Ambient nitrogen dioxide and cardiovascular diseases in rural regions: a time-series analyses using data from the new rural cooperative medical scheme in Fuyang, East China

Most of studies relating ambient nitrogen dioxide (NO2) exposure to hospital admissions for cardiovascular diseases (CVDs) were conducted among urban population. Whether and to what extent these results could be generalizable to rural population remains unknown. We addressed this question using data from the New Rural Cooperative Medical Scheme (NRCMS) in Fuyang, Anhui, China. Daily hospital admissions for total CVDs, ischaemic heart disease, heart failure, heart rhythm disturbances, ischaemic stroke, and haemorrhagic stroke in rural regions of Fuyang, China, were extracted from NRCMS between January 2015 and June 2017. A two-stage time-series analysis method was used to assess the associations between NO2 and CVD hospital admissions and the disease burden fractions attributable to NO2. In our study period, the average number (standard deviation) of hospital admissions per day were 488.2 (117.1) for total CVDs, 179.8 (45.6) for ischaemic heart disease, 7.0 (3.3) for heart rhythm disturbances, 13.2 (7.2) for heart failure, 267.9 (67.7) for ischaemic stroke, and 20.2 (6.4) for haemorrhagic stroke. The 10-μg/m3 increase of NO2 was related to an elevated risk of 1.9% (RR: 1.019, 95% CI: 1.005 to 1.032) for hospital admissions of total CVDs at lag0–2 days, 2.1% (1.021, 1.006 to 1.036) for ischaemic heart disease, and 2.1% (1.021, 1.006 to 1.035) for ischaemic stroke, respectively, while no significant association was observed between NO2 and hospital admissions for heart rhythm disturbances, heart failure, and haemorrhagic stroke. The attributable fractions of total CVDs, ischaemic heart disease, and ischaemic stroke to NO2 were 6.52% (1.87 to 10.94%), 7.31% (2.19 to 12.17%), and 7.12% (2.14 to 11.85%), respectively. Our findings suggest that CVD burdens in rural population are also partly attributed to short-term exposure to NO2. More studies across rural regions are required to replicate our findings.


Introduction
Cardiovascular diseases (CVDs) were the major disease burden around the world, and the burden of CVDs has been rising in last decades. Prevalent cases of CVDs raised 271 to 523 million, and the number of the death from CVDs raised 12.1 to 18.6 million from 1990 to 2019 (Roth et al. 2020). Over 60% of CVDs are borne in China. In 2019, there were 330 million patients with cardiovascular diseases, contributing to 46.66% of total deaths in rural regions as compared to 43.8% in urban regions (Writing Committee of the Report on Cardiovascular Health and Diseases in China 2022). Clearly, prevention and control of CVDs may be long term challenges for China, which is particularly true for rural population due to a large rural population and their insufficient health resources.

Responsible Editor: Lotfi Aleya
Teng-Fei Dong, Zhen-Qiu Zha, Liang Sun, and Ling-Li Liu have equal contributions to this work.
Air pollution was one of well-known risk factors to human health around the world (Tong 2019) and the fourth threats for the disease burden in China (Yang et al. 2013). Growing studies has indicated adverse effects of air pollution on cardiovascular health (Adebayo-Ojo et al. 2022;Dastoorpoor et al. 2019;Sepandi et al. 2021). Among air pollutants, nitrogen dioxide (NO 2 ) was one of toxic components within air pollution mixtures (Latza et al. 2009), and exposure to NO 2 was related to elevated risk for CVDs. For instance, a study in Lanzhou city, Gansu province, China showed a positive relationship between NO 2 and daily CVD admissions (Ma et al. 2017a). Another study using data from 172 cities in China also showed a positive relation between short-term exposure to NO 2 and ischaemic stroke admissions . However, most of previous studies on the relation between NO 2 and CVDs focused on urban population, leaving the question unanswered of whether and to what extent these findings could be generalizable to rural population.
So far, limited studies indicated the urban-rural differences in the associations of air pollution (including NO 2 ) with CVDs, whereas did not receive consist results. A study using China Death Cause Monitoring Dataset found that the mortality attributed to air pollution was higher in rural regions than in urban regions, indicating that rural population may be more sensitive to air pollution than urban population in China (Zhao et al. 2021). Another study, however, found a positive association between NO 2 and CVD hospitalizations in urban regions but not in rural regions . The mixed findings on urban-rural differences on the relation of air pollution with CVDs suggested that more studies should be conducted in rural population. A recent study used the data from NRCMS in Pingliang, Northwest China, to examine the effects of short-term exposure to air pollutants on the health burden of hospitalized patients with CVDs, and received some novel results in rural population: NO 2 was the most contributor to hospital admission costs of CVDs; among six air pollutants include particulate matter ≤ 2.5 μm (PM 2.5 ), particulate matter ≤ 10 μm (PM 10 ), carbon monoxide (CO), NO 2 , sulphur dioxide (SO 2 ), and ozone(O 3 ) (Zha et al. 2022). Given that NRCMS can cover over 98% of local farmers without sampling error, these findings may be believable. However, whether or not these findings are repeatable across rural regions remain unknown.
This study aimed to replicate and extend previous studies on the associations between short-term exposure to NO 2 and CVD admissions using data from NRCMS in Fuyang city, Anhui province, China, with a rural population of 4.76 million, accounting for 58.03% of total population (Anhui Bureau of Statistics 2020). We used constrained distributed lag model (CDLM) to examine the short-term association between NO 2 and hospital admissions for five CVDs (ischaemic heart disease, heart failure, heart rhythm disturbances, ischaemic stroke, haemorrhagic stroke) and total CVDs, and to calculate their attributable fractions.

Study site
Fuyang (114° 52′ to 116° 49′ east longitude, 32° 25′ to 34° 04′ north latitude), located in the northwest of Anhui province, China (Fig. 1). Fuyang has a four-season and temperate monsoon climate. According to China's seventh census in 2020, Fuyang city has a higher proportion of the rural population (58.03% in Fuyang vs. 41.67% in Anhui province and 36.11% in China) (National Bureau of Statistics of China 2020). With the rapid urbanization in recent years, air pollution in Fuyang has also become more prominent.

Health data
The data used in our study were extracted from the NRCMS database from January 2015 to June 2017 in Fuyang. Based on the tentative diagnosis of hospital admissions, five CVDs (ischaemic heart disease (ICD-10: I20-I25), heart failure (ICD-10: I50), heart rhythm disturbances (ICD-10: I47-I49), ischaemic stroke (ICD-10: I63), and haemorrhagic stroke (ICD-10: I60-I61)) were included in this study (Dominici et al. 2006). Given that there are large exporting laborers in rural adults of Fuyang, we excluded the cases hospitalized in areas other than Anhui province.

Environmental data
Air pollution data of five monitoring stations between January 2015 and June 2017 in Fuyang were collected by the local Municipal Environmental Protection Bureau. Twenty-four-hour daily average concentrations of PM 2.5 , PM 10 , NO 2 , CO, SO 2 , and mean maximum 8-h average concentrations of O 3 were calculated. In addition, the daily mean temperature and the relative humidity were extracted from the China Meteorological Data Sharing Service System (https:// data. cma. cn/ en). The daily environmental data of missing days were inputted using mean values of the 2 days before and after the missing day.

Statistical analysis
A two-stage model was applied in our study. In the first stage, a CDLM model was used to assess the association between NO 2 and daily CVD hospital admissions, as bellows: where E(Y t ) was the expected counts of hospital admissions in the tth day; βX t,l was a cross-basis function with the concentrations of NO 2 in the tth day using a CDLM with 4 df polynomials for 6 lag days. Based on previous studies (Xie et al. 2021;Yin et al. 2017), five covariates were adjusted. First, the natural cubic spline (ns t ) function of calendar time was adjusted with 7 df per year to control time trends and seasonality (Chen et al. 2017b). Second, we selected 6 df for 3-day moving average temperature and 3 df for 3-day moving average relative humidity to control the confounding effects of potential nonlinearity and delay . Third day of week and holidays were added as confounding variables. Associations between ambient NO 2 and hospital admissions for total and cause-specific CVDs were assessed separately.
Moreover, RR and 95% CI were calculated for the associations between daily hospital admissions for cause-specific CVDs and per 10 μg/m 3 increase in Log E Y t = + X t,l + ns t (Calendar time, df = 7∕year) + ns t (Temperature, df = 6) + ns t (Relative humidity, df = 3) + Day of week + Holiday NO 2 (single-day lag and cumulative effects). We found that the cumulative effects of NO 2 were stable at lag0 − 2 days. Subgroup analyses by gender, age (18-64, 65-74, and ≥ 75), and seasons (warm and cold) were conducted at lag0-2 days. Specifically, the study period was divided into warm seasons (April to September) and cold seasons (October to March). We used Z test for assessing the differences between subgroups (Altman and Bland 2003).
In the second stage, we calculated attributable number (AN) and attributable fraction (AF) of health outcomes over the study period at lag 0-2 day.
where N i was the annual total counts of health outcomes in year i and β was health outcome-specific coefficients from the first stage. ΔX were the annual mean NO 2 concentration in year i. AF was obtained by the AN divided by total counts of the corresponding health outcome over years (Gasparrini et al. 2012;Tian et al. 2019). Finally, we calculated the 95% CI of AN and AF by using the 95% CI values of the β (Chen et al. 2017a).

Fig. 1 The geographical location of Fuyang city analysed in this study
The R software (version 4.0.5) and R packages dlnm (author: A. Gasparrini) were used, and a two-sided P value < 0.05 was considered statistically significant.

Sensitivity analysis
Sensitivity analyses were performed to assess the stability of the estimates. (1) We used two-pollutant models to adjust for PM 2.5 , PM 10 , O 3 , SO 2 , or CO. (2) Covariates were set at different degrees of freedom to evaluate the robustness of our estimations: 3-5 for polynomials function, 6-9 per year for calendar time, 3-6 for temperature, and 3-6 for relative humidity, respectively. (3) Poisson distribution was set instead of quasi-Poisson distribution to fit the model. (4) The cases hospitalized in areas other than in Anhui province were included. Table 1 presents demographic characteristics of CVD cases enrolled in the NRCMS in Fuyang between January 2015 and June 2017. Overall, this study included 47.10% males, 26.29% older adults aged 75 or more, and 51.24% admissions in cold seasons. A total of 445,216 hospital admissions for total CVDs, 164,004 for ischaemic heart disease, 6,342 for heart rhythm disturbances, 12,064 for heart failure, 244,367 for ischaemic stroke, and 18,441 for haemorrhagic stroke were identified.  Table 2 shows the daily hospital admissions for total and cause-specific CVDs, NO 2 levels, and meteorological variables. Mean hospital admissions per day (standard deviation) were 488.2 (117.1) for total CVDs, 179.8 (45.6) for ischaemic heart disease, 7.0 (3.3) for heart rhythm disturbances, 13.2 (7.2) for heart failure, 267.9 (67.7) for ischaemic stroke, and 20.2 (6.4) for haemorrhagic stroke occurred over the study period, respectively. The annual mean value for NO 2 concentrations was 36.3 (12.4) μg/m 3 .

Results
The Spearman correlations between six air pollutants and meteorological factors are showed in Table S1. Daily NO 2 levels exhibited positive correlations with PM 2.5 , PM 10 , CO, and SO 2 (r = 0.40 ~ 0.53), but negative correlations with O 3 , humidity, and temperature (r = − 0.12 ~ − 0.23). Figure 2 shows the relation between NO 2 and hospital admissions for total and cause-specific CVDs at different lag days. Similar lag results were observed on the effects of NO 2 on total CVDs, ischaemic heart disease, and ischaemic stroke. With per 10 μg/m 3 increase in NO 2 concentrations, the highest estimates in significant increments of the hospital admissions were found at lag0-2 days for total CVDs (RR: 1.019, 95%CI: 1.005 to 1.032), ischaemic heart disease (1.021, 1.006 to 1.036), and ischaemic stroke (1.021, 1.006 to 1.035), but not for heart rhythm disturbances (0.996, 0.963 to 1.030), heart failure (1.018, 0.994 to 1.042), and haemorrhagic stroke (0.986, 0.966 to 1.005) (Table S2). Figure 3 shows the associations between NO 2 levels at lag0-2 days and hospital admissions for total and causespecific CVDs, stratified by gender, age, and seasons. The associations were stronger in older individuals than in young individuals for total CVDs, ischaemic heart disease, and ischaemic stroke, although associations were only marginally significant in some instances. The estimates were consistently higher in females than in males for total CVDs and ischaemic stroke, but lower in females than in males for ischaemic heart disease although the differences were not significant. We found that the relation was also stronger in cold seasons than in warm seasons for total CVDs, ischaemic heart disease, and ischaemic stroke (all P < 0.05). Table 3 presents the results of two-pollutant models at lag0-2 days. The effects of NO 2 were similar with those from single-pollutant models. Table 4 presents the ANs and AFs of all CVD incidence attributable to NO 2 . The ANs were 29,012 (8323 to 48,710) for cardiovascular disease, 11,996 (3594 to 19,962) for ischaemic heart disease, and 17,397 (5218 to 28,946) for ischaemic stroke, with AFs of 6.52% (1.87% to 10.94%) for total CVDs, 7.31% (2.19% to 12.17%) for ischaemic heart disease, and 7.12% (2.14% to 11.85%) for ischaemic stroke, respectively.
In the subgroup analyses by gender, age, and seasons, AN and AF estimates of total CVDs and ischaemic stroke attributable to NO 2 at lag0-2 days in males were more than Fig. 2 RRs and 95% CIs for the associations of daily hospital admissions for cause-specific cardiovascular diseases with per 10 μg/m 3 increase in NO 2 concentrations on lag0-2 days in rural region of Fuy-ang city between January 2015 and June 2017. NO 2 , nitrogen dioxide; RR, relative risk; RR, relative risk; CI, confidence interval in females, while AN and AF estimates of ischaemic heart disease in males were lower than in females. In addition, for total CVDs, ischaemic heart disease, and ischaemic stroke, AN and AF estimates were higher in older adults than in younger adults, and in cold seasons than in warm seasons (Tables S3, S4, and S5). Table S6 presents the results of the sensitivity analyses. The relation between NO 2 and hospital admissions for total CVDs remained unchanged across different df for polynomials function (3-5), calendar time (6-9 per year), temperature (3-6), and relative humidity (3-6). The df for polynomials function was 3-5, and the RR were 1.019 (1.006 to 1.033), 1.019 (1.005 to 1.032), and 1.020 (1.007 to 1.035), respectively. The df for calendar time was 6-9, and the RR were 1.020 (1.006 to 1.033), 1.019 (1.005 to 1.032), 1.021 (1.007 to 1.034), and 1.016 (1.002 to 1.030) respectively. The df for temperature was 3-6, and the RR were 1.020 (1.007 to 1.034), 1.020 (1.006 to 1.033), 1.019 (1.005 to 1.032), and 1.019 (1.005 to 1.032) respectively. The df for relative humidity was 3-6, and the RR were 1.019 (1.005 to 1.032), 1.018 (1.005 to 1.032), 1.019 (1.005 to 1.032), and 1.019 (1.006 to 1.033) respectively. The cases hospitalized in areas other than Anhui province were included, and similar effect estimate was found (RR: 1.018, 95% CI: 1.004 to1.032).

Discussion
In our study, short-term exposure to NO 2 exhibited positive association with increased hospital admissions for total CVDs, ischaemic heart disease, and ischaemic stroke, but did not for heart rhythm disturbances, heart failure, and haemorrhagic stroke. The associations of NO 2 with CVDs were more pronounced in females, in older individuals, and in cold seasons. Moreover, 6.52% of total CVDs were attributable to NO 2 . Our study confirmed and extended previous reports on the positive associations between short-term exposure to NO 2 and hospital admissions for CVDs using data from rural population.

Fig. 3 RRs and 95%
CIs for the associations of daily hospital admissions for cause-specific cardiovascular diseases with per 10 μg/m 3 increase in NO 2 concentrations on lag0-2 days stratified by gender, age, and seasons. NO 2 , nitrogen dioxide; RR, relative risk; RR, relative risk; CI, confidence interval Our study used CDLM model, which has been proved by previous literature to be applicable to the study of air pollutants and health variables in urban and rural regions (Xie et al. 2021;Wu et al. 2022). Cross basis function for the air pollutant in CDLM model, with polynomial of 4 df, was performed according to previous literature. The covariables included calendar time, temperature, relative humidity, public holiday, and day of the week in CDLM model. The df of calendar time, temperature, and relative humidity were 7, 6, and 3 respectively according to the previous literatures (Xie et al. 2021;Wu et al. 2022). Finally, we conducted a sensitivity analysis. The association between NO 2 and hospital admissions for total CVDs did not change with different df for polynomials function (3-5), calendar time (6-9 per year), temperature (3-6), and relative humidity (3-6). The stability of the model suggested that the associations between short-term exposure to NO 2 and increased hospital admissions for some CVDs may be robust.
We observed an increased risk for hospital admissions of total CVDs, ischaemic heart disease, and ischaemic stroke with per 10 μg/m 3 increment of NO 2 in Fuyang rural regions. Our results were similar with most of previous reports on the relation between NO 2 and cardiovascular hospital admissions in urban regions. For instance, the total CVD hospitalizations in the inland city were increased by 2.65% (1.61 to 3.70) per 10 μg/m 3 increase in NO 2 (Liu et al. 2020). The RR of the cardiovascular hospital admissions was 1.005 (1.001 to 1.009) with per 10 μg/m 3 increase in NO 2 at lag 6 day in Iran city (Dastoorpoor et al. 2019). The RR of the cardiovascular emergency room admissions was 1.014 (1.003 to 1.024) with per 10 μg/m 3 increase in NO 2 in Beijing, China (Ma et al. 2017b). However, related studies carried out in rural regions were limited and received inconsistent results. A study conducted in Pingliang, Northwest China, indicated that NO 2 was the most contributor to hospital admission costs of CVDs among six air pollutants (Zha et al. 2022), whereas another study performed in Guangxi province, China, did not find a significant relation between NO 2 and cardiovascular hospital admission . The reasons for the inconsistent results remain unclear. One possible explanation is that the associations between NO 2 and CVDs admission may depend on the NO 2 concentrations. Previous studies have shown that NO 2 has a greater impact on CVDs in areas with high pollution compared to areas with better air quality (Zhou et al. 2022;Ma et al. 2017b). Mean NO 2 concentrations in Guangxi were 22.53 μg/m 3 , lower than recommended concentration of 25 μg/m 3 by the World Health Organization, whereas higher NO 2 concentrations during the study period were reported in Pingliang (34.2 μg/m 3 ) and Fuyang (36.3 μg/m 3 ) (Zha et al. 2022;Zhang et al. 2022). Moreover, higher NO 2 concentrations in Fuyang may also explain the consistent results between our study and those conducted in urban areas. More studies in rural regions are needed to confirm this possibility. There were no significant associations between shortterm NO 2 exposure and heart rhythm disturbances, heart failure, and haemorrhagic stroke in this study. So far, no comparable research conducted in rural regions could be found. In contrast, most of previous studies carried out in urban regions reported positive associations. For instance, the increased risk of outpatient visits for atrial fibrillation were associated with short-term exposure to NO 2 in Xi'an city (Zhou et al. 2022). An interquartile range increase in NO 2 concentrations were related with 1.6% (0.6 to 2.5%) increase in heart failure admissions in 26 largest cities in China . A meta-analysis also showed that per 10 ppb increase in NO 2 were associated with 2.4% (0.3 to 4.5%) increase in haemorrhagic stroke (Shah et al. 2015). Several possibilities could explain the differences between our findings and previous reports from urban population. First, there are higher NO 2 concentrations in urban regions than in rural regions because it mainly sources from automobile exhaust and industry, possibly contributing to higher incidence of NO 2 related cardiovascular diseases in urban population. Second, there exist differences in the availability of medical resources between rural and urban population, leading to possible delays or decreases for cardiovascular hospital admissions in rural regions.
We found that the associations between NO 2 and total CVDs were stronger in females than in males, which were consistent with previous studies. A study in Beijing showed that short-term effects of air pollutions on emergency room admissions for CVDs were more pronounced in females than in males (Ma et al. 2017b). A recent study found that females had higher prevalence of obesity (females/males: 26.6% / 19.6%), dyslipidaemia (61.9%/47.9%), and hypertension (98.1%/94.2%) than males based on 1.7 million adults in China (Lu et al. 2019). The gender differences in the association between NO 2 and total CVDs may be attributed to complex mechanisms and beyond the scope of our research. Slightly greater airway reactivity (Yunginger et al. 1992) and lower socioeconomic status (Zhao et al. 2014) in females than in males may partly explain the sex differences. We also found a stronger relation between NO 2 and CVDs in older individuals than in younger individuals although no significance was found. Similar to our findings, a study conducted in Lanzhou, China, showed that the association between NO 2 and cardiocerebrovascular diseases was pronounced in older adults than in younger adults (Zheng et al. 2013). One possible explanation was that elderly people were considered to be more susceptible to the effects of air pollution than younger groups which their body function were gradually declining over time.
We found a significant association between NO 2 and CVDs in cold seasons rather than in warm seasons. Similar to our results, seasonal differences in the association of air pollutions with CVDs have been reported in previous studies. A study conducted in Shanghai found that cold seasons significantly heightened the effect of NO 2 on CVD hospital admissions (Chen et al. 2010). A recent meta-analysis study provided robust evidence that season modified the associations between air pollution on CVDs (Bergmann et al. 2020). Another review has also shown that the risk of CVDs was slightly higher in winter and inversely correlated with outdoor temperature (Stewart et al. 2017). One possible explanation for the seasonal differences was the higher NO 2 level in cold seasons than in warm seasons. Another potential reason was the interaction of NO 2 and temperatures on CVDs. The fall of temperature could lead to an increase in heart rate and blood pressure via triggering sympathetic activation, platelet activation (Zhang et al. 2004), and cumulative increases of inflammatory blood markers (Halonen et al. 2010), all of which increased susceptibility to air pollution.
The CVD hospital admissions attributable to NO 2 would reach 29,012 cases in the Fuyang rural regions. Our results were consistent with previous report which the national CVD hospital admissions attributable to NO 2 were 0.20 (0.11 to 0.30) million in 2017 (Yao et al. 2020). The AFs of hospital admissions for by CVDs were higher in females, elderly, and cold seasons. This was consistent with the results of the above subgroup analysis. Given that severe air pollution and the huge amount of population in China, protecting populations from NO 2 -related cardiovascular burden is essential.
A major strength of this study was use of NRCMS which nearly covered over the whole farmers in Fuyang city without sampling error. However, several limitations in this study should be noted. First, limited fixed-site air monitoring stations were used in this study to reflect exposures to ambient NO 2 in Fuyang as a whole, which could mask the  . Moreover, monitoring data may lead to non-differential estimate error for individual exposure (Goldman et al. 2011), biasing the effect estimates downward. Second, individual covariates, such as comorbidities, cigarette smoking, alcohol consumption, and physical activities, were not available in our study, thus confounding by these individual factors could not be ruled out.

Conclusions
Our results indicate that short-term increase in NO 2 levels is associated with increased hospital admissions for total CVDs, ischaemic heart disease, and ischaemic stroke in rural population. The associations for NO 2 with CVDs are larger in females, older individuals, and cool seasons. The burden of CVDs attributed to NO 2 is enormous. This is a preliminary study in rural population, and more research is required to replicate our findings.