Associations between Different Ozone Indicators and Cardiovascular Hospital Admission: A Time-Stratified Case-Crossover Analysis in Guangzhou, China

Epidemiological studies reported that ozone (O3) is associated with cardiovascular diseases. However, only few of these studies examined the impact of multiple O3 indicators on cardiovascular hospital admissions. This study aimed to explore and compare the impacts of different O3 indicators on cardiovascular hospital admissions in Guangzhou, China. Based upon the data on daily cardiovascular hospital admissions, air pollution, and meteorological factors in Guangzhou from 2014 to 2018, a time-stratified case-crossover design model was used to analyze the associations between different O3 indicators and cardiovascular hospital admissions. Moreover, the sensitivities of different age and gender groups were analyzed for the whole year and different seasons (i.e., warm and cold). During the warm season, for the single-pollutant model, the odds ratio (OR) value of cardiovascular hospital admissions was 1.0067 (95% confidence interval (CI): 1.0037, 1.0098) for every IQR increase in MDA8 O3 at a lag of five days. The effect of O3 on people over 60 year was stronger than that on the 15–60 years age group. Females were more sensitive than males to O3 exposure. These results provided valuable references for further scientific research and environmental improvement in Guangzhou. Given that short-term O3 exposure poses a threat to human health, the government should therefore pay attention to prevention and control policies to reduce and eliminate O3 pollution and protect human health.


Introduction
Ozone (O 3 ), a secondary pollutant, is formed by the photochemical reactions of precursors that mainly include volatile organic compounds (VOCs) and nitrogen oxides (NOx). It can negatively impact human and animal health, plant growth, climate change, and ecological balance on a local, regional, or even global scale [1][2][3][4]. O 3 pollution poses a huge threat to the global public health, especially in developing countries such as China and India [5,6]. In recent years, rapid economic development and urbanization have been closely related to severe O 3 pollution. Although China implemented an important policy in 2013 (i.e., Air Pollution Prevention and Control Action Plan), the concentration of particulate matter dominated by PM 2.5 (particulate matter with an aerodynamic diameter ≤2.5 µm) shows an obvious decreasing trend, while O 3 concentration shows an increasing trend, gradually becoming the dominant air pollutant in China [7,8]. Notably, the reduction divided into two age groups: 15-60 and 60+ years old. Screening was performed according to the residential address of the patients, who were all residents of Guangzhou.

Air Pollutants and Meteorological Data
The data on pollutants from January 1, 2014 to August 31, 2018, including O 3 (MDA8 O 3 and MDA1 O 3 ), PM 2.5 , and NO 2 , were obtained from the Guangzhou Air Quality Monitoring Station. Among them, PM 2.5 and NO 2 concentrations were used to identify the stability of the relationship between O 3 and cardiovascular admissions in the multipollution model. The pollutant data were collected from the National Environmental Monitoring Center of China (http://quotsoft.net/air/ accessed on 8 July 2022). The 10 air quality monitoring stations in Guangzhou can be divided into three categories, namely, urban, suburban, and mountainous stations, which are located in the densely populated area with high traffic pressure, areas where the industrial base is less developed and the population is relatively sparse, and areas that have high altitude and location in mountainous scenic spots. More importantly, due to data maybe being affected by unpredictable weather conditions, equipment/data transmission failures, etc., the O 3 monitoring data contain a large number of missing values and potential outliers. To guarantee the stability of the results, we adopt a strict quality control of the data according to [24]. For each city, when we have not less than 20 h of data, then the daily O 3 concentration was the average of the hourly data of all monitoring stations in this city [8]. Moreover, if a large amount of data was lost on a day, then the data for that day were considered as invalid and replaced with the average value from the corresponding month [25]. Considering that the hospitalized patients came from different districts in Guangzhou, the average value of pollutants from 10 monitoring stations was thus used in this study.
The meteorological data from 2014 to 2018 were obtained from the China Meteorological Data Sharing Service System (http://data.cma.gov.cn/). Daily average temperature and relative humidity were included to test the sensitivity of the results. All the data were divided into two seasons, warm and cold, according to their monthly average temperature. In particular, months with an average temperature above 20 • C were classified as warm season (May to October) and those below 20 • C were classified as cool season (November to April).

Statistical Analysis
Given that each case serves as its own control, a TSCC design can well control the temporal and seasonal confounders, as well as individual confounders, such as age, gender, and socioeconomic status. Therefore, TSCC is widely used in epidemiological studies [26,27]. Time-stratified refers to the method of comparing the day a patient was hospitalized (case day) with the same day in other weeks within one previous month (control day) [28]. Specifically, if a patient visited on a Tuesday in August 2015, all Tuesdays a month prior would be control days; according to this rule, every case will thus have three to four controls.
In the present study, TSCC was used to evaluate the effect of short-term exposure to different O 3 indicators on cardiovascular hospital admissions. We used conditional logistic regression to generate the odds ratio (OR) values and their 95% CI. If the value of p was less than 0.05, the results were considered statistically significant. Additionally, daily average temperature and relative humidity were involved as covariates. The model is as follows: where t is the date of hospital admission, and X represents the independent variable, including pollutants and meteorological factors. The item log (h (t, X)) represents the risk function of exposure to X on the day t, and log (h 0 (t)) refers to the baseline risk function [28]. C s is the daily concentration of pollutants; and Temp and RH represent average temperature and relative humidity, respectively, with the coefficients β 1 , β 2 , and β 3 . To examine whether O 3 has a lag effect on cardiovascular hospital admissions, we analyzed the impacts of O 3 with a single-day lag model (from lag0 to lag5) and a multi-day lag model (from lag01 to lag05). The pollutant concentration of lag0 refers to the concentration on the current day, and the pollutant concentration of lag1 refers to the O 3 concentration of the previous one day. The pollutant concentration of lag01 represents the 2-day average concentration of the pollutant concentrations on the current day and the previous one day, and the pollutant concentration of lag05 represents the concentration averaged on the current day and the previous five days. In the lag model, we obtained the ORs and optimal lag period for O 3 based on the maximum value of the ORs.
We also performed a series of subgroup analyses stratified by age (15-60 years age group and more than 60 years age group) and gender (male and female) to identify potentially susceptible subgroups. These age stratifications refer to previously published studies [5,29]. Seasonal analysis of O 3 -related effects was also performed by dividing the whole year into warm season (May to October) and cold season (November to April). The effects of different O 3 indicators on daily cardiovascular hospital admissions in Guangzhou were both examined in these subgroup analyses.

Sensitivity Analysis
To further verify the stability of the results, we (1) performed a multi-pollutant analysis, where we added other severe pollutants (e.g., NO 2 and PM 2.5 ) one by one to test whether the results were stable; and (2) we changed the lag days of meteorological factors (i.e., average temperature and relative humidity) from lag0 to lag3 to test whether the results were sensitive to changes in meteorological factors.
All of the above analyses were performed in R software (version 3.6.3). The p-values less than 0.05 were regarded as statistically significant. Notably, PM 2.5 and NO 2 concentrations were higher in the cold season, whereas O 3 was higher in the warm season. The daily concentration range of MDA8 O 3  daily concentration of NO 2 was in the range of 14.0-148.0 µg/m 3 , with an annual average of 46.1 µg/m 3 . During the study period, NO 2 exceeded the level II of the Chinese Ambient Air Quality Standards (40 µg/m 3 ) for 963 days, accounting for 54.93% of the total number of days. The annual average value of PM 2.5 was 38.2 µg/m 3 , which was 46.8% higher than the 35 µg/m 3 threshold recommended by the level II of the Chinese Ambient Air Quality Standards, and it was 99.0% higher than the annual average reported by the WHO (10 µg/m 3 ). Moreover, the daily average temperature and relative humidity ranged from 3.3 • C to 31.7 • C (annual average was 22.2 • C) and from 28% to 97% (annual average was 78.7%), respectively, which were included in the model as confounding factors. Figure 1 shows the temporal evolution of the four pollutants and the two meteorological factors.

Basic Information of Data
was higher in the warm season. The daily concentration range of MDA8 O3 and MDA1 O3 during the study period (2014-2018) were 4.0-271.0 and 4.3-311.1 μg/m 3 , respectively. The daily concentration of NO2 was in the range of 14.0-148.0 μg/m 3 , with an annual average of 46.1 μg/m 3 . During the study period, NO2 exceeded the level Ⅱ of the Chinese Ambient Air Quality Standards (40 μg/m 3 ) for 963 days, accounting for 54.93% of the total number of days. The annual average value of PM2.5 was 38.2 μg/m 3 , which was 46.8% higher than the 35 μg/m 3 threshold recommended by the level Ⅱ of the Chinese Ambient Air Quality Standards, and it was 99.0% higher than the annual average reported by the WHO (10 μg/m 3 ). Moreover, the daily average temperature and relative humidity ranged from 3.3 °C to 31.7 °C (annual average was 22.2 °C) and from 28% to 97% (annual average was 78.7%), respectively, which were included in the model as confounding factors. Figure 1 shows the temporal evolution of the four pollutants and the two meteorological factors.  Figure 2 shows the percentage change in cardiovascular hospital admissions with an IQR increase in MDA8 O3 and MDA1 O3. Figure S1 shows the association between two O3 indicators and the daily cardiovascular hospital admissions in a single pollutant model. Similar lagged effects were observed for the two O3 indicators. For MDA8 O3, significant   Figure S1 shows the association between two O 3 indicators and the daily cardiovascular hospital admissions in a single pollutant model. Similar lagged effects were observed for the two O 3 indicators. For MDA8 O 3 , significant and negative OR values were observed at lag1-lag05, and only the result of lag0 was not statistically significant ( Figure S1a). For MDA1 O 3 , significant and negative OR values of 0.9966 (95% CI: 0.9940, 0.9993), 0.9965 (95% CI: 0.9940, 0.9990), 0.9951 (95% CI: 0.9912, 0.9990), and 0.9947 (95% CI: 0.9907, 0.9989) were observed at lag3, lag4, lag04, and lag05 days, respectively (Figure S1b). and negative OR values were observed at lag1-lag05, and only the result of lag0 was not statistically significant ( Figure S1a). For MDA1 O3, significant and negative OR values of 0.9966 (95% CI: 0.9940, 0.9993), 0.9965 (95% CI: 0.9940, 0.9990), 0.9951 (95% CI: 0.9912, 0.9990), and 0.9947 (95% CI: 0.9907, 0.9989) were observed at lag3, lag4, lag04, and lag05 days, respectively (Figure S1b). The risk of cardiovascular hospital admissions in different age groups was further examined by applying the same model to determine whether there were differences between age groups. As shown in Figure S1a, MDA8 O3 was negatively related to cardiovascular hospital admissions. At the lag of five days, the effect of MDA8 O3 on cardiovascular hospital admissions in the 15-60 years age group was the highest and significant, with an OR value of 0.9999 (95% CI: 0.9952, 0.9909). The highest OR value at the day of lag3 was significant for the 60+ years age group, with an OR value of 0.9960 (95% CI: 0.9929, 0.9991).

The Relationship between Ozone and Cardiovascular Hospitalization
For MDA1 O3, its effect on the 15-60 years age group was significant and the highest at the day of lag4, with an OR value of 0.9946 (95% CI: 0.9901, 0.9991), while its effect on the 60+ years age group had no significant association ( Figure S1b).
The OR estimates for different gender groups were further analyzed using the same model ( Figure S2). For MDA8 O3, at the lag of three days, the effect of MDA8 O3 on cardiovascular hospital admissions in the male group was the highest and significant, with an OR value of 0.9965 (95% CI: 0.9931, 0.9998). For the female group, the highest and significant association occurred at the day of lag 2, with an OR value of 0.9951 (95% CI: 0.9913, 0.9989) ( Figure S2a).
For MDA1 O3, no significant association was found for the male group. Meanwhile, for the female group, significant associations were found at the days of lag3 and lag4, with OR values of 0.9958 (95% CI: 0.9918, 0.9997) and 0.9951 (95% CI: 0.9913, 0.9989), respectively ( Figure S2b). Figure 3 shows the percentage change in cardiovascular hospital admissions with an IQR increase in MDA8 O3 and MDA1 O3 for different gender people. The risk of cardiovascular hospital admissions in different age groups was further examined by applying the same model to determine whether there were differences between age groups. As shown in Figure S1a, MDA8 O 3 was negatively related to cardiovascular hospital admissions. At the lag of five days, the effect of MDA8 O 3 on cardiovascular hospital admissions in the 15-60 years age group was the highest and significant, with an OR value of 0.9999 (95% CI: 0.9952, 0.9909). The highest OR value at the day of lag3 was significant for the 60+ years age group, with an OR value of 0.9960 (95% CI: 0.9929, 0.9991).
For MDA1 O 3 , its effect on the 15-60 years age group was significant and the highest at the day of lag4, with an OR value of 0.9946 (95% CI: 0.9901, 0.9991), while its effect on the 60+ years age group had no significant association ( Figure S1b).
The OR estimates for different gender groups were further analyzed using the same model ( Figure S2). For MDA8 O 3 , at the lag of three days, the effect of MDA8 O 3 on cardiovascular hospital admissions in the male group was the highest and significant, with an OR value of 0.9965 (95% CI: 0.9931, 0.9998). For the female group, the highest and significant association occurred at the day of lag 2, with an OR value of 0.9951 (95% CI: 0.9913, 0.9989) ( Figure S2a). For MDA1 O 3 , no significant association was found for the male group. Meanwhile, for the female group, significant associations were found at the days of lag3 and lag4, with OR values of 0.9958 (95% CI: 0.9918, 0.9997) and 0.9951 (95% CI: 0.9913, 0.9989), respectively ( Figure S2b). Figure 3 shows the percentage change in cardiovascular hospital admissions with an IQR increase in MDA8 O 3 and MDA1 O 3 for different gender people.

Seasonal Estimates of the Effects of Different Ozone Indicators on Cardiovascular Hospital Admissions
The continuous high temperature and strong sunshine in the warm season are conducive to the photochemical reaction of VOCs and NOx, which will accelerate the formation of O3. That is, there is a phenomenon that the O3 concentration in summer is higher than that in winter. Therefore, we evaluated the effect of different O3 indicators on cardiovascular hospital admissions in the warm season (May to October) and the cold season (November to April) (Figs. S3 and S4, respectively). The results showed that O3 had different effects on cardiovascular hospital admissions in different seasons. Specifically, a significant and positive association between MDA8 O3/MDA1 O3 and cardiovascular hospital admissions in the warm season and a negative association in the cold season.

Seasonal Estimates of the Effects of Different Ozone Indicators on Cardiovascular Hospital Admissions
The continuous high temperature and strong sunshine in the warm season are conducive to the photochemical reaction of VOCs and NOx, which will accelerate the formation of O 3 . That is, there is a phenomenon that the O 3 concentration in summer is higher than that in winter. Therefore, we evaluated the effect of different O 3 indicators on cardiovascular hospital admissions in the warm season (May to October) and the cold season (November to April) ( Figures S3 and S4, respectively). The results showed that O 3

Seasonal Estimates of the Effects of Different Ozone Indicators on Cardiovascular Hospital Admissions
The continuous high temperature and strong sunshine in the warm season are conducive to the photochemical reaction of VOCs and NOx, which will accelerate the formation of O3. That is, there is a phenomenon that the O3 concentration in summer is higher than that in winter. Therefore, we evaluated the effect of different O3 indicators on cardiovascular hospital admissions in the warm season (May to October) and the cold season (November to April) (Figs. S3 and S4, respectively). The results showed that O3 had different effects on cardiovascular hospital admissions in different seasons. Specifically, a significant and positive association between MDA8 O3/MDA1 O3 and cardiovascular hospital admissions in the warm season and a negative association in the cold season.   During the warm season, similar lagged effects on cardiovascular hospital admissions were observed for the two different O3 indicators. MDA8 O3 was significantly and positively related to all the people from lag4 and lag5 days, with OR values of 1.0043 (95% CI: 1.0011, 1.0075) and 1.0067 (95% CI: 1.0037, 1.0098), respectively ( Figure S3). For the moving average lag model, the results were not significant. As for MDA1 O3, there was also a positive association at a lag of four and five days, with the OR values of 1.0034 (95% CI: 1.0001, 1.0067) and 1.0057 (95% CI: 1.0024, 1.0089), respectively ( Figure S4). In the cold season, MDA8 O3/MDA1 O3 was significantly and negatively correlated with cardiovascular hospital admissions ( Figure S3 and S4).  During the warm season, similar lagged effects on cardiovascular hospital admissions were observed for the two different O3 indicators. MDA8 O3 was significantly and positively related to all the people from lag4 and lag5 days, with OR values of 1.0043 (95% CI: 1.0011, 1.0075) and 1.0067 (95% CI: 1.0037, 1.0098), respectively ( Figure S3). For the moving average lag model, the results were not significant. As for MDA1 O3, there was also a positive association at a lag of four and five days, with the OR values of 1.0034 (95% CI: 1.0001, 1.0067) and 1.0057 (95% CI: 1.0024, 1.0089), respectively ( Figure S4). In the cold season, MDA8 O3/MDA1 O3 was significantly and negatively correlated with cardiovascular hospital admissions ( Figure S3 and S4). We further examined the OR value in different seasons by using the same model to determine whether there were differences between different seasons. As shown in Figure  4, MDA8 O3 was positively associated with the cardiovascular hospital admissions in the warm periods. At the lag of five days, the effect of MDA8 O3 on cardiovascular hospital admissions in the 15-60 years age group was the highest and significant, with an OR value During the warm season, similar lagged effects on cardiovascular hospital admissions were observed for the two different O 3 indicators. MDA8 O 3 was significantly and positively related to all the people from lag4 and lag5 days, with OR values of 1.0043 (95% CI: 1.0011, 1.0075) and 1.0067 (95% CI: 1.0037, 1.0098), respectively ( Figure S3). For the moving average lag model, the results were not significant. As for MDA1 O 3 , there was also a positive association at a lag of four and five days, with the OR values of 1.0034 (95% CI: 1.0001, 1.0067) and 1.0057 (95% CI: 1.0024, 1.0089), respectively ( Figure S4). In the cold season, MDA8 O 3 /MDA1 O 3 was significantly and negatively correlated with cardiovascular hospital admissions ( Figure S3 and S4).
We further examined the OR value in different seasons by using the same model to determine whether there were differences between different seasons. As shown in Figure 4, MDA8 O 3 was positively associated with the cardiovascular hospital admissions in the warm periods. At the lag of five days, the effect of MDA8 O 3 on cardiovascular hospital admissions in the 15-60 years age group was the highest and significant, with an OR value of 1.0059 (95% CI: 1.0004, 1.0115). Meanwhile, significant and positive associations for the 60+ years age group were found at the days of lag3 to lag5, with OR values of 1.0040 (95% CI: 1.0000, 1.0081), 1.0051 (95% CI: 1.0013, 1.0090), and 1.0070 (95% CI: 1.0033, 1.0107), respectively ( Figure S3a).
For MDA1 O 3 , its significant effect on the 60+ years age group was observed at the days of lag4 and lag5, with OR values of 1.0043 (95% CI: 1.0002, 1.0083) and 1.0061 (95% CI: 1.0022, 1.0101), respectively. Meanwhile, its effect on the 15-60 years age group had no significant association ( Figure S4a). For MDA8 O 3 , significant associations were found for the male group at the days of lag3 to lag5, with OR values of 1.0064 (95% CI: 1.0020, 1.0108), 1.0068 (95% CI: 1.0026, 1.0110), and 1.0066 (95% CI: 1.0025, 1.0107), respectively. Meanwhile, for the female group, the significant and highest association was found at the day of lag5, with an OR value of 1.0069 (95% CI: 1.0023, 1.0106) ( Figure S3b). For MDA1 O 3 , significant associations were found for the male group at the days of lag4 and lag5, with OR values of 1.0057 (95% CI: 1.0012, 1.0101) and 1.0050 (95% CI: 1.0007, 1.0093), respectively. Meanwhile, for the female group, the significant and highest association was found at the day of lag5, with an OR value of 1.0064 (95% CI: 1.0016, 1.0114) ( Figure S4b).

Sensitivity Analysis
Sensitivity analyses demonstrated that these results are stable by implementing the single-and multi-pollutant model. In the multi-pollutant model analysis, after PM 2.5 was added, the OR values of different O 3 indicators decreased slightly, but still without statistical significance (purple color in Figure 6). For the MDA8 O 3 , after adding NO 2 or adding PM 2.5 and NO 2 simultaneously, the change in the effect of O 3 on cardiovascular hospital admissions was small, with OR values of −0.0312%, −0.0015%, −0.0698%, and −0.0078%, respectively. For MDA1 O 3 , the values were 0.0049%, 0.0297%, −0.0207, and 0.0178, respectively (purple color in Figure 6).
To further verify the stability of the results, we likewise adjusted the different lag days for meteorological factors. The results showed that when the lag days of meteorological factors were 0, 1, 2, and 3 days, the OR values changed only slightly, which for MDA8 O 3 were −0.031%, −0.017%, −0.007% and −0.008%, respectively (pink color in Figure 6). The changed OR values for the MDA1 O 3 were 0.005%, 0.010%, 0.016%, and 0.015%, respectively (pink color in Figure 6). The very minimal change in results indicated that these results are stable. To further verify the stability of the results, we likewise adjusted the different lag days for meteorological factors. The results showed that when the lag days of meteorological factors were 0, 1, 2, and 3 days, the OR values changed only slightly, which for MDA8 O3 were −0.031%, −0.017%, −0.007% and −0.008%, respectively (pink color in Figure  6). The changed OR values for the MDA1 O3 were 0.005%, 0.010%, 0.016%, and 0.015%, respectively (pink color in Figure 6). The very minimal change in results indicated that these results are stable.

Discussion
As far as we know, this is one of the few studies that reported the effect of different O3 indicators on daily cardiovascular hospital admissions. In this study, we employed a TSCC design to evaluate the short-term effect of MDA8 O3/MDA1 O3 on daily cardiovascular hospital admissions in Guangzhou from 2014 to 2018. The results showed that the associations between cardiovascular hospital admissions and O3 varied with different O3 indicators, age, gender groups, and season. The association between MDA8 O3/MDA1 O3 and daily cardiovascular hospital admissions was significantly positive in the warm season and negative in the cold season. Compared with MDA1 O3, MDA8 O3 was more strongly associated with cardiovascular hospital admissions. The effects of O3 on cardiovascular hospital admissions were stronger in the 60+ years age group than in the 15-60 years age group, and females were slightly more sensitive than males to O3. These findings not only help us understand the differences in impact of different O3 indicators on cardiovascular hospital admissions, but also present the differences in the effect of O3 on different groups.
Three O3 indicators (MDA8 O3, MDA1 O3, and daily 24 h average O3) are commonly used in O3-related epidemiological studies. In this study, we reviewed studies about the effects of different O3 indicators on cardiovascular disease worldwide since 2000 ( Table 2) and found that different studies used different O3 indicators. For example, in the United States, daily all-cause, cardiovascular, and respiratory mortality are closely associated with MDA8 O3 levels [30]. Based on a generalized additive model, Qin et al. (2017) evaluated the relationship between daily O3 and cardiovascular mortality in 2013-2015 [31]. Their results showed that MDA8 O3 was positively associated with cardiovascular mortality. Gryparis et al. (2004) reported that, in Europe, a 10 μg/m 3 increase in MDA1 O3 was associated with a 0.33%, 0.45%, and 1.13% increase in all-cause, cardiovascular, and respiratory mortality, respectively [32]. The daily 24 h average O3 indicator is also widely

Discussion
As far as we know, this is one of the few studies that reported the effect of different O 3 indicators on daily cardiovascular hospital admissions. In this study, we employed a TSCC design to evaluate the short-term effect of MDA8 O 3 [31]. Their results showed that MDA8 O 3 was positively associated with cardiovascular mortality. Gryparis et al. (2004) reported that, in Europe, a 10 µg/m 3 increase in MDA1 O 3 was associated with a 0.33%, 0.45%, and 1.13% increase in all-cause, cardiovascular, and respiratory mortality, respectively [32]. The daily 24 h average O 3 indicator is also widely used in the previous epidemiological studies [17,20,33] [37]. Their results showed that MDA8 O 3 and MDA1 O 3 were positively related to daily cardiovascular hospital admissions, and the association was stronger for MDA8 O 3 than for MDA1 O 3 . The potential reason may be that the MDA8 O 3 indicator is the most relevant to individual exposure levels, as people come into cities for work during the daytime and return home at night. Furthermore, high O 3 exposure during the daytime may have more negative health impacts than commute period and nighttime [38]. Li et al. (2015) also demonstrated that the health effects caused by short-term O 3 exposure was more related to higher O 3 concentration, such as MDA8 O 3 , rather than peak concentration, such as MDA1 O 3 [38].
The strong solar radiation and higher temperature in summer are helpful to the photochemical reactions of NOx and VOCs, resulting in the production of the secondary pollutant O 3 . Therefore, there is a high O 3 concentration in summer and low O 3 concentration in winter [8,39]. According to the characteristics of O 3 , we evaluated the effects of two O 3 indicators on daily cardiovascular hospital admissions in Guangzhou in different seasons: warm and cold. Notably, there are different associations in the warm and cold seasons, which was consistent with the previous studies. We also examined the susceptibility of different age and gender groups to O 3 pollution during the warm season and the cold season. The results showed that, in the warm season, both MDA8 O 3 and MDA1 O 3 were significantly and positively related to daily cardiovascular hospital admissions. By contrast, in the cold season, the two O 3 indicators were significantly and negatively related to daily cardiovascular hospital admissions. That is, in the warm season, there are positive associations between MDA8 O 3 /MDA1 O 3 and daily cardiovascular hospital admissions, whereas in the cold season, there are negative associations between them, results that are consistent with other studies [2,40]. Similarly, Yang et al. (2004) aimed to use a case-crossover method to determine the association between O 3 concentration and cardiovascular hospital admissions in Taiwan [21]. Their results showed that O 3 was positively associated with cardiovascular hospital admissions in the warm season, as well as negatively in the cold season. Based upon O 3 data from 2014 to 2017, Lu et al. (2022) used cardiovascular disease emergency data collected from famous hospitals in Shanghai to explore the relationship between O 3 and cardiovascular hospital visits in different seasons [41]. Their results also showed that O 3 had different relationships with cardiovascular hospital visits in different seasons. The underlying reason may be that temperature modulates the impacts of O 3 on the health risks, and some studies have found that the health risks related to O 3 exposure are enhanced in low temperature environment [42,43]. In addition, temperature has a long-term lag effect on human health, especially low temperatures, which are delayed and may last for 2-3 weeks [44]. Therefore, the modifying effects of temperature on the O 3cardiovascular mortality association may also result in insufficient control for temperature effects and increased O 3 mortality risks [45].
In addition, O 3 pollution had a lagged effect on daily cardiovascular hospital admissions. Compared to the single-day lag model, O 3 showed similar effects on the results of the moving average lag model for different age and gender subgroups. These results suggested a cumulative effect of O 3 pollutants on cardiovascular hospital admissions, which was similar to the previous studies [2,38,46].
There are some limitations in this study that should be mentioned. First of all, the data on air pollution in this study were collected from monitoring stations. During the study period, these may be affected by distance, climate, and other factors and could not fully reflect individual exposure, which may cause a certain impact on the estimated association between O 3 and cardiovascular hospital admissions. Second, we did not consider some individual behavioral factors in this study, that is, we did not consider individual-level O 3 exposure (i.e., are they smokers or not, educational background, body mass, or occupation); thus, some uncertainty may be involved. Third, our hospital admissions data include all cardiovascular and cerebrovascular diseases. The absence of detailed data about hypotension disease or the use of data that were not divided into more detailed disease subgroups may introduce some uncertainty in the results. Fourth, we do not consider the long-term trend in our model, which also may introduce some uncertainties. Finally, as our cardiovascular hospital admissions data were obtained from a large general hospital, and not all hospitals in Guangzhou, the sample size is lower than that in other epidemiological studies. Hence, care should be taken when generalizing the results to other regions.

Conclusions
This study provided evidence to examine the impact of different O 3

Data Availability Statement:
The data analyzed in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare that they have no known competing financial interest or personal relationships that could have appeared to influence the work reported in this paper.