally evaluated a binary indicator for binge alcohol consumption across all surveys (appendix pp 15). We also evaluated potential self-report biases in cardiovascular disease risk factors by comparing objectively measured hypertension (or self-reported blood pressure medication use) with self-reported physician diagnosis of hypertension when data were available. To test if this bias was related to development status, we estimated the Pearson correlation between log-transformed gross domestic product (GDP) per capita and the difference between objective and self-reported hypertension prevalence. Finally, we considered versions of hearing and vision loss measures in KLoSA and CHARLS without applying language translation adjustments.

To identify patterns related to differences in economic contexts, we ordered studies by country GDP per capita for all results. A significance threshold of α=0·05 was used where statistical testing was applied; however, formal hypothesis testing was not conducted systematically across all analyses. All analyses used survey weights and accounted for survey design features when available (appendix p 17). For all analyses, we excluded respondents with missing data on risk factors of interest; principal component analysis used correlation matrices constructed from non-missing data on pairwise comparisons. Missing data were generally minimal, except for health variables in NICOLA and physical measures in HRS and MHAS (appendix pp 18-29). Because high levels of missingness in these variables were attributable to study design (these measures were not collected for the full sample), this missingness is less likely to be informative or cause bias in the results. All analyses were conducted using R version 4.5.1.

Role of the funding source

The funders had no role in the study design, data collection, data analysis, data interpretation, or writing of the report.

Results

We included 214251 respondents from 14 countries and regions in the primary wave with non-zero survey weights and available data on at least one risk factor. Prevalence varied substantially across settings for most dementia risk factors (table). The prevalence of low formal education showed large variability, ranging from 85·6% (95% CI 84·8–86·5) in CHARLS (China) to 12·0% (95% CI 11·3–12·7) in HRS (the USA). Although low education was most prevalent in LMICs, other risk factors showed different patterns. For example, the prevalence of obesity was the highest in HRS (the USA; 44·9%, 95% CI 43·3–46·5), followed by MHAS (Mexico; 35·7%, 95% CI 31·5–39·9), and the lowest in CHARLS (China; 14·5%, 95% CI 13·7–15·3), LASI (India; 13·3%, 95% CI 12·9–13·7), and KLoSA (South Korea; 6·2%, 95% CI 5·6–6·8). Cardiovascular disease risk factors were more prevalent in HICs than in LMICs. The prevalence of cardiovascular disease risk factors, such as hypertension, was also higher in the two Latin American countries than in other countries with similar socioeconomic status. For example, the prevalence of hypertension was 53·1% (95% CI 51·5–54·7) in MHAS (Mexico) and 49·0% (95% CI 47·7–50·3) in ELSI (Brazil), compared with 40·2% (95% CI 39·1–41·2) in CHARLS (China) and 43·4% (95% CI 41·8–44·9) in MARS (Malaysia).

Table:

Weighted prevalence of 12 risk factors for dementia across 14 geographical countries and regions

TILDA HRS SHARE
W		 ELSA SHARE N NICOLA KLoSA SHARE S SHARE E MHAS CHARLS MARS ELSI LASI Cross-national
challenges*
Ireland USA Western
Europe		 England Northern
Europe		 Northern
Ireland		 South
Korea		 Southern
Europe		 Eastern
Europe		 Mexico China Malaysia Brazil India
Low education 63·4% 12·0% 24·6% 20·4% 26·4% 62·5% 47·7% 69·5% 34·0% 84·0% 85·6% 70·0% 76·4% 76·7% None identified
Hearing loss 15·2% 19·4% 20·4% 21·0% 17·7% 24·0% 7·1% 20·8% 23·0% 33·3% 14·9% 17·0% 26·4% 7·4% Low awareness of risk factor
High LDL cholesterol 37·6% 53·1 19·9% 41·1% 17·6% 36·4% ·· 27·1% 21·6% ·· 23·0% 25·2% 23·0% 2·9% Low awareness of risk factor
Depression 9·8% 19·8% 28·1% 25·1% 19·6% 14·1% 23·5% 32·2% 37·6% 30·7% 37·2% ·· 26·0% 12·9% Interpretation
Physical inactivity ·· 27·8% ·· 24·5% ·· 43·0% ·· ·· ·· ·· 15·4% 31·9% 52·3% ·· Interpretation
Smoking 57·1% 53·8% 58·9% 53·0% 54·3% 51·6% 33·4% 40·5% 53·2% 37·3% 44·0% 25·8% None identified 19·8% None identified
Diabetes 8·0% 22·6% 14·5% 12·9% 11·5% 11·3% 15·9% 17·6% 16·8% 26·1% 13·4% 23·0% 17·7% 13·1% Low awareness of risk factor
Hypertension 37·8% 55·0% 46·7% 41·8% 44·1% 40·6% 36·6% 50·1% 54·3% 53·1% 40·2% 43·4% 49·0% 28·5% Low awareness of risk factor
Obesity 35·3% 44·9% 21·3% 32·2% 18·2% 35·5% 6·2% 18·2% 29·4% 35·7% 14·5% 36·1% 30·7% 13·3% None identified
Alcohol consumption 13·6% 8·0% 12·4% 16·8% 10·5% ·· 13·8% 7·9% 5·4% 3·7% ·· 0·6% 4·1% ·· None identified
Social isolation 22·9% 22·3% 29·7% 22·6% 33·4% 26·1% 13·1% 23·2% 23·2% 10·8% 10·0% 4·9% 21·2% 3·9% None identified
Vision loss 11·0% 21·0% 20·2% 15·7% 16·4% 15·4% 17·8% 31·4% 38·8% 47·2% 30·4% 30·8% 49·6% 69·7% Low awareness of risk factor
Open in a new tab

All data are in percentages. The regions are ordered by per capita gross domestic product (high->low) from left to right. CHARLS=China Health and Retirement Longitudinal Study. E=East. ELSA=English Longitudinal Study of Ageing. ELSI=Brazilian Longitudinal Study of Ageing. HRS=Health and Retirement Study. KLoSA=Korean Longitudinal Study of Ageing. LASI=Longitudinal Ageing Study in India. MARS=Malaysia Ageing and Retirement Survey. MHAS=Mexican Health and Aging Study. N=North. NICOLA=Northern Ireland Cohort for the Longitudinal Study of Ageing. S=South. SHARE=Survey of Health, Ageing and Retirement in Europe. TILDA=The Irish Longitudinal Study on Ageing. W=West.

*

Cross-national challenges identify potential methodological or measurement reasons that could explain observed differences across countries or regions.

Examining the rank order of risk factor prevalence across countries further emphasised cross-national patterns (figure 1; appendix pp 30-31). Low education (first rank in nine of 14 countries or regions), hypertension (rank ≤4 in all countries or regions), and smoking (rank ≤4 in 12 of 14 countries or regions) were among the most prevalent risk factors in most settings. Nevertheless, low education was less prevalent in HICs, whereas smoking showed the opposite pattern. Social isolation, high cholesterol, and obesity had higher ranks in HICs, whereas poor vision had a higher rank in many LMICs. There were a few notable exceptions to the general pattern of contrasting risk factor prevalence in HICs and LMICs. For example, excessive alcohol consumption was the seventh-ranked risk in KLoSA (South Korea) and the eighth-ranked risk in TILDA (Ireland), whereas it ranked lower (ranging from 10th to 12th ranking) in all other contexts. In addition to excessive alcohol consumption, social isolation (in LMICs) and diabetes were among the lowest ranked factors in many countries.

Figure 1: Heatmap of risk factor prevalence across countries and regions.

Figure 1:

Open in a new tab

The overall ordering of risk factors is based on their average prevalence across settings. Colours represent within-country ranking, scaled to the number of risk factors available in each country. For this study, we considered education below the level of secondary education as low education. Physical activity was quantified using the Leisure Score Index. CHARLS=China Health and Retirement Longitudinal Study. E=East. ELSA=English Longitudinal Study of Ageing. ELSI=Brazilian Longitudinal Study of Ageing. HRS=Health and Retirement Study. KLoSA=Korean Longitudinal Study of Ageing. LASI=Longitudinal Ageing Study in India. MARS=Malaysia Ageing and Retirement Survey. MHAS=Mexican Health and Aging Study. N=North. NICOLA=Northern Ireland Cohort for the Longitudinal Study of Ageing. S=South. SHARE=Survey of Health, Ageing and Retirement in Europe. TILDA=The Irish Longitudinal Study on Ageing. W=West.

Patterns by demographic variables showed that adults aged 70 years and older were more likely to have most risk factors across settings, except for smoking, obesity, and excessive alcohol consumption (figure 2; appendix pp 32-34). Additionally, age differences in low education were greater in HICs than in LMICs. For example, older adults in England were 157% (95% CI 122–198) more likely to have low education, whereas older adults in Malaysia were 30% (95% CI 25–35) more likely to have low education. Cardiovascular disease risk factors were more common in men than in women in HICs but more common in women than men in LMICs. For example, women were 35% (95% CI 24–44) less likely to have diabetes than men in TILDA (Ireland) but 21% (95% CI 7–36) more likely to have diabetes than men in ELSI (Brazil). The gender difference in poor hearing, with higher prevalence in men, was also greater in HICs. In contrast, behavioural risk factors (smoking and alcohol consumption) were more common in men only in LMICs. Across almost all (13 of 14) countries and regions, symptoms of depression and social isolation were more common in women. Many risk factors were more common in those with low education, with significantly higher prevalence among those with low education for at least half of the countries or regions with data for eight of 11 of included risk factors. Exceptions in some countries included cardiovascular risk factors, namely high cholesterol, diabetes, hypertension, and obesity, which were more common in those with high education in LASI (India) and CHARLS (China; only for cholesterol and diabetes). Physical inactivity co-occurred with low education in HICs, but this pattern was less apparent in LMICs.

Figure 2: Comparative patterns of risk factors by (A) age, (B) gender, and (C) education across countries and regions.

Figure 2:

Open in a new tab

Risk ratios were estimated using country-specific and risk factor-specific Poisson models with robust standard errors. Models were mutually adjusted for all three factors. For this study, we considered education below the level of secondary education as low education. Physical activity was quantified using the Leisure Score Index. CHARLS=China Health and Retirement Longitudinal Study. E=East. ELSA=English Longitudinal Study of Ageing. ELSI=Brazilian Longitudinal Study of Ageing. HRS=Health and Retirement Study. KLoSA=Korean Longitudinal Study of Ageing. LASI=Longitudinal Ageing Study in India. MARS=Malaysia Ageing and Retirement Survey. MHAS=Mexican Health and Aging Study. N=North. NICOLA=Northern Ireland Cohort for the Longitudinal Study of Ageing. S=South. SHARE=Survey of Health, Ageing and Retirement in Europe. TILDA=The Irish Longitudinal Study on Ageing. W=West.

Education gradients differed by gender for many risk factors (appendix p 35). For example, education gradients were larger in women for several cardiovascular disease risk factors, including diabetes, hypertension, and obesity, across most countries and regions.

Risk factors frequently co-occurred across all settings, with the prevalence of at least two risk factors exceeding 50% in all countries and regions (figure 3). In most (11 of 14) countries and regions, the prevalence of at least four risk factors also exceeded 20%. Estimates of the average number of co-occurring risk factors of the total number of risk factors measured (ie, a proportion, scaled from 0 to 1, representing the burden of co-occurring risk factors) for each country or region showed the highest levels of co-occurring risk factors in the two Latin American studies (MHAS [35%, 95% CI 34–36; Mexico] and ELSI [34%, 95% CI 33–34; Brazil]), followed by CHARLS (31%, 95% CI 31–31; China), HRS (29%, 95% CI 29–30; the USA), and NICOLA (29%; 95% CI 29–30; Northern Ireland). Risk factor clusters showed similarities across contexts (figure 4). A cardiovascular disease cluster, including diabetes, high cholesterol, hypertension, and in some cases obesity, was present across all contexts. Smoking and alcohol use also frequently co-occurred in the same cluster; poor hearing, poor vision, and low education co-occurred frequently as well.

Figure 3: Patterns of co-occurring risk factors across countries and regions.

Figure 3:

Open in a new tab

(A) The proportion of individuals with two or more risk factors. (B) The average proportion of risk factors present across respondents in each country or region after accounting for differences in the number of available risk factors across studies. CHARLS=China Health and Retirement Longitudinal Study. E=East. ELSA=English Longitudinal Study of Ageing. ELSI=Brazilian Longitudinal Study of Ageing. HRS=Health and Retirement Study. KLoSA=Korean Longitudinal Study of Ageing.

LASI=Longitudinal Ageing Study in India. MARS=Malaysia Ageing and Retirement Survey. MHAS=Mexican Health and Aging Study. N=North. NICOLA=Northern Ireland Cohort for the Longitudinal Study of Ageing. RF=risk factor. S=South. SHARE=Survey of Health, Ageing and Retirement in Europe. TILDA=The Irish Longitudinal Study on Ageing. W=West.

Figure 4: Clusters of risk factors across countries and regions.

Figure 4:

Open in a new tab

Parallel analysis was used to identify the number of clusters, and principal component analysis was used with modal assignment to assign risk factors to clusters based on the highest factor loadings. Values represent standardised loadings of risk factors (scale range 0–1). Higher values indicate stronger associations of the risk factors with the cluster. Mix 1 and Mix 2 denote heterogeneous clusters that varied across countries and represent less common, inconsistent combinations of risk factors. For this study, we considered education below the level of secondary education as low education. Physical activity was quantified using the Leisure Score Index. CHARLS=China Health and Retirement Longitudinal Study. E=East. ELSA=English Longitudinal Study of Ageing. ELSI=Brazilian Longitudinal Study of Ageing. HRS=Health and Retirement Study. KLoSA=Korean Longitudinal Study of Ageing. LASI=Longitudinal Ageing Study in India. MARS=Malaysia Ageing and Retirement Survey.

MHAS=Mexican Health and Aging Study. N=North. NICOLA=Northern Ireland Cohort for the Longitudinal Study of Ageing. S=South. SHARE=Survey of Health, Ageing and Retirement in Europe. TILDA=The Irish Longitudinal Study on Ageing. W=West.

Sensitivity analyses showed that the observed relative cross-country comparisons in primary analyses were largely similar across alternative specifications of binary or continuous measures of risk factors (appendix pp 18-29). Hypertension prevalence based on measured blood pressure or self-reported medication use was consistently higher than prevalence based on self-reported physician diagnosis; the difference ranged from 19% higher in CHARLS (China) to 71% higher in NICOLA (Northern Ireland). However, the rank ordering of hypertension prevalence across countries remained more stable across different hypertension measurement methods, with an average absolute rank difference of 1·33 (appendix pp 18-29). No pattern was observed in the magnitude of self-report bias (the difference between self-reported and measured hypertension prevalence) by GDP; the Pearson correlation coefficient between log(GDP) and prevalence difference was 0·23 (p=0·55).

Discussion

Prevalence of the 12 selected dementia risk factors varied across settings, although there were some consistent patterns across HICs and LMICs. Low education (particularly in LMICs), hypertension, and smoking were among the most prevalent risk factors in most settings. Although patterns stratified by age, gender, and education showed substantial heterogeneity, similarities were observed among countries at similar levels of development. Despite cross-country differences in risk factor prevalence, risk factors co-occurred and clustered together in a similar manner. These patterns can inform strategic design choices that leverage commonality across countries to design intervention components or features meant to be shared across settings, while also highlighting areas where the customisation of interventions or policies aimed at dementia prevention are needed.

Low education was a priority risk factor in this analysis and in previous analyses that focused on PAFs.2,3,6 However, for risk factors other than low education, the prevalence rankings observed in this study differed from those based on PAFs, as previous studies considered both risk factor prevalence (the only focus of our study) and each risk factor’s effect on dementia. For example, although many analyses reported a high PAF for hearing loss,3,6 a low rank was observed for hearing loss in our study when examining prevalence across countries; this low rank might be partly attributable to our focus on prevalence rather than PAFs. Although PAFs provide a more appropriate metric for ranking overall importance because they consider both prevalence and effect size, comparisons of risk factor prevalence and patterns provide important information for planning and designing interventions or policies as they shape potential population-level impact and intervention feasibility. Additionally, our focus on prevalence enabled the inclusion of a broader range of countries without requiring data on dementia or assumptions of effect sizes being consistent across settings.

Including more wide-ranging country contexts than in previous cross-country analyses of dementia risk factors6 makes it easier to identify differences and similarities across countries and patterns by development levels. Similarities within HICs and LMICs in the prevalence and ranking of different risk factors suggest that insights from countries with similar development levels might be informative when designing interventions for dementia risk factors in settings with no empirical data. For example, low education is often the most prevalent risk factor in LMICs and might also influence other downstream risk factors,23,24 emphasising the importance of addressing low education in LMIC settings and other similar contexts. Estimates can also be compared with previous studies collating risk factor data across countries (eg, for the Western Pacific region),25 although other studies might provide poor-quality estimates based on models with sparse input data. New efforts to expand the availability of harmonised data sources are needed to improve data quality.

Methodological reasons for patterns should also be considered. Although cardiovascular disease risk factors were more prevalent in HICs, observed gradients were likely influenced by low or poor awareness of risk factors in LMICs or poor health-care access given surveys collected data on physician diagnoses. The low rank of diabetes prevalence compared to other risk factor prevalence across most settings might also be partly attributed to poor awareness of diabetes status. In addition to identifying patterns across countries, exceptions and potential outliers should be considered. For example, although obesity ranked high (between fourth and sixth) in most HICs, its prevalence was only 6·2% in South Korea. An understanding of country-specific contexts, including the social, political, and community structures in place, can help to identify outliers of broader patterns by development level and support tailoring of policies and interventions using frameworks for intervention adaptation.26

We also observed differences between HICs and LMICs in the patterns of risk factors by age, gender, and education. For example, women were less likely than men to engage in risky health behaviours such as smoking or alcohol use in LMICs, whereas this gender difference was smaller in HICs. Where evidence indicated substantial gender differences, findings can inform gender-specific intervention strategies and maximise the effects of available resources. Additionally, individuals with low education were more likely to have low physical activity in HICs, whereas this association was weaker in LMICs.

Patterns by gender and education highlight the ways in which exposure to dementia risk factors is shaped by broader social and societal contexts rather than solely by individual-level decision-making. Facets of identity and social or economic status (eg, age, gender, and education) influence an individual’s interactions with employers, community and government services, and health-care systems, which in turn affect the social, economic, and health factors associated with dementia risk. These patterns also emphasise how interventions that target poverty or broader sociodemographic inequalities might have sizeable downstream effects on late-life health outcomes such as dementia.27

Findings on the co-occurrence and clustering of dementia risk factors align with those from a previous analysis done in Denmark, which reported that 82% of respondents had more than one of 16 risk factors (in our study, this value is 62–93% across 10–12 risk factors).10 Although our data suggest that the average number of co-occurring risk factors varied across countries and regions (from 21% in Korea to 35% in Mexico), findings generally emphasised the importance of a holistic approach to dementia prevention across the range of established risk factors. The low effectiveness of many single-factor interventions might partly reflect failure to address multiple co-occurring risk factors.28 Risk factors might co-occur for many reasons, including biological links via mechanistic pathways (eg, cardiovascular disease risk factors), shared upstream determinants (eg, risky behaviours or socioeconomic factors), or mediating mechanisms (eg, education influencing access to correction for sensory impairment). In all cases, the co-occurrence and clustering of risk factors could guide efforts to target multiple risk factors that commonly cluster together; however, the reason behind clustering could influence design choices. For example, if multiple risk factors lie on the same biological pathway, targeting the most upstream factor might be most effective. Although multidomain interventions that include various components, such as personalised guidance on diet and physical activity, cognitive training, and increased social engagement, are highly effective,13,14,29 the effects of such interventions might not persist over the long term.30 Cluster-tailored multidomain trials would help to evaluate whether cluster-informed intervention designs can improve intervention response.

Strengths of this study include the large number of countries and regions, the inclusion of population-based samples to ensure generalisability of findings, the use of harmonised data to maximise comparability between contexts, and the in-depth analysis of risk factor prevalence and patterns. Limitations of the study include the analysis being restricted to the dementia risk factors reported in the Lancet Commission report; although these risk factors were selected based on evidence-based approaches with sufficient epidemiological support,3 future updates should consider additional risk factors as evidence emerges. We also used binary versions of all risk factors to simplify the interpretation of findings and comparisons, although, continuous measures (eg, pack-years of smoking) can provide more detailed information if available. Furthermore, despite efforts to maximise harmonisation during study design, data preparation, and data analysis, inconsistency remained. For example, survey questions on alcohol use relied on different response options across settings, and available data from the LASI survey in India did not include comparable alcohol consumption-related questions. In some cases, inconsistency reflects necessary cultural adaptations to ensure that questions are interpreted correctly and remain relevant to populations of interest. In other cases, despite the use of identical survey items, factors related to study context might influence comparisons. For example, although all surveys included questions on physician diagnosis of diabetes or hypertension, differences in health-care access and disease awareness across countries complicate comparisons. We compared self-reported and objective measures of hypertension to assess this bias, given the availability of objective measures across many surveys. Although we did not observe a strong association of self-report bias and GDP for hypertension, previous single-setting studies have reported higher self-report bias in groups with lower socioeconomic status, raising concerns of differential bias.31-33 Similar bias patterns might occur for diabetes as well.32 More standardised implementation of objective biomarker measurement protocols across additional surveys and conditions would help to improve such measurements. Self-report bias in other risk factors and survival bias should also be considered, especially given that self-reports of life course risk factors are based on cross-sectional data of respondents in mid to late life. Moreover, efforts to harmonise data across studies also led to the oversimplification of some risk factor definitions in our study. For example, we used living alone as an indicator of social isolation, simplifying a complex construct to enable harmonised measurement, and we acknowledge that cultural differences in multi-generational living could alter the interpretation of this indicator despite consistent measurement. Although rates of missing data were generally low, with instances of higher missing data explained by study design features, potential bias can arise from excluding respondents with missing data. Additionally, despite our efforts to balance the competing priorities of using the most up-to-date information with representativeness by selecting the most recent wave with a refresher sample, some data were outdated (eg TILDA in Ireland; 2009–10). Regular refresher sampling would ensure the availability of timely, population-representative data and enhance the utility of comparative analyses such as those in our study; however, such efforts require substantial investment. Future iterations of our study should include emerging data on additional risk factors (eg, sleep34) and new country contexts and should consider examining temporal changes in risk factors.

Our findings provide important insights into differences in the prevalence and patterns of dementia risk factors across contexts and can inform the optimal design of interventions and policies to address modifiable dementia risk factors. Although the development of national dementia plans and individual-level recommendations and guidelines is a useful initial step, such approaches often do not result in behavioural changes in individuals.35,36 Population-level approaches focused on implementing policy changes might be more effective and produce substantial cost savings for health systems.37,38 Although differences across contexts likely preclude a one-size-fits-all approach, similarities by development level suggest that policy makers can draw on evidence from countries with comparable contexts to inform adaptation. Moreover, areas of commonality, including the high levels of co-occurring risk factors and similar risk factor clusters, can be leveraged as components of a universal design strategy that can be adapted to specific contexts.26 The availability of harmonised data, such as those analysed here, is essential for generating cross-national insights and supporting the development of context-specific solutions in participating countries. Thus, our findings highlight opportunities to ensure that efforts to prevent or delay dementia are based on high-quality quantitative evidence, leading to effective interventions with maximum benefits.

Supplementary Material

1

Research in context.

Evidence before this study

We searched PubMed on Dec 15, 2025, for studies on dementia risk factors using the terms “dementia”, “risk factor”, and (“global” or “international” or “cross-country”), without language restrictions, and also screened the reference lists of relevant articles to identify additional eligible studies. Few studies presented primary data on multiple risk factors across diverse contexts. A previous study reported population attributable fractions (PAFs) for dementia in Latin America, India, and China, but data were restricted to small geographical areas within these regions, and analyses focused on the overall PAF measure rather than examining patterns of risk factor prevalence in detail.

Added value of this study

Our study extends previous small-scale efforts by analysing harmonised data (N=214 251) from population-based studies across 14 countries and regions to explore patterns of risk factor prevalence in greater detail. Using these unique data, we assessed the prevalence and rank order of risk factors, demographic patterns, and risk factor co-occurrence and clustering in diverse settings. Owing to the large number of countries and regions included, we improved on previous efforts by performing cross-country comparisons and identifying both common patterns and context-specific differences in dementia risk factors across development levels.

Implications of all the available evidence

The findings highlight the value of generating cross-country comparative evidence, while also presenting the associated challenges. Similarities across settings with similar development levels can be leveraged to guide the adaptation of existing interventions when country-specific data do not exist. Moreover, consistency in patterns of risk factor co-occurrence can inform universal design frameworks that leverage elements of a common structure while ensuring flexibility and adaptability across diverse settings. Although a one-size-fits-all approach would not succeed, general patterns, combined with country-specific data (when available), can be used to ensure that the design of dementia policies and interventions is based on high-quality quantitative evidence. Future research should prioritise improving the availability of harmonised, high-quality data across diverse populations and incorporate standardised measurement approaches to strengthen cross-country comparisons.

Acknowledgments

Data harmonisation for the Gateway to Global Ageing Data project was supported by the National Institutes of Health (NIH) and National Institute on Aging (NIA; grant number R01AG030153). The funding sources for the different longitudinal studies are listed below.

The Irish Longitudinal Study on Ageing (TILDA): TILDA waves 5 and 6 were funded by the Department of Health, Government of Ireland, under the administration of Health Research Board (TILDA-2017–1) and the Atlantic Philanthropies. Researchers interested in using TILDA data can access the data for free from Irish Social Science Data Archive (ISSDA) at University College Dublin and Interuniversity Consortium for Political and Social Research (ICPSR) at the University of Michigan.

Health and Retirement Study (HRS): HRS is sponsored by NIA (grant number NIA U01AG009740) and conducted by the University of Michigan. Survey of Health, Ageing and Retirement in Europe (SHARE): This paper uses data from SHARE. The SHARE data collection was funded by the European Commission, Directorate-General for Research and Innovation (DG RTD) through Framework Programme (FP) 5 (grant number QLK6-CT-2001-00360), FP6 (SHARE-I3, grant number RII-CT-2006-062193; COMPARE, grant number CIT5-CT-2005-028857; and SHARELIFE, grant number CIT4-CT-2006-028812), FP7 (SHARE-PREP, grant number GA N°211909; SHARE-LEAP, grant number GA N°227822; SHARE M4, grant number GA N°261982; and DASISH, grant number GA N°283646), and Horizon 2020 (SHARE-DEV3, grant number GA N°676536; SHARE-COHESION, grant number GA N°870628; SERISS, grant number GA N°654221; SSHOC, grant number GA N°823782; and SHARE-COVID19, grant number GA N°101015924`). Additional funding was provided by DG Employment; Social Affairs and Inclusion (grant numbers VS 2015/0195, VS 2016/0135, VS 2018/0285, VS 2019/0332, and VS 2020/0313); SHARE-EUCOV (grant number GA N°101052589); EUCOVII (grant number GA N°101102412); German Federal Ministry of Research, Technology and Space (grant numbers 01UW1301, 01UW1801, and 01UW2202); the Max Planck Society for the Advancement of Science; the US NIA (grant numbers U01_AG09740-13S2, P01_AG005842, P01_AG08291, P30_AG12815, R21_AG025169, Y1-AG-4553-01, IAG_BSR06-11, OGHA_04-064, BSR12-04, R01_AG052527-02, R01_AG056329-02, R01_AG063944, HHSN271201300071C, and RAG052527A); and various other national funding sources.

English Longitudinal Study of Ageing (ELSA): ELSA is funded by the NIA (grant number R01AG017644) and departments of UK Government coordinated by the National Institute for Health and Care Research (NIHR). Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA): The Northern Ireland Cancer Registry is funded by the Public Health Agency (PHA). NICOLA uses data provided by patients and collected by the Health Service as part of their care and support. We thank all participants of NICOLA and the NICOLA team consisiting of nursing staff, research scientists, clerical staff, computer and laboratory technicians, managers, and receptionists.

Korean Longitudinal Study of Ageing (KLoSA): We thank the Korea Employment Information Service (KEIS) for conducting and sharing the KLoSA data. KLoSA was supported by the Occupational Safety and Health Research Institute, South Korea.

Mexican Health and Aging (MHAS): MHAS acknowledges infrastructure support from the Sealy Center on Aging and the WHO–Pan American Health Organization (PAHO) Collaborating Center on Aging and Health, University of Texas Medical Branch (UTMB). We acknowledge the participation of the Instituto Nacional de Estadística y Geografía (INEGI), Mexico, and the Instituto Nacional de Salud Pública (INSP), Mexico, in data collection and the University of Wisconsin and the Instituto Nacional de Geriatría (INGer), Mexico, in the conceptual design of the study.

China Health and Retirement Longitudinal Study (CHARLS): CHARLS is supported by the Behavioural and Social Research division of the NIA, the Natural Science Foundation of China, the World Bank, and Peking University for financial support. We thank the CHARLS research and field team and every respondent in the study for their contributions.

Malaysia Ageing and Retirement Survey (MARS): MARS is conducted by the Social Wellbeing Research Center (SWRC) and funded by the Employees Provident Fund, Malaysia.

Brazilian Longitudinal Study of Ageing (ELSI): The ELSI-Brazil baseline study was funded by the Department of Science and Technology from the Secretariat of Science, Technology and Strategic Inputs (DECIT/SCTIE), the Brazilian Ministry of Health (grant number 404965/2012-1); COSAPI/DAPES/SAS—Healthcare Coordination of Elderly, Department of Strategic and Programmatic Actions from the Secretariat of Health Care (grant numbers 20836, 22566, and 23700); and the Brazilian Ministry of Science, Technology, Innovation and Communication. We thank the study participants and the consultant team for their contribution. The list of consultants is available at the ELSI-Brazil’s homepage.

Longitudinal Ageing Study in India (LASI): LASI is jointly funded by the Ministry of Health and Family Welfare, Government of India, and NIA/NIH, USA (grant number R01 AG042778). The UN Population Fund–India office provided additional financial support during the preparatory phase of LASI wave 1.

During the preparation of this work the authors used Codex in order to update and format analytical code. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.

Footnotes

Declaration of interests

We declare no competing interests.

Editorial note: The Lancet Group takes a neutral position with respect to territorial claims in published maps and institutional affiliations.

See Online for appendix

For more on Gateway to Global Ageing Data, see https://www.g2aging.org

For more details, see www.share-eric.eu

For more on ELSI-Brazil, see http://elsi.cpqrr.fiocruz.br/

Data sharing

Harmonised data from the Gateway to Global Ageing Data platform are available on study-specific websites. Data access across surveys typically requires users to register, provide an email address, and sign a data-use agreement. Additional details and instructions on accessing data from each survey are available at Gateway to Global Ageing Data platform.

References

  • 1.Nichols E, Steinmetz JD, Vollset SE, et al. Estimation of the prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the Global Burden of Disease Study 2019. Lancet Public Health 2022; 7: e105–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Stephan BCM, Cochrane L, Kafadar AH, et al. Population fractions of modifiable risk factors for dementia: a systematic review and meta-analysis. Lancet Healthy Longev 2024; 5: e406–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Livingston G, Huntley J, Liu KY, et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. Lancet 2024; 404: 572–628. [DOI] [PubMed] [Google Scholar]
  • 4.Kobayashi LC, Jones RN, Briceño EM, et al. Cross-national comparisons of later-life cognitive function using data from the Harmonized Cognitive Assessment Protocol (HCAP): considerations and recommended best practices. Alzheimers 2024; 20: 2273–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Glymour MM, Whitmer RA. Using cross-cultural studies to improve evidence on dementia prevention: lessons from the special issue sponsored by the International Research Network on Dementia Prevention (IRNDP). J Alzheimers Dis 2019; 70 (suppl 1): S5–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Mukadam N, Sommerlad A, Huntley J, Livingston G. Population attributable fractions for risk factors for dementia in low-income and middle-income countries: an analysis using cross-sectional survey data. Lancet Glob Health 2019; 7: e596–603. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Angrisani M, Nichols E, Meijer E, et al. Modifiable risk factors for dementia in India: a cross-sectional study revisiting estimates and reassessing prevention potential and priorities. BMJ Public Health 2024; 2: e001362. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Vergara RC, Zitko P, Slachevsky A, San Martin C, Delgado C. Population attributable fraction of modifiable risk factors for dementia in Chile. Alzheimers Dement (Amst) 2022; 14: e12273. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Suemoto CK, Borelli WV, Calandri IL, et al. The potential for dementia prevention in Brazil: a population attributable fraction calculation for 14 modifiable risk factors. Lancet Reg Health Am 2025; 49: 101209. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Janbek J, Laursen TM, Jørgensen K, et al. Prevalence and co-occurrence of dementia risk factors in Denmark: a nationwide study. J Prev Alzheimers Dis 2025; 12: 100365. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Meader N, King K, Moe-Byrne T, et al. A systematic review on the clustering and co-occurrence of multiple risk behaviours. BMC Public Health 2016; 16: 657. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Peters R, Booth A, Rockwood K, Peters J, D’Este C, Anstey KJ. Combining modifiable risk factors and risk of dementia: a systematic review and meta-analysis. BMJ Open 2019; 9: e022846. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Ngandu T, Lehtisalo J, Solomon A, et al. A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): a randomised controlled trial. Lancet 2015; 385: 2255–63. [DOI] [PubMed] [Google Scholar]
  • 14.Kivipelto M, Mangialasche F, Snyder HM, et al. World-Wide FINGERS Network: a global approach to risk reduction and prevention of dementia. Alzheimers Dement 2020; 16: 1078–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Lee J, Phillips D, Wilkens J, Gateway to Global Aging Data Team. Gateway to Global Aging Data: resources for cross-national comparisons of family, social environment, and healthy aging. J Gerontol B Psychol Sci Soc Sci 2021; 76 (suppl 1): S5–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.UN Statistics Division. Composition of macro geographical (continental) regions, geographical sub-regions, and selected economic and other groupings (M49). https://unstats.un.org/unsd/methodology/m49/ (accessed Feb 25, 2026).
  • 17.UNESCO Institute for Statistics. International standard classification of education ISCED 2011. 2012. isced-2011-en.pdf (accessed Feb 20, 2026).
  • 18.Godin G. Godin leisure-time exercise questionnaire. Med Sci Sports Exerc 1997; 29: 36–38. [Google Scholar]
  • 19.Centers for Disease Control and Prevention (USA). WHO, What is excessive alcohol use? June 26. 2014. https://stacks.cdc.gov/view/cdc/30624 (accessed Feb 20, 2026).
  • 20.Radloff LS. The CES-D Scale: a self-report depression scale for research in the general population. Appl Psychol Meas 1977; 1: 385–401. [Google Scholar]
  • 21.Guerra M, Ferri C, Llibre J, Prina AM, Prince M. Psychometric properties of EURO-D, a geriatric depression scale: a cross-cultural validation study. BMC Psychiatry 2015; 15: 12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.WHO Expert Consultation. Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies. Lancet 2004; 363: 157–63. [DOI] [PubMed] [Google Scholar]
  • 23.World Bank. GDP (current US$). https://data.worldbank.org/indicator/NY.GDP.MKTP.CD (accessed Feb 25, 2026).
  • 24.Office for National Statistics. Regional economic activity by gross domestic product, UK: 1998 to 2023. April 17, 2025. https://www.ons.gov.uk/economy/grossdomesticproductgdp/bulletins/regionaleconomicactivitybygrossdomesticproductuk/1998to2023#:~:text=Of%20the%20ITL1%20regions%2C%20Northern%20Ireland%20experienced,at%20%C2%A369%2C077%2C%20while%20the%20North%20East%20had (accessed Feb 25, 2026).
  • 25.Clarke AJ, Brodtmann A, Irish M, et al. Risk factors for the neurodegenerative dementias in the Western Pacific region. Lancet Reg Health West Pac 2024; 50: 101051. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Moore G, Campbell M, Copeland L, et al. Adapting interventions to new contexts-the ADAPT guidance. BMJ 2021; 374: n1679. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Walsh S, Merrick R, Brayne C. The relevance of social and commercial determinants for neurological health. Lancet Neurol 2022; 21: 1151–60. [DOI] [PubMed] [Google Scholar]
  • 28.Williams JW, Plassman BL, Burke J, Benjamin S. Preventing Alzheimer’s disease and cognitive decline. Evid Rep Technol Assess (Full Rep) 2010: 1–727. [PMC free article] [PubMed] [Google Scholar]
  • 29.Baker LD, Espeland MA, Whitmer RA, et al. Structured vs self-guided multidomain lifestyle interventions for global cognitive function: the US Pointer randomized clinical trial. JAMA 2025; 334: 681–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Ngandu T, Solomon A, Lehtisalo J, et al. Long-term adherence to lifestyle changes and association with cognitive change: 11-year results from the FINGER randomized, controlled trial. Alzheimer’s Dement 2025; 21: e106542. [Google Scholar]
  • 31.Nichols E, Hu P, Bloom DE, Lee J, Sekher TV. The impact of using self-report versus objective measures of cardiometabolic conditions in epidemiologic research: a case study from India using data from the longitudinal aging study in India. Front Epidemiol 2025; 1372972. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Ning M, Zhang Q, Yang M. Comparison of self-reported and biomedical data on hypertension and diabetes: findings from the China Health and Retirement Longitudinal Study (CHARLS). BMJ Open 2016; 6: e009836. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Mulyanto J, Kringos DS, Kunst AE. The accuracy of self-report versus objective assessment for estimating socioeconomic inequalities in disease prevalence in Indonesia. Int J Public Health 2019; 64: 1233–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Xu W, Tan CC, Zou JJ, Cao XP, Tan L. Sleep problems and risk of all-cause cognitive decline or dementia: an updated systematic review and meta-analysis. J Neurol Neurosurg Psychiatry 2020; 236–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Marteau TM, Rutter H, Marmot M. Changing behaviour: an essential component of tackling health inequalities. BMJ 2021; 372: n332. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Adams J, Mytton O, White M, Monsivais P. Why are some population interventions for diet and obesity more equitable and effective than others? The role of individual agency. PLoS Med 2016; 13: e1001990. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Mukadam N, Anderson R, Walsh S, et al. Benefits of population-level interventions for dementia risk factors: an economic modelling study for England. Lancet Healthy Longev 2024; 5: 100611. [DOI] [PubMed] [Google Scholar]
  • 38.Mukadam N, Anderson R, Knapp M, et al. Effective interventions for potentially modifiable risk factors for late-onset dementia: a costs and cost-effectiveness modelling study. Lancet Healthy Longev 2020; 1: e13–20. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS2198887-supplement-1.pdf (969.3KB, pdf)

Data Availability Statement

Harmonised data from the Gateway to Global Ageing Data platform are available on study-specific websites. Data access across surveys typically requires users to register, provide an email address, and sign a data-use agreement. Additional details and instructions on accessing data from each survey are available at Gateway to Global Ageing Data platform.

RESOURCES