Introduction

Population aging is one of the major challenges facing countries worldwide, especially China. In recent years, with the continuous decline in China’s birth rate and the steady increase in average life expectancy, the aging phenomenon in China has become increasingly severe1. Meanwhile, the overall health status of middle-aged and elderly individuals in China is not optimistic. The psychological and physiological health issues that arise with aging are also becoming more prominent2. Cognitive function is closely related to self-care ability and quality of life in middle-aged and elderly individuals. As an important indicator of overall health status, it directly impacts daily functioning and well-being3,4. Cognitive function encompasses the psychological abilities to acquire, apply knowledge, and perform corresponding activities, including attention, episodic memory, and executive function5. With the rapid aging of the population, the prevalence of mild cognitive impairment in China continues to rise6. About 12% to 15% of mild cognitive impairment cases in China progress to Alzheimer’s disease and related dementias annually, compared to only 1% to 2% in healthy adults7. In 2019, the top five causes of death among the Chinese population were stroke, ischemic heart disease, chronic obstructive pulmonary disease, lung cancer, and Alzheimer’s disease, with Alzheimer’s disease rising five places since 19908. Approximately 20% of global deaths from Alzheimer’s disease and related dementias occur in China8. In recent years, the age of diagnosis for Alzheimer’s disease has shown a trend toward younger ages, with the diagnostic age decreasing from 65 to 55 years old8. This indicates that the situation of Alzheimer’s disease in China is concerning, as its seeds are often sown before old age, and without timely intervention, the likelihood and severity of its onset in later life may increase significantly. Existing pharmaceutical and non-pharmaceutical measures have shown limited efficacy, and the underlying pathogenic mechanisms remain unclear9. While aging is a crucial factor, it is irreversible, highlighting the importance of early identification of other risk factors and interventions for the prevention and delay of Alzheimer’s disease9.

Related epidemiological studies indicate that cognitive function in populations is associated with particulate matter in air pollutants. Cognitive dysfunction is associated with exposure to air pollutants, such as particulate matter, and air pollution has been explicitly listed as a risk factor for cognitive impairment10,11. Therefore, advancing the work on air pollution prevention and control not only helps to mitigate cognitive decline but is also crucial for reducing premature deaths associated with exposure to air pollution12, ultimately being vital for the health and well-being of the people in China and similar countries. In most parts of China, particulate matter remains the primary air pollutant, with concentrations far exceeding the latest standards set by the World Health Organization13. Additionally, particulate matter is not a uniform air pollutant but consists of various fine particulate substances with different chemical components, including but not limited to carbon particles, sulfides, and nitrides. These components may have different associations with cognitive function14, so studying only the overall concentration of particulate matter cannot fully reveal its detailed association with cognitive function.

Given that developing countries have a large proportion of the global population, relatively poor living conditions, and poor air quality, most of the increases in cognitive impairment and Alzheimer’s disease occur in these countries. Currently, 60% of dementia patients live in low- and middle-income countries, and this number is expected to rise to 71% by 205015. The fastest growth in the elderly population is taking place in China, India, and their South Asian and Western Pacific neighbors15. This study focuses on China, the largest developing country, to examine the relationship between particulate matter and its components in air pollution with cognitive scores in middle-aged and elderly populations. It further explores whether improvements in air quality under different Shared Socioeconomic Pathways (SSPs) can offset the cognitive decline associated with rapid aging and related economic costs. The findings may offer valuable insights for reducing the burden of Alzheimer’s disease and related dementias, enhancing public health, and supporting sustainable economic and social development in China and other countries facing similar demographic and environmental challenges.

Results

Descriptive statistics

This study involved 7035 adults, each of whom was interviewed three times. The baseline mean age was 59.45 (standard deviation of 8.41). The average cognitive scores for the 2011, 2013, and 2015 waves of the China Health and Retirement Longitudinal Study (CHARLS)16 were 15.01, 15.31, and 14.81. Corresponding PM1 (Particulate matter with a diameter smaller than 1 µm) concentrations were 31.75, 33.50, and 27.06 µg/m3, with the highest values in eastern China. The corresponding PM2.5 (Particulate matter with a diameter smaller than 2.5 µm) concentrations were 57.01, 60.72, and 48.33 µg/m3, with the highest values in eastern China. The corresponding PM10 (Particulate matter with a diameter smaller than 10 µm) concentrations were 96.76, 103.20, and 82.27 µg/m3, with high values in the desert regions of northwest China. Table 1 summarizes the basic statistical information in the study, and Fig. 1 shows the regional distribution of particulate matter and its components.

Table 1 Summary table of population variables of the three waves of research subjects
Full size table
Fig. 1: Average Exposure Concentration Distribution of Particulate Matter and Its Components from 2010 to 2015.
figure 1

Color scales represent the average concentrations of particulate matter and its components (μg/m³), with red indicating higher concentrations and blue indicating lower concentrations. The maps, arranged from top to bottom and left to right, show the spatial distributions of PM₁, PM₂.₅, PM₁₀, BC (black carbon), OM (organic matter), NH₄⁺ (ammonium), SO₄²⁻ (sulfate), and NO₃⁻ (nitrate).

Full size image

There was a highly significant positive correlation between PM1, PM2.5, PM10, black carbon, organic matter, ammonium, nitrate, and sulfate (as shown in Table 2), indicating that they may have similar sources and co-exist in the air.

Table 2 Pearson correlation coefficient of particulate matter and its components
Full size table

The association between particulate matter and its components with cognitive scores in single-pollutant models

The results of the single-pollutant linear mixed-effects models indicated that most pollutants were statistically significantly associated with lower cognitive scores, except for nitrate, which showed no statistically significant association. Specifically, for every 0.1 µg/m3 increase in PM₁, the cognitive score decreased by 0.53 points (95% CI: [−0.95, −0.09]); for every 0.1 µg/m3 increase in PM₂.₅, the cognitive score decreased by 0.30 points (95% CI: [−0.54, −0.07]); for every 0.1 µg/m3 increase in PM₁₀, the cognitive score decreased by 0.14 points (95% CI: [−0.27, −0.01]); for every 0.1 µg/m3 increase in organic matter, the cognitive score decreased by 1.44 points (95% CI: [−2.19, −0.62]); for every 0.1 µg/m3 increase in ammonium, the cognitive score decreased by 1.55 points (95% CI: [−2.60, −0.36]); for every 0.1 µg/m3 increase in sulfate, the cognitive score decreased by 1.70 points (95% CI: [−2.58, −0.70]); and for every 0.1 µg/m3 increase in black carbon, the cognitive score decreased by 7.23 points (95% CI: [−8.26, −5.56]). The standard adjustment model and the fully adjusted model after adding ozone control showed similar relationships (see Fig. 2 for details). The cognitive score decreased by 0.13 for every year increase in age (95% CI: [−0.14, −0.12]).

Fig. 2: The association between particulate matter and its components and cognitive scores under different single-pollutant models.
figure 2

Estimated correlation coefficients between various pollutants and cognitive scores under three single-pollutant models: fully adjusted (blue), fully adjusted with ozone (yellow), and standard adjusted (red). Pollutants include ammonium, black carbon, nitrate, organic matter, PM₁, PM₂.₅, PM₁₀, and sulfate. Error bars represent 95% confidence intervals. Negative coefficients indicate a potential inverse association between pollutant exposure and cognitive scores. The consistency of estimates across models suggests the robustness of the observed associations.

Full size image

The single-pollutant generalized additive mixed-effects models were used to examine the potential non-linear associations between particulate matter and its components and cognitive scores. The results indicated that PM₁ and nitrate exhibited linear associations with cognitive scores (P = 0.08 and P = 0.13, respectively), while PM₂.₅ (P = 0.01), PM₁₀ (P = 0.01), black carbon (P = 1.00e-5), ammonium (P = 0.01), sulfate (P = 1.10e-3), and organic matter (P = 1.20e-3) demonstrated non-linear associations. These results indicated that the relationships between most pollutants and cognitive scores are not strictly linear. Detailed information can be found in Support Table 1 and Support Fig. 1 to 8.

The association between particulate matter and its components with cognitive scores in multi-pollutant models

The results of the multi-pollutant linear mixed-effects model indicated that nitrate, ammonium, and PM₁₀ were not statistically significantly associated with cognitive scores. In contrast, several other pollutants showed statistically significant negative associations. Specifically, for every 0.1 µg/m3 increase in PM₁, the cognitive score was associated with a decrease of 0.13 points (95% CI: [−0.25, −0.01]); for PM₂.₅, a 0.1 µg/m3 increase was associated with a 0.13-point decrease (95% CI: [−0.26, −0.01]). For organic matter, a 0.1 µg/m3 increase was associated with a 0.58-point decrease (95% CI: [−1.15, −0.01]); for sulfate, a 1.22-point decrease (95% CI: [−2.32, −0.11]); and for black carbon, a 5.75-point decrease (95% CI: [−9.21, −2.29]) was observed per 0.1 µg/m3 increase (see Fig. 3 for further details). The cognitive score was associated with a decrease of 0.13 for every year increase in age (95% CI: [−0.14, −0.12]).

Fig. 3: The association between particulate matter and its components and cognitive scores under different multi-pollutant models.
figure 3

Estimated correlation coefficients between various pollutants and cognitive scores under three multi-pollutant models: fully adjusted (blue), fully adjusted with ozone (yellow), and standard adjusted (purple). Pollutants include ammonium, black carbon, nitrate, organic matter, PM₁, PM₂.₅, PM₁₀, and sulfate. Error bars represent 95% confidence intervals. The comparison across models reflects the impact of co-pollutant adjustments and indicates the robustness of the associations observed.

Full size image

The multi-pollutant generalized additive mixed-effects models were used to examine the potential non-linear associations between particulate matter and its components and cognitive scores. The results indicated that PM₁, PM₂.₅, black carbon, and organic matter exhibited non-linear associations with cognitive scores (P = 0.02, P = 1.00e-5, P = 1.00e-5, and P = 0.01, respectively). In contrast, PM₁₀, ammonium, nitrate, and sulfate demonstrated linear associations with cognitive scores (P = 0.06, P = 0.14, P = 0.68, and P = 0.16, respectively). Detailed information can be found in Support Table 2 and Support Figures 9 to 16.

Average age of China’s population over 45 years old under different SSPs

By 2030, the average age of adults over 45 in China will be approximately 62 years old under different SSPs, including 62 years old under the Sustainable Development Scenario (SSP1: rapid economic growth with reduced use of energy and resource-intensive agricultural products, a significant reduction in inequality within and between countries, and strong controls on air pollution) and Moderate Development Scenario (SSP2: various food consumption and energy production patterns similar to current trends, with corresponding measures to control air pollutants, as developing economies catch up with developed countries leading to gradual emission reductions over time) and 61 years old under the Regional Development or Inequality Scenario (SSP3: high inequality both within and between countries, ineffective policies in land use regulation, air pollution control, and greenhouse gas emissions leading to the highest levels of pollutant and aerosol emissions). By 2050, the average age under each scenario will increase to approximately 66 years old, including 67 years old under the SSP1, 66 years old under the SSP2, and 65 years old under the SSP3. The trend is shown in Support Fig. 17.

Comparison of changes in particulate matter and its components under different SSPs

Under the SSP1, the concentrations of particulate matter and its components will exhibit a significant downward trend, with the most pronounced decrease occurring by 2050. Under the SSP2, the concentrations of particulate matter and its components will also decline, but the reduction will be relatively modest. In contrast, under the SSP3, the concentrations of some components will increase by 2050, followed by a slight decline or stabilization. Overall, air quality improvement in terms of particulate matter and its components will be most significant under the SSP1, followed by SSP2. In contrast, the improvement under the SSP3 will remain limited, with concentrations of certain components expected to rise (see Table 3 and Support Figs. 18–33 for details).

Table 3 The changes in particulate matter and its components under different SSPs
Full size table

In future development scenarios, the average concentrations of particulate matter and its components in the northeastern area will be lower than those in the central, eastern, and western regions. Under the SSP1 and SSP2, the concentration declines in the eastern and central areas will be more substantial than in the northeastern and western regions. In contrast, under the SSP3, changes across all areas will be relatively small, but an overall upward trend will still be observed (see Support Figs. 34–48 for details).

Benefits of particulate matter and components vs. aging under different SSPs

Suppose the estimated correlation coefficient is based on the single-pollutant models by 2030 and 2050 under the SSP1. In that case, reducing PM1 concentration will have a more significant positive benefit on cognition than the negative impact of population aging. Under the SSP2, the decrease in PM1 concentration will also offset the negative impact of aging on cognition. Under the SSP3, the increase in PM1 concentration will exacerbate the negative impact of aging. Similar trends were observed for other pollutants. Since nitrate SSP3 had fewer patterns, comparing it may be less accurate (see Support Figs. 43 to 45 for details). Estimates based on the multi-pollutant model also produced similar results (see Support Figs. 46–48 for details).

Comparison with the benefit of aging after reaching the current national standard threshold

In 2030, assuming that particulate matter of different diameters in China meets the current national standards (annual averages of PM1 ≤ 15 µg/m3, PM2.5 ≤ 35 µg/m3, PM10 ≤ 70 µg/m3), the positive benefit of the reduction in particulate matter is greater than the negative impact brought about by aging. By 2050, as aging continues to develop, the negative impact of aging continues to increase, but it is still lower than the positive benefit of reducing particulate matter.

Economic cost analysis

Under the SSP1, the reduction in particulate matter and the concentrations of its components will potentially decrease healthcare costs related to Alzheimer’s disease and cognitive impairment-induced dementia by ~116.25 billion Chinese Yuan (CNY) (calculated at a 7:1 CNY to dollar exchange rate, equivalent to about 16.61 billion dollars); under the SSP2, costs will be reduced by around 114.39 billion CNY (~16.34 billion dollars); and under the SSP3 development scenario, costs will potentially increase by 112.59 billion CNY (around 16.08 billion dollars). Looking ahead to 2050, if China continues to implement the SSP1, the reduction in particulate matter and its components concentrations will likely reduce costs by 135.46 billion CNY (around 19.35 billion dollars); under the SSP2, costs will be reduced by 126.67 billion CNY (around 18.09 billion dollars); whereas under the SSP3, costs will likely increase by 118.31 billion CNY (around 16.90 billion dollars).

Discussion

Based on retrospective survey data across China, this study identified a significant link between increased concentrations of particulate matter and its components and cognitive impairment in middle-aged and older Chinese adults, and this association exhibited a robust exposure-response relationship. Further analysis revealed an encouraging finding: even under the pressure of rapid aging, the positive benefit of the reduction in particulate matter concentration can offset the negative impact of aging on cognitive function. Because the human body’s natural aging is irreversible, studying the association of air pollution with cognitive function has more public health significance for the stable and sustainable development of human society to some extent.

Although many studies have examined the relationship between air pollutant concentrations and cognitive function in populations, the underlying biological mechanisms are not fully understood9. Existing research showed that air pollution may have a negative association with the central nervous system and lead to central nervous system diseases17,18,19,20,21. The association between air pollution and the central nervous system is primarily mediated through the inhalation of particulate matter via the respiratory system22. The use of traced radioactive carbon spots demonstrated that inhaled particles could pass through the delicate tissue within the rodent’s nasal cavity, travel along neurons, and ultimately reach the cerebellum at the back of the brain, triggering an inflammatory response23. The accumulation of particulate matter in the brain induces oxidative stress and neuroinflammation, which can damage the central nervous system and lead to neurodegenerative diseases. Neuroinflammatory responses may lead to brain synaptic dysfunction, which is one of the main mechanisms of particulate matter-induced cognitive impairment24.

Relevant animal experiments showed that in mice that inhale polluted air, microglia in the brain release a large number of inflammatory molecules, including tumor necrosis factor alpha, which is elevated in the brains of Alzheimer’s disease patients. Mice exposed to polluted air also showed other signs of brain damage, such as accumulation of amyloid beta, axonal atrophy, and brain atrophy25. These findings provided important insights into the relationship between air pollutants and cognitive function. Related brain imaging and air pollution studies further supported these findings. Long-term exposure to PM2.5 and PM10 is associated with changes in cortical thickness and subcortical volume in adults: as the concentration of particulate matter increased, the thickness of the frontal lobe, temporal lobe, parietal lobe, and insula became thinner, while the thickness of the occipital lobe and cingulate cortex became thicker; at the same time, the thickness of the thalamus, caudate nucleus, putamen, hippocampus, amygdala, and the nucleus accumbens also decreased in size. These changes in brain structure are closely related to cognitive dysfunction, indicating that the negative impact of air pollution on the brain is widespread and far-reaching26. The above strong evidence supports the findings of our nationwide retrospective study.

This study explored the association between long-term exposure to particulate matter of varying diameters and cognitive scores. The findings indicated that smaller particle diameters were associated with stronger negative correlations with cognitive scores. This phenomenon can be attributed to the unique biochemical properties of small-diameter particles. Specifically, tiny particles are more likely to cross the blood-brain barrier and reach the alveoli and other target sites, thereby exerting a more significant impact on the nervous system27,28.

In terms of particulate matter components, the results of this study showed statistically significant associations between black carbon, organic matter, and sulfate and cognitive scores. Although the content of black carbon in particulate matter is relatively small, this study showed that the association between black carbon and cognitive scores in middle-aged and elderly individuals is the strongest among the five components. Therefore, limiting black carbon emissions will bring considerable benefits to improving the cognitive health of middle-aged and elderly people in China. Research by Segersson showed that black carbon produced by traffic exhaust is the primary source of black carbon and is closely associated with human health29. In rural areas of China, the primary source of black carbon is inefficient cooking systems that use polluting fuels, including wood, charcoal, animal manure, crop waste, coal, and kerosene30,31. Organic matter, organic salts, and sulfate mainly originate from fossil fuel combustion and motor vehicle emissions32,33,34. However, most current studies on the relationship between particulate matter and population health generally assumed that the health impacts of all particulate matter components are the same or that the toxicity of each element is consistent across different geographical regions, which ignored the specific conditions of various components and regions35,36.

Therefore, based on this study’s results, China needs to refine the key treatment priorities for different pollutants. Based on the current focus on PM1, PM2.5, and PM10 for air pollution prevention and control, a more specific and comprehensive air pollution prevention and control plan targeting different air pollutants and their components should be further developed. Of course, as China vigorously promotes sustainable development, especially its efforts in new energy transportation and rural clean energy development plans, it will help reduce the emission of harmful particulate matter, including black carbon. Regarding policy formulation, standards should be further refined and improved, especially emission standards for black carbon, organic matter, and sulfate, to achieve more effective air pollution prevention and control.

This study found that by 2030 and 2050, if the particulate matter meets the currently formulated standards, the positive benefit of reduced particulate matter concentrations on cognition can offset the negative impact of aging. However, China lacks the implementation standards for key components such as black carbon, organic matter, and sulfate. In summary, in the future, China needs to formulate and improve the quality standards for particulate matter components to protect the cognitive health of middle-aged and older adults.

This study found that under the SSP1, with the improvement of lifestyle, environmental quality, and socioeconomic development, the mortality rate of middle-aged and elderly people will decrease, increasing the total population of this group. As the middle-aged and elderly population grows, the demand for medical and social services will increase significantly, increasing the burden on the country. Although such changes bring new challenges, this study indicated that the short-term benefits of implementing sustainable development strategies may be limited; however, their long-term adoption can not only significantly improve cognitive function but also contribute to maintaining national peace and prosperity. This study also suggested that continued reductions in pollutant concentrations will bring significant economic benefits. According to the study’s rough estimates, related costs could be reduced by $19.35 billion by 2050 (Without considering inflation in the economy), and these financial savings are pretty significant. It can be seen that adopting sustainable development policies can not only safeguard the cognitive function of residents but also lead to substantial savings in economic costs. The results of this study send a clear positive message to China and countries in similar situations to China, whose populations are aging rapidly. Although in the sustainable development scenario, the extension of the average lifespan of the middle-aged and elderly may lead to an increase in the number of people suffering from cognitive impairment and Alzheimer’s disease, improving air quality can offset this negative impact and save a large amount of economic costs. This kind of environmental governance is not only beneficial to China but also has positive significance for the health systems, social care, and families of many countries with similar situations to China, especially low- and middle-income countries.

Therefore, we believe that in order to fully reap the benefits of life extension, relevant countries must actively adopt policies, take environmental governance actions, and develop more state-funded new energy or low-energy consumption projects to meet the growing demand of the middle-aged and elderly population for a good air environment. In line with the United Nations Sustainable Development Goals37, we can adopt a series of effective measures to promote sustainable development. For example, we can widely promote the use of clean energy and reduce dependence on fossil fuels; improve public transportation infrastructure to reduce vehicle emissions; strengthen urban greening efforts by increasing green spaces and parks, providing more recreational areas for residents; and implement strict air pollution control policies, enhancing industrial emission monitoring and reduction. Through these initiatives, we can not only significantly improve the quality of life for middle-aged and elderly individuals but also bring dual health and economic benefits to society as a whole.

This study is the first to focus on China, the largest developing country, and to explore the relationship between cognitive scores of middle-aged and elderly people and particulate matter and its components, which provides a robust scientific basis for relevant countries. It includes particulate matter of different diameters (PM1, PM2.5, PM10) and components (SO42−, NO3−, NH4−, OM, BC), which is rare in previous studies and provides a more comprehensive perspective. Secondly, this study covers the life cycle from middle to old age, which is of great significance for prevention and research starting from middle age. Thirdly, this study explores and quantifies the economic cost relationship between particulate matter and its components and the cognitive function of middle-aged and elderly people under different shared socioeconomic paths and national levels. Finally, the results of this study point out the significance of meeting the current national implementation standards for particulate matter in safeguarding the cognition of the middle-aged and elderly. It also suggests that China needs to improve the implementation standards for particulate matter and its components in order to achieve better public health and improvement in cognitive function, which has seldom been compared before.

However, this study is subject to the following limitations. Firstly, due to the lack of specific addresses, assessing the exposure to environmental particulate matter and its components at the city level may lead to mismatching of exposures and overlook the heterogeneity of concentrations within cities. Secondly, although this study adjusts for potential confounding factors, including a series of healthy behaviors, there may still be some confounding factors (such as noise38, medication history9, blue space39, etc.) that have not been included, thus affecting the observed associations between particulate matter and its components and cognitive scores. Thirdly, physiological data (such as central nervous system inflammatory factors) were not collected in this study to examine the physiological and biochemical response processes of particulate matter on cognition discussed in the previous section. Therefore, further research is needed to examine its mechanisms. Fourthly, in terms of economic cost calculations, this study only has rough estimates, so when formulating strategies and decisions, more in-depth analysis and evaluation of the impact of various factors are needed. Finally, missing data are often unavoidable for large longitudinal studies, which may bias the results.

In conclusion, this study identified a significant link between increased concentrations of particulate matter and its components and cognitive impairment in middle-aged and older Chinese adults. Under the sustainable development scenario (SSP1), particulate matter and its concentration decline rapidly, improving cognitive function conditions. This study not only enriched the epidemiological evidence on the adverse impact of air pollution on cognitive health but also showed that the implementation of sustainable development policies can improve the cognitive health of middle-aged and elderly people and, to a certain extent, compensate for the impact of aging on the cognitive health of middle-aged and elderly people in China, and bring significant health and economic benefits.

Methods

Study design and population

The primary data for this study were sourced from the CHARLS. The CHARLS aimed to establish a high-quality micro-database representing middle-aged and elderly residents aged 45 and above in mainland China16. This study utilized data from three survey waves (2011, 2013, and 2015) to explore the association between particulate matter and its components with cognitive function. Individuals under 45, those missing baseline age data, and participants who did not have cognitive measurements in all three survey rounds were excluded (see Support Fig. 49 for details). Ultimately, this study included 7035 participants (20,115 observations) from 123 cities, with the specific population distribution across cities shown in Fig. 4.

Fig. 4: Distribution of participants (Notes: The population density in Northwestern China is relatively low, rather than gaps in coverage).
figure 4

Spatial distribution of survey participants across China. Colored circles represent the number of participants at each ___location, ranging from green (approximately 100 participants) to red (up to 500 participants). The sparse distribution of circles in northwest China reflects lower population density rather than insufficient survey coverage, supporting the representativeness of the sampling across regions.

Full size image

Outcome

The CHARLS project assessed the cognitive function of middle-aged and elderly individuals using the internationally recognized Mini-Mental State Examination and the cognitive function telephone interview. This comprehensive, accurate, and rapid cognitive assessment reflects the intellectual status and degree of cognitive impairment of the subjects, specifically measuring orientation, attention, memory, and visuospatial abilities40. The assessment process involves the interviewer reading out ten Chinese nouns to each respondent, who then attempts to repeat as many of these nouns as possible. The number of correctly repeated nouns constitutes the immediate memory score (0–10). Following a 2–3 min delay, the interviewer requests the respondent to recall the previously mentioned nouns, with the count of accurately recalled nouns (ranging from 0 to 10) determining the delayed memory score. The sum of these two scores provides the episodic memory score, ranging from 0 to 20. To measure the mental status score, the interviewer asks the respondent to subtract 7 from 100 five times in a row and state the results while also writing down the current date and day of the week, with the score ranging from 0 to 9. Additionally, the interviewer shows the respondent a picture of two overlapping pentagons. The respondent earns 1 point if they can accurately replicate the drawing; otherwise, they score 0. The total score of the aforementioned tests constitutes the cognitive function score, ranging from 0 to 30. Higher scores indicate better cognitive function, while lower scores indicate poorer cognitive function (A score below 17 points typically indicates possible cognitive impairment. Note: Since the episodic memory score accounts for 20 out of the total 30 points, this scoring system places greater emphasis on memory function.).

Covariates

Based on existing research on the association between air pollution and cognitive function and utilizing the extensive demographic data provided by the CHARLS database16,41, this study controlled for a series of demographic characteristic variables (residence, gender, age, education levels (considering China’s context, the education levels for most middle-aged and elderly individuals are categorized as: no formal education, primary school, middle school, and high school or above.), marital status, retirement status, BMI). In terms of personal health behaviors, this study controlled for exercise status, smoking status (defined as any past or present smoking behavior), and drinking status (defined as any past or present drinking behavior). Given that chronic illnesses and medication use can influence cognitive decline42, individual health status was also controlled for in this study. Comfortable weather conditions can enhance cognitive function43. Therefore, this study controlled for temperature and relative humidity at the 2-meter level at the individual’s residence (grid resolution: 0.1° × 0.1°, temporal resolution: hourly, data source: [The European Centre for Medium-Range Weather Forecasts] (https://cds.climate.copernicus.eu/datasets/reanalysis-era5-land)). Additionally, a good greening environment has been found to slow cognitive decline44, so this study also controlled for the NDVI (grid resolution: 0.03° × 0.03°, temporal resolution: annual, data source: [National Ecosystem Science Data Center] (http://www.nesdc.org.cn/)). Table 1 presents the covariates controlled in this study.

Exposure

From 2010 to 2015, annual grid data of PM1, PM2.5, and PM10 concentrations at a resolution of 0.01° × 0.01° were obtained from the China High-resolution Air Quality Monitoring dataset (temporal resolution: daily, data source: https://weijing-rs.github.io/product_cn.html). This dataset integrated multiple sources, including satellite remote sensing, machine learning algorithms, ground-based observations, atmospheric reanalysis, and model simulations to produce high-resolution near-surface air pollution data45,46,47. The predicted PM1, PM2.5, and PM10 concentrations have shown good consistency with ground-based measurements, with validation metrics as follows: PM1 cross-validation coefficient of determination (CV-R2) = 0.83, root mean square error (RMSE) = 9.50 µg/m3, mean absolute error (MAE) = 6.17 µg/m3. PM2.5 CV-R2 = 0.92, RMSE = 10.76 µg/m3, MAE = 6.32 µg/m3. PM10 CV-R2 = 0.90, RMSE = 21.12 µg/m3, MAE = 11.22 µg/m3. Since the exact residential addresses of the participants were not available for the CHARLS, this study assessed the two-year moving average exposure to particulate matter at the city level by matching and extracting grid data based on the investigators’ locations. Figure 1 illustrates the average exposure distribution of PM1, PM2.5, and PM10.

The component data for PM2.5 were sourced from the TAP dataset (temporal resolution: annual, grid resolution: 0.1° × 0.1°, data source: [TAP Data] (http://tapdata.org.cn)), which included five main components: SO42−, NO3−, NH4−, OM, BC. These components were derived using a combination of weather research and forecasting-community multi-scale air quality modeling systems, ground observations, machine learning algorithms, and integrated PM2.5 data. The component dataset aligned well with existing observations (monthly correlation coefficients ranging from 0.64 to 0.75, daily correlation coefficients ranging from 0.67 to 0.80; most normalized average biases within ±20%)48. Like particle data, component data were extracted by matching and cropping grid data based on the investigators’ ___location. Figure 1 displays the average exposure distribution of each component.

Future exposure

Future air pollution data were sourced from the Scenario Model Intercomparison Project within Coupled Model Intercomparison Project Phase 6 (CMIP6). Scenarios within CMIP6 were composed of different SSPs and Representative Concentration Pathways forming scenario matrices49. SSPs depict future societal development scenarios in the absence of climate change or climate policies, including Sustainable Development, Moderate Development, and Regional Development or Inequality. This study employed data from 10 Earth System Models (BCC-ESM1, CESM2-WACCM, EC-Earth3-AerChem, GFDL-ESM4, IPSL-CM5A2-INCA, MIROC-ES2L, MPI-ESM-1-2-HAM, MRI-ESM2-0, NorESM2-LM, and UKESM1-0-LL)50. Specific details of these Earth System Models are provided in Support Table 3. The average age of Chinese adults aged 45 years and above in the future was derived from global population projections for China (http://dataexplorer.wittgensteincentre.org/wcde-v2/). Support Table 4 for further details. Since some CMIP6 models lacked PM2.5, this study referenced empirical formulas from Chowdhury51 and Van Donkelaar52 to estimate PM2.5 concentrations.

Statistical analysis

Continuous variables for participant characteristics were presented as means (±standard deviation), while categorical variables were presented as counts (%). The methodology for calculating future average age in a specific year involves multiplying the midpoint age of each age interval by its corresponding population size, summing these products, and then dividing by the total population. The Pearson correlation coefficient test assessed the relationship between air pollutants.

Linear mixed-effects models with a random intercept at the individual level were employed to longitudinally analyze cognitive scores in relation to exposure to particulate matter and its components53. The regression model that includes all covariates described in the covariates subsection is referred to as the fully adjusted model. The model may be too complex to produce stable estimates, which increases the uncertainty of the model. To assess whether estimated associations are sensitive to these limitations, this study also applied a standard adjustment, including demographic factors (age, sex, region), lifestyle factors (smoking, drinking, exercise), health risk factors (diabetes, hypertension), and environmental factors (temperature, humidity, NDVI). Furthermore, considering recent evidence on the impact of ozone on cognition54 and China’s increasingly severe ozone situation55, the analysis was conducted by adding ozone into the fully adjusted model. (Ozone also comes from https://weijing-rs.github.io/product_cn.html, temporal resolution: daily, grid resolution: 0.01° × 0.01°). To analyze the association between co-exposure to different pollutants and cognitive scores, this study regressed highly correlated pollutants against each other and used the residuals as covariates in model56. Generalized additive mixed-effects models were used to explore the nonlinear association between cognitive scores and exposure to particulate matter and its components.

Based on the estimated correlation coefficients among particulate matter and its components, age, and cognitive scores, this study assessed the potential cognitive benefits for middle-aged and older adults in China from future changes in particulate matter and its components under different SSPs, particularly the sustainable development scenario (SSP1), as well as the negative impact of future age-related changes on cognition. Subsequently, this study compared the potential cognitive benefits resulting from improvements in particulate matter and its components with the negative impact of aging. If the cognitive benefits from improvements in particulate matter and its components outweigh the negative impact of aging, reducing the concentrations of particulate matter and its components may help mitigate age-related cognitive decline.

In addition, this study evaluated whether achieving China’s current particulate matter standards by 2030 and 2050 (annual averages: PM₁ ≤ 15 µg/m3, PM₂.₅ ≤ 35 µg/m3, PM₁₀ ≤ 70 µg/m3) could yield cognitive benefits and whether these benefits could offset the negative impact of aging. Due to the lack of standardized concentration thresholds, comparisons involving specific particulate matter components were not conducted.

This study projected the population size for 2030 and 2050 to evaluate the impact of improved particulate matter and its components on the total healthcare costs for China’s population aged 45 and above. Assuming a national Alzheimer’s disease prevalence rate of 3.48%57, a reduction in long-term exposure to air pollution could lead to a 4% decrease in incidence9, with the medical cost per Alzheimer’s and related dementia patient being 122,523 CNY58. Based on these assumptions, this study analyzed the cost savings under different scenarios.

All statistical analyses were conducted using R (version 4.2.1), and a 2-tailed P < 0.05 was considered statistically significant. Missing values were handled using the mice package for multiple imputation. Linear mixed-effects models were employed using lme4, while generalized additive mixed-effects models utilized gamm4, mgcv, mgcViz, and Splines packages.