Heatwave and dementia: a systematic analysis for the Global Burden of Disease Study 2019 Authors: Yinhan Chen, Yanan Luo, Le Yu, Xiaoying Zheng, Qiang Luo Published: 2025 ABSTRACT AbstractIntroductionClimate change has increased extreme weather conditions, such as heatwave, but its effects on dementia have not been studied globally. This study aims to assess the global association between heatwave and dementia incidence and to examine its heterogeneity across different countries.MethodsGridded daily maximum temperature from the European Centre for Medium-Range Weather Forecasts Reanalysis V.5 database and the age-standardised incidence of dementia from the Global Burden of Disease Study 2019 were derived for 153 countries from 1990 to 2019. The associations between heatwaves and annual growth in incidence of dementia were tested via mixed-effect models and two-stage meta-regressions. The relationships were further analysed in the high-income countries (HICs) and the low- to middle-income countries (LMICs) separately. Finally, supporting analyses at state level in the USA and individual level in the UK were conducted.ResultsWe found a rising trend of heatwaves but significant between-country variation in incidence of dementia. Heatwave was positively associated with the acceleration of within-country annual growth in incidence of dementia worldwide (0.0436 SD, 95% CI 0.0391 to 0.0481; p<0.0001). The association was significant in the HICs (0.0294, 95% CI 0.0178 to 0.041; p<0.0001), and in the LMICs (0.0482, 95% CI 0.0437 to 0.0528; p<0.0001). A better economic situation mitigated the association in the HICs but exacerbated it in the LMICs. These findings remained significant after sensitivity analyses and were consistent at the state and individual levels.ConclusionsThis finding underscores the global association between heatwave and incidence of dementia and its heterogeneity, indicating the importance of public policy to address the heatwave-associated incidence of dementia, especially in the LMICs. WHAT IS ALREADY KNOWN ON THIS TOPIC The global effects of heatwave on mortality and morbidity related to heart disease and stroke have been reported, but few studies have analysed such effects on dementia. The global effects of heatwave on mortality and morbidity related to heart disease and stroke have been reported, but few studies have analysed such effects on dementia. Until recently, research mainly focused on how heat worsens neurological disorders or the difficulties patients face in protecting themselves from heat. Emerging experimental and epidemiological evidence now suggests that heat stress may impair hippocampal neurogenesis, trigger neuroinflammation and increase the risk of cognitive decline and dementia incidence. Until recently, research mainly focused on how heat worsens neurological disorders or the difficulties patients face in protecting themselves from heat. Emerging experimental and epidemiological evidence now suggests that heat stress may impair hippocampal neurogenesis, trigger neuroinflammation and increase the risk of cognitive decline and dementia incidence. Previous studies in high-income countries have mainly examined the effects of heatwaves on mortality and hospitalisations in patients with dementia, while little is known about their impact on dementia incidence at a global scale. Previous studies in high-income countries have mainly examined the effects of heatwaves on mortality and hospitalisations in patients with dementia, while little is known about their impact on dementia incidence at a global scale. WHAT THIS STUDY ADDS This is the first global study of the association between heatwaves and dementia incidence across 153 countries over 30 years. This is the first global study of the association between heatwaves and dementia incidence across 153 countries over 30 years. A global association has been identified, particularly between cumulative exposure to heatwaves and the growth rate of incidence. More research is needed to further explore and validate this finding. A global association has been identified, particularly between cumulative exposure to heatwaves and the growth rate of incidence. More research is needed to further explore and validate this finding. The association was evident in both high-income and low-to-middle-income countries, with a stronger association observed in the latter. The association was evident in both high-income and low-to-middle-income countries, with a stronger association observed in the latter. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY These findings call for the attention of the policymakers and the public to be aware and to address the heatwave-associated global incidence of dementia, especially in low- to middle-income countries. These findings call for the attention of the policymakers and the public to be aware and to address the heatwave-associated global incidence of dementia, especially in low- to middle-income countries. Introduction Heatwave, a period of prolonged abnormally hot weather, has become more common during the past 30 years due to climate change.1 It has drawn significant public attention as exposure to heatwave has been associated with both excessive morbidity2 and mortality.3 Exposure to a heatwave can provoke maladaptive physiological responses to heat stress in both the human body and the brain.4 5 It has been reported in the literature that exposure to heat affects neurotransmitters by engaging in thermoregulation in animal brains, and heat is also associated with disruptions in functional neuroimaging features in the human brain.6 Emerging evidence also suggests that heat stress can trigger neuroinflammatory responses, impair hippocampal neurogenesis and lead to neuronal damage, thereby contributing to cognitive decline and increased risk of neurodegeneration.79 Dementia, as a condition of the brain, is one of the greatest burdens on human society with an estimated 152 million patients worldwide by 2050, especially in low- to middle-income countries (LMICs).10 Advancing knowledge of how heatwaves affect dementia across regions can inform region-specific prevention strategies targeted at vulnerable populations, guide the allocation of healthcare resources and support the development of climate resilience policies. However, little is known about the impact of heatwaves on the global incidence of dementia, primarily due to the lack of paired location‒clinical record data.11 Previous studies on the mortality and hospital admission of patients with dementia are limited to a few developed countries.12 13 The Global Burden of Disease (GBD) data set and high-resolution long-term climate satellite data have offered a valuable opportunity to explore the relationship between heatwave and dementia at a global scale. Recent studies have reported that annual heatwaves can be estimated for each country via satellite data to match annual and national health records, such as medical staffing levels14 and dengue cases.15 Therefore, we aimed to construct a database of satellite-derived heatwave and GBD-provided dementia incidence data that are paired at the annual and national levels for 30 years covering hundreds of countries. In previous studies examining different health outcomes (mortality and hospitalisations) in patients with dementia, the impact of heatwave varies across different countries, including both significant13 16 and non-significant17 findings. Many factors may contribute to this heterogeneity, such as the population structure, heatwave definition, possible lag effect and socioeconomic conditions,18 ethnicity19 and illness history (eg, heat-related illness).20 Thus, incorporating these factors into the model to investigate the heterogeneity in the impact of heatwave on dementia is necessary. This study aimed to evaluate the relationship between heatwave and the incidence of dementia at a global scale and the heterogeneity of this association across different countries. First, we constructed a database of heatwave and dementia incidence rates in 153 countries over 30 years from 1990 to 2019 using air temperature data from the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis V.5 (ERA5) satellite data set. Second, we tested the associations between the incidence of dementia and annual heatwave globally and separately in high-income countries (HICs) and LMICs. Third, we explored the heterogeneity of this association across different countries. Fourth, to provide more supporting evidence, we explored the association in specific, data-rich contexts with different strengths (eg, controlling for unmeasured country-level confounders in the US analysis, and controlling for individual-level confounders in the UK analysis). Materials and methods Data collection The study was conducted from July 2022 to September 2023. We obtained dementia data from the GBD 2019 via the Global Health Data Exchange. This database provides the age-standardised and age-specific prevalence and incidence of 369 diseases and injuries from 1 January 1990 to 31 December 2019.21 Estimation methods for dementia prevalence, incidence and mortality have been described in detail elsewhere.22 Drawing on a wide range of global data sets, including administrative records and published studies, the data were standardised to a single reference definition using meta-regression—Bayesian, regularised, trimmed (MR-BRT), an meta-regression (MR) framework developed for GBD 2019.23 DisMod-MR V.2.1, a Bayesian MR tool developed for GBD, was used to generate internally consistent models of prevalence, incidence, remission and excess mortality. DisMod-MR first fits a global model based on all available data, the resulting estimates are then used as Bayesian priors for seven GBD super-regions. This process is repeated hierarchically down to the country level, after which estimates are aggregated back up to the global level to produce the final incidence estimates. This Bayesian hierarchical approach allows the model to borrow strength from predictive covariates and data-rich regions to inform estimates in countries with limited or missing data. Here, we extracted age-standardised incidence rates of dementia across 204 countries and territories between 1990 and 2019. For sensitivity analysis, we also extracted the rate of deaths caused by dementia. Dementia (ie, Alzheimer’s disease and other dementias; International Classification of Diseases (ICD)-10 codes: F00-02.0, F02.8-F03.91, F06.2, G30-G31.1, G31.8-G32.89) was defined as a clinical diagnosis by a physician with or without confirmation via positron emission tomography imaging.22 We obtained the daily temperatures for these countries and territories via the ERA5 database, which is a gridded reanalysis data set produced by the ECMWF with a 1° × 1° spatial resolution on the Google Earth Engine.24 The daily mean, maximum and minimum temperature estimates were used for the period from 1 January 1990 to 31 December 2019. We also obtained the estimated numbers of people (per 30 arc-second grid cell) from the Gridded Population of World V.4 on Google Earth Engine,25 which is a raster data collection of globally integrated national population data from the 2010 round of Population and Housing Censuses, that occurred between 2005 and 2014. The World Bank provides the annual per capita gross domestic product (GDP)26 and population27 of countries around the world from 1960 to 2021. The countries were classified as HIC or LMIC according to the Gross National Income (GNI) provided by the World Bank.28 To test whether the contributions of heatwaves to the incidence of dementia could be simply explained by nuisance variables, we obtained the national sex structure from the World Population Review as an additional sociodemographic variable.29 We selected the latest sex ratio for how many males correspond to 100 females. Studied regions We began with dementia incidence data from 204 countries (GBD), 167 countries with heatwave data and 266 countries with GDP data. After harmonising the data sets, 43 small or disputed territories were excluded due to spatial or naming inconsistencies, resulting in 161 countries. Two more were removed because of missing gridded population data, and six were excluded for having over 20% missing GDP data. The final analysis included 153 countries and territories (online supplemental figure S1). Heatwave calculation Following the literature,30 the strength of the heatwave was quantified by the daily heatwave intensity index (Md) and then added to the annual cumulative heatwave intensity index (HW; online supplemental figure S2). For any given day d, we defined a reference set of temperatures Ad as the collection of all maximum temperatures within 30 days before and after the dth day each year from 1990 to 2019. Next, we calculate the daily index Md as the IQR of the temperature anomaly with respect to T30𝑦25𝑝 on day d. T30𝑦25𝑝 and T30𝑦75𝑝 are the 25th and 75th percentiles of the reference temperature set Ad. The IQR, defined as the difference between T30𝑦25𝑝 and T30𝑦75𝑝, is the definite predetermined heatwave magnitude unit. Finally, we estimated the HW by calculating the sum of the magnitude throughout the year. The detailed information and formulas can be found in online supplemental section 2.2.1. Following the literature,14 we also used alternative indicators of heatwave, including the annual heatwave count, annual heatwave days, annual cumulative heatwave intensity of heatwave days, annual average cumulative intensity of heatwave days, annual maximum cumulative intensity of heatwave days (also known as HWMId30) and annual average duration of heatwave (online supplemental section 2.2.2). These indices were first calculated at each latitude and longitude grid location (i, j), and then summed within each country via the population-weighted approach14 31 (online supplemental figure S3). This step ensured that the heatwave exposure index was aligned with the national-level dementia incidence data, which is reported by country in the GBD data set. Detailed definitions are provided in online supplemental section 2.2.3. Statistical analysis To investigate the association between heatwave and dementia incidence, we first conducted an association analysis using the mixed-effect model at the national level. The advantage of the mixed-effect model is that, unlike the between-country approach, it leverages the within-country effects of heatwave on the incidence of dementia, employing a self-contrast mechanism that aids in alleviating the influence of unobserved confounders. We used this model to obtain the global average effect. Second, we carried out heterogeneity analysis by employing the two-stage MR model to reveal the differences in this heatwave–dementia relationship across different countries. Third, we conducted two supporting analyses, including a state-level analysis for the USA and an individual-level analysis for the UK. Global ecological association analysis To test the impact of heatwave on the annual growth rate of the incidence of dementia, we built a mixed-effect model of the age-standardised incidence of dementia. Mathematically, In this model, the fixed effects of HW, Year, GDP and their interactions were considered. Including the main effect of the Year helps us to control for the linear trend in dementia incidence. The random intercept of the Country variable was considered for the initial difference in the incidence of dementia among countries. The random slope of the Year was considered for the variation in the annual growth rates of this incidence among countries. The interaction effect of HW×Year can estimate the effect of a heatwave on the within-country annual growth rate of incidence. Its positive coefficient indicates an association between heatwave and acceleration in the incidence growth rate. Additionally, we tested for the possible modulation effect of GDP on this HW-by-Year interaction via a three-way interaction, HW × Year × GDP. Its positive coefficient indicates an aggravating association, suggesting that the acceleration of dementia incidence associated with heatwave over time is stronger in regions with higher GDP. These analyses were further carried out for the HICs and LMICs separately, as socioeconomic development has been implicated in research on the health impacts of climate change.32 Considering the possible lag effect of climate events,18 we defined the n-year cumulative heatwave (ACCn) as the sum of HW in the present and previous consecutive (n–1) years (t0, t1, …, t(n–1)). We calculated the partial-eta-square for our mixed-effect model33 and selected the best cumulative year with the highest partial-eta-square for HW×Year. We further employed 1000 cross-validation tests, extracting 80% of the data each time based on the ratio between the HICs and LMICs to verify the stability of the peaked ACC. Sensitivity analyses To test the robustness of the findings, we reran the above models while controlling for additional variables. These variables included between-country effects of heatwave and GDP, the average maximum temperature of the hottest month, the sex structure and cold waves. We also developed an excess mortality index for the dementia population to account for the impact of heatwave-related mortality increases on the observed incidence rates. In addition, we used alternative heatwave indicators focusing on either the duration or the intensity of the heatwave to evaluate the significance of these aspects. Furthermore, we tested the effects of heatwave on the prevalence of dementia. Given that some years had extreme heatwaves, we also reran the model after excluding these extreme data points. To test for non-linearity in the heatwave-incidence relationship, we also included the quadratic and cubic terms of HW in the model. To further assess the impact of country exclusions on model robustness, we additionally reconstructed a near-complete data set via GDP time-series imputation and country-level Graph-Laplacian interpolation of heatwave indices and re-estimated the mixed-effect models on the resulting 199-country data set. Details are provided in the online supplemental section 2.3. Heterogeneity analysis In the first stage, we constructed a linear regression model for incidence, including HW as the main exposure and incorporating GDP as a modifier of its effect, while adjusting for the long-term temporal trend. Mathematically, In the second stage, we applied the meta-analysis to the coefficients of the HW in these models to estimate the average effect all over the world, in the HICs and in the LMICs. If the 95% CI of this averaged global effect does not include zero, a significant effect is detected. To explore the potential effects of heatwave on the incidence of dementia, independent of high temperatures, we controlled for the average maximum temperature of the hottest month in the first-stage analysis. The main effect of HW, as well as the main effect of the average maximum temperature of the hottest month, was then taken into the second stage for meta-analysis, aiming to discern their distinct influences. Next, several potential regulators were added to the MR in the second stage to explain the heterogeneity between countries. These regulators included the yearly mean and SD of GDP, population density, excessive mortality among the dementia population and cold waves. A p value less than 0.05 indicates that the regulator can partly explain the heterogeneity in the extent to which different countries are affected by heatwaves. Supporting analyses in the USA and UK State-level analysis in the USA Controlling for unmeasured country-level confounders, we conducted our analysis at the state level in the USA. For 46 states in the USA, we derived the state-level age-standardised incidence of dementia between 2008 and 2017 from the GBD database and the GDP data from the U.S. Bureau of Economic Analysis.34 Heatwave was aggregated at the state level using the population-weighted approach based on gridded HW data, rather than at the national level as in the primary analysis. We analysed relationship between heatwave and the incidence of dementia in the USA via the mixed-effect model described above. The fixed effects of HW, Year, GDP and their interactions were still considered. However, we replace the random intercept of the country variable with the state variable to account for the initial difference in the incidence of dementia among states, and the random slope of the Year was considered to capture variation in annual trends across states. A positive coefficient for the interaction term HW×Year suggests that heatwave is associated with an accelerated increase in dementia incidence over time. Individual-level analysis in the UK Controlling for individual-level confounders, we conducted our analysis at the individual level in the UK. It is worthy of noting that this individual-level analysis addresses the hazard of individual dementia onset rather than population-level incidence trends. Therefore, this shift from a compositional ecological effect to an individual etiological effect means the UK findings cannot directly confirm or refute the global association. We used the home location and diagnosis data of patients with dementia from the UK Biobank cohort.35 The UK Biobank cohort is a prospective cohort study based on a large-scale population that enrolled over 500 000 participants from 22 assessment centres across the UK between 2006 and 2010. Dementia is defined using ICD10 codes according to the GBD criteria. We included white British individuals who resided in the UK from 1990 to 2010 and had no diagnosis of dementia prior to enrolment. A total of 499 075 participants were included, with 4356 dementia cases (online supplemental section 2.4.1). Gridded daily temperatures at a 1 km resolution from 1 January 1990 to 31 December 2019, were obtained from HadUK-Grid Gridded Climate Observations36 and assigned to participants based on their residential addresses. We first reconstructed the yearly heatwave exposure for each participant according to the literature37 (online supplemental section 2.4.2) and defined the outcomes (online supplemental section 2.4.3). Covariates were selected based on prior literature38 (online supplemental section 2.4.4). We then applied time-varying Cox proportional hazards models39 to estimate the relationship between the incidence of dementia after enrolment and heatwave exposure in the UK (online supplemental section 2.4.5). An HR significantly greater than one suggests increased dementia risk associated with heatwave exposure. Results Global trends and distributions of dementia incidence and heatwave Between 1990 and 2019, we found that the age-standardised incidence of dementia was on average higher (95.95 per 100 000 population, 95% CI 95.51 to 96.38) in the HICs than in the LMICs (88.34 per 100 000 population, 95% CI 87.92 to 88.77; figure 1A; online supplemental table S1). The LMICs had an increasing trend in the age-standardised incidence of dementia (annual growth rate=0.0215 per 100 000 population, 95% CI 0.00834 to 0.0346; figure 1B), whereas the HIC did not. Among the 153 countries, the annual maximum temperature rose by 0.0387°C (95% CI 0.0347°C to 0.0427°C; p<0.0001) per year from 1990 to 2019 (online supplemental figure S4). Weighted by population density, the annual cumulative heatwave intensity index HW was 216.60 (95% CI 215.13 to 218.07; figure 2A; online supplemental table S1). From 1990 to 2019, we detected a rising trend in heatwave (annual growth rate=3.21, 95% CI 2.93 to 3.48; figure 2B). Global view of age-standardised incidence of dementia. Distribution of averaged age-standardised incidence of dementia for 30 years within each country (A); trend of the age-standardised incidence of dementia. Each grey line represents one country. The blue line indicates the average trend in high-income countries and the red line indicates that in the low- to middle-income countries (B). Global view of heatwave. Distribution of averaged HW for 30 years within each country (A); trend of the HW. Each grey line represents one country. The blue line indicates the average trend in high-income countries and the red line indicates that in the low- to middle-income countries (B). HW, the annual cumulative heatwave intensity index. Heatwave associated with dementia incidence increase Among the 153 countries analysed, heatwave was associated with an increase in the incidence of dementia globally (0.0436, 95% CI 0.0391 to 0.0481, p<0.0001 according to a permutation test; figure 3A; online supplemental table S2). The mixed-effect model fit the data well (X2 = 2199.5, p<0.0001), and the random effect on the slope of the year was significant (log likelihood ratio=5197.6, p<0.0001). On average, this association between heatwave and acceleration in dementia incidence was significant in the LMICs (0.0482, 95% CI 0.0437 to 0.0528; p<0.0001) and in the HICs (0.0294, 95% CI 0.0178 to 0.041; p<0.0001; figure 3A; online supplemental table S2). A relatively high GDP mitigated this association in HICs (−0.0101, 95% CI –0.0169 to −0.00325; p=0.0039; online supplemental table S2) but further exacerbated it in LMICs (0.00402, 95% CI 0.000717 to 0.00732; p=0.017; online supplemental table S2). Relationship between the heatwave and age-standardised incidence of dementia. Associations between the growth rate in the age-standardised incidence of dementia and two measures of heatwave, including the HW and the ACC3. These associations were measured by their interaction terms with Year (* × Year) in the mixed-effect models (A); such associations for different heatwave measurements were evaluated cumulatively from 1 to 8 years. The effect sizes of these associations were assessed by partial eta squared (B); Associations between ACC3 and the annual growth of dementia incidence in the high-income countries (blue) and the low- to middle-income countries (red). Data points from various regions and years were grouped into several bins according to the ACC3 values. The mean value (dot) and the 95% CI (bar) of the annual growth of dementia incidence are shown. A linear regression line and its CI are shown (C). ACC3, 3-year cumulative heatwave intensity index; HW, the annual cumulative heatwave intensity index. Notably, the heatwave-associated acceleration in the incidence of dementia remained in the following sensitivity analyses: (1) adjustment for the average maximum temperature in the hottest month, excessive mortality among people with dementia, cold waves and sex ratios in the population; (2) using alternative heatwave indexes; (3) using the prevalence of dementia instead of incidence; (4) exclusion of years with exceptionally high heatwaves; (5) testing for non-linear effect (online supplemental table S2); and (6) testing based on imputed data (online supplemental table S3). Exploratorily, we examined the association between dementia incidence and cumulative heatwave. We found a significant association between cumulative heatwave and acceleration of dementia incidence in both the LMICs and HICs (figure 3B,C). This association peaked at the 3-year cumulative value for both the LMICs and HICs (figure 3B; online supplemental table S2). The significance of these peaks was confirmed by 1000 cross-validations (online supplemental figure S5). Heterogeneity across countries and possible contributors The first-stage linear regression revealed that the association between heatwave and the incidence of dementia varies across countries (figure 4; online supplemental table S4). The second-stage meta-analysis confirmed the association between heatwave and the incidence of dementia both globally among 153 countries (0.0282, 95% CI 0.00851 to 0.0478; p=0.0050) and in the LMICs (0.0386, 95% CI 0.0132 to 0.0639; p=0.0029). These associations remained significant after removing the outliers from the meta-analysis (all countries: 0.0167, 95% CI 0.0001 to 0.0333; p=0.049; LMICS: 0.0367, 95% CI 0.0154 to 0.0580; p=0.00070). When the mean temperature was included as a moderator in the model, it was not a significant predictor of the effect size (–0.0283, 95% CI –0.183 to 0.127; p=0.72), and it did not account for any of the observed heterogeneity. The MR analysis revealed that more cold waves were associated with a stronger relationship between heatwave and dementia incidence across 153 countries (−0.00685, 95% CI –0.0114 to −0.00231; p=0.0031; online supplemental table S5) and in the LMICs (−0.00768, 95% CI –0.0138 to −0.00157; p=0.0014; online supplemental table S5). Global view of the association between heatwave and dementia incidence. Supporting analyses in the USA and UK The first-stage linear regression revealed that, in the USA, heatwave was associated with an increased dementia incidence (0.518, 95% CI 0.0813 to 0.954; p=0.028), whereas in the UK, this association was not significant (0.163, 95% CI –0.0892 to 0.414; p=0.26; online supplemental table S4). Our analysis of higher-resolution data at the state level in the USA confirmed this significant association (0.00401, 95% CI 0.00122 to 0.00681; p=0.0054; online supplemental table S6). Meanwhile, the individual-level analysis in the UK reproduced the non-significant association (online supplemental tables S7 and 8). Discussion This study constructed the first global database of heatwave and dementia incidence rates over 30 years in 153 countries to investigate the impacts of heatwave on dementia incidence. We found a significant increasing trend of heatwaves from 1990 to 2019 worldwide. The incidence of dementia had an increasing trend in both the HICs and the LMICs. Two different methods confirmed a significant heatwave-associated acceleration in the dementia incidence globally. This association with acceleration was most prominent and stable in the LMICs, and its national variations can be explained by factors such as GDP and cold waves. Exploratorily, a similar association has been observed at the subnational level within the USA, suggesting the global pattern was not solely driven by unmeasured cross-national differences. The individual-level analysis in the UK, while null, helps refine the hypothesis by suggesting that the population-level effect might not be driven solely by direct biological pathways in low-risk individuals, pointing to more complex societal or indirect mechanisms. To the best of our knowledge, the current findings provide the first estimation of heatwave-associated global incidences of dementia, which might have wide implications for public health policy. The current findings reveal for the first time that heatwaves are significantly associated with the acceleration of changes in the incidence of dementia globally. Previous research on the effects of heat on dementia often relied on regional or hospital data, focusing on the hospital admission rates and mortality.13 40 In this study, we combined the database with unprecedented coverage of both time (ie, for 30 years) and space (ie, included 153 countries) with the mixed-effect model focusing on the within-country effects and found the heatwave-associated acceleration in the dementia incidence. We found that heatwave was associated not only with dementia incidence but also with prevalence, showing similar patterns. However, we focused on incidence rather than prevalence because incidence more directly reflects new cases and is better suited to detecting potential environmental triggers, such as heatwaves.41 Our findings identified a spatial heterogeneity in the heatwave–dementia association across countries and territories. In HICs, dementia incidence in temperate-zone regions was more sensitive to heatwaves (eg, the USA and New Zealand), whereas the impact was minimal or even reversed in colder regions such as the UK, Northern Europe and Canada. In LMICs, this association was more evident in tropical regions, including Southeast Asia, sub-Saharan Africa and Latin America. The non-significant association in the UK might be understood by the temperate oceanic climate. In our findings, the associations were also non-significant in the countries sharing the same climate, such as Belgium, the Netherlands and France (online supplemental table S3). Consistent with this, previous studies have reported only a slight, non-significant increase in hospital admissions during heatwaves in the UK.42 Especially, we found that cold waves contributed to this spatial heterogeneity. The underlying mechanism might be that frequent shifts between cold and hot extremes can impair the body’s thermal adaptation function,43 leading to greater hospital admission rate of dementia.40 Our findings provide the first evidence that the influence of heatwaves on the growth rate of the incidence of dementia is most prominent when heatwaves cumulate over 3 years. The identification of a multiyear cumulative effect fundamentally reframes the public health challenge posed by heatwaves. It demonstrates that the dementia risk is not merely an acute consequence of individual extreme events, but a pathology that accrues over time,6 in line with the long latency period of dementia. It can also be supported by previous studies reporting the delayed influences of heatwave on mental illnesses.44 45 This finding reveals that the long-term health burden of a warming climate may be substantially underestimated by models that only consider immediate effects. Consequently, our findings necessitate a paradigm shift in adaptation policy—from reactive crisis management during heatwaves to proactive, long-term strategies focused on systematically reducing the cumulative heat exposure of populations over years. Notably, our findings provide evidence for economic-related inequality in the association between heatwave and dementia incidence. Previous studies have focused on HICs, such as those conducted in the USA40 and the UK,13 with few studies from developing countries.11 As a result, it is difficult to observe inconsistencies in the impact of heatwave on dementia across countries with different levels of economic development. However, evidence suggests that heatwaves may have a stronger impact on low-income populations; for example, a 2022 study reported a heatwave-associated increase in sexual violence, which is particularly significant in low-income households in South Asian countries.46 This inequality might be related to both better healthcare conditions47 (eg, closer proximity to health facilities, adequate drug stocks and lower out-of-pocket burden) and less exposure to external climate change48 (eg, greater protection from hazards through better infrastructure and housing in safer locations) in high-income populations. In contrast, in LMICs, a higher average GDP might exacerbate the effects of heatwaves due to rapid urbanisation,49 lagging infrastructure development50 and significant social inequality.51 As the global socioeconomic burden of dementia has been growing,10 addressing the heatwave-associated incidence of dementia has become an increasingly pressing issue worldwide,52 especially in LMICs through targeted policies such as improved heatwave warning systems, access to cooling infrastructure, public health education and climate-resilient healthcare planning. Alternatively, the association observed in LMICs may reflect the fact that there is a rising prevalence of cardiovascular and metabolic conditions which may lead to dementia, and heatwaves are compounding this problem.53 Future strategies in LMICs should integrate poverty reduction with climate resilience by strengthening infrastructure, improving healthcare access and implementing community-based risk management to mitigate the adverse impacts of heatwaves on vulnerable populations. This study also has several limitations. First, although the GBD is currently the most authoritative source of incidence rate data at a global scale, it is not a flawless data set, with limitations mainly arising from sparse or inconsistent primary data and methods used to adjust dementia deaths.22 In regions with limited or inconsistent primary data, the GBD framework uses covariate-driven models and spatiotemporal Gaussian process regression to fill data gaps and correct inconsistencies. While this enables global coverage, it may introduce uncertainty or bias, especially in LMICs where estimates rely more on predictive covariates than direct observations. Moreover, different types of dementia were pooled together, highlighting the need for future studies with more precise data to examine heatwave effects on specific dementia subtypes. Second, the current study used an annual cumulative index for heatwave, but heatwave in daytime/nighttime54 and summer/winter40 may have different effects on dementia incidence. In literature, a heatwave has different definitions by considering additional factors, such as humidity.55 Future studies are needed to examine these variations. The spatial resolution (ie, grid size) of our climate data varies with latitude, which may smooth local heat extremes in coarser, high-latitude grids. However, given the sparse population in these areas, this has a minimal impact on the overall population-based association. Additionally, exposure misclassification may occur due to the use of aggregated heatwave measures, which may not accurately reflect individual-level exposures. Although our national-level findings were consistent with the state-level findings in the USA and individual-level findings in the UK, these limitations highlight the need for future studies using higher-resolution climate and population data to improve spatial precision in exposure assessment. Third, our exploratory analyses in the US and the UK data support triangulation and hypothesis generation, but do not serve as true confirmatory tests of the main global association. The country-level, state-level and individual-level analyses were based on different sources, assumptions and even slightly different research questions (ecological association vs individual aetiology). The null finding in the UK Biobank analysis, contrasted with the positive ecological associations, may suggest that heatwaves influence dementia rates through pathways beyond individual biological risk, such as by stressing healthcare systems or exacerbating social isolation, which are more readily detectable at a population level. This multilevel approach, while complex, is valuable because it provides a more comprehensive picture and highlights the need for future research to disentangle these mechanisms. Fourth, our analysis excluded certain nations, primarily small island nations and some low-income countries, due to missing covariate data. This systematically omits populations that are among the most vulnerable to climate change, potentially underestimating the global risk distribution and limiting the generalisability of our findings to these most-at-risk regions. Consequently, while our results robustly characterise the heatwave–dementia association across 153 countries, they may not fully represent the extremity of risk faced by excluded populations, highlighting a critical equity gap in global environmental health research. Fifth, although the models adjusted for several covariates, unmeasured confounding may remain, particularly related to differences in healthcare access, diagnostic practices, air pollution and public awareness not captured in the data set. Sixth, the observational nature of this study precludes definitive causal conclusions. Future preclinical studies are warranted for the mechanistic roles of heat in the incidence and pathophysiology of dementia. Additionally, due to the collinearity between heatwave and year, it is difficult for a linear model to distinguish the interaction effect (HW × Year) and the non-linear effect (Year2).56 Although the interaction effect was confirmed by the permutation test, future research is warranted to accurately assess these competing hypotheses.57 Seventh, as the primary analysis is based on country-level data, the findings may be subject to ecological fallacy, whereby population-level associations do not necessarily reflect individual-level risk. Conclusion In summary, by constructing a data set of both heatwave and the incidence of dementia in 153 countries over 30 years, the current study uncovered an association between heatwave and an accelerated increase in dementia incidence, as well as marked geographical and economic heterogeneity in this relationship. Supplementary material