Temperature Modification of Ambient Ozone Association with Outpatient Visits for Atherosclerotic Cardiovascular Diseases

Abstract

Scarce evidence exists on the impact of ozone (O₃) on atherosclerotic cardiovascular disease (ASCVD) morbidity in tropical urban settings, and whether temperature modifies this effect remains unclear. To bridge this gap, we assessed the association between ambient O₃ and ASCVD outpatient visits, and the potential effect modification by temperature, in Haikou, China. A time-series analysis was performed on data from 163,348 daily hospital outpatient visits for ASCVD collected between 1 January 2020 and 31 December 2022. The association between O₃ exposure and daily visits was evaluated with an over-dispersed Poisson generalized additive model (GAM), and the modifying effect of temperature was scrutinized using a nonparametric bivariate response surface model. A 10 μg/m³ increment in the daily maximum 8 h average concentration of O₃ was associated with a 1.35% (95% CI: 0.63, 2.07) increase in ASCVD outpatient visits at lag0. Stratified analyses revealed that the association between O₃ and ASCVD visits was only significant during the warm season, with stronger effects observed above 30 °C, peaking at 34 °C (lag06). The combined exposure to high temperature and O₃ concentrations significantly amplified ASCVD outpatient visits. Ambient O₃ exposure was associated with increased ASCVD outpatient visits in the tropical city, and this risk was enhanced under high temperatures. These results highlighted the importance of considering temperature interaction in O₃-related risk assessments.

1. Introduction

Cardiovascular diseases (CVD) continue to be the leading cause of morbidity and mortality worldwide, particularly in China [1,2,3]. According to the 2022 Cardiovascular Health and Disease Report in China, the estimated number of people affected by CVD reached 330 million, concurrently highlighting a notable increase in the incidence of atherosclerotic cardiovascular disease (ASCVD). ASCVD is a major and growing component of CVD, which refers to a spectrum of conditions driven by atherosclerosis, including acute coronary syndrome, myocardial infarction, angina pectoris, ischemic stroke, transient ischemic attack, and peripheral artery disease [4]. While behavioral and metabolic risk factors are well-established, ambient air pollution has been increasingly recognized as a critical environmental determinant of CVD [5].

Among air pollutants, tropospheric ozone (O₃) has become a particularly pressing concern, with concentrations rising steadily in many regions, even surpassing particulate matter as the primary pollutant in some Chinese cities [6,7,8]. Current WHO Air Quality Guidelines, informed by meta-analyses linking O₃ to non-accidental mortality, recommend a long-term, peak-season standard of 60 μg/m³ and a short-term level of 100 μg/m³ [9]. The understanding of ozone’s health effects has significantly evolved, shifting the focus from initially observed cardiopulmonary pathological changes to issues encompassing an increased risk of cardiopulmonary morbidity and mortality, a paradigm shift greatly facilitated by the deepened awareness of the exposure–response relationship [9]. Accumulating epidemiological evidence indicates that both short- and long-term exposure to O₃ is associated with elevated CVD mortality, even at concentrations below current regulatory standards. Furthermore, pathophysiological studies support the biological plausibility of the relationship between O₃ exposure and CVD morbidity and mortality [10].

Despite this progress, important knowledge gaps remain. Most epidemiological studies have focused on O₃-associated cardiovascular mortality, with limited and inconsistent evidence regarding morbidity endpoints, especially in outpatient settings [11,12,13,14,15,16,17,18,19,20]. In China, where outpatient visits substantially exceed inpatient and emergency care volumes, outpatient data represent a highly sensitive metric for early, sub-acute health effects. However, their application in O₃ health studies has received limited attention [21]. Furthermore, regional and seasonal variations in O₃ levels have been widely reported, with higher concentrations typically occurring during warm seasons due to favorable photochemical conditions [6,7,10]. Similarly, the incidence of CVD exhibits seasonal patterns influenced by climatic factors, particularly ambient temperature [22,23]. Climate change is increasing the frequency of co-occurring heatwaves and high O₃ episodes, raising the possibility of synergistic health effects [24]. However, it remains unclear whether and how temperature modifies this association, notably in tropical urban environments characterized by prolonged high temperatures. This study aimed to investigate the association between ambient O₃ exposure and outpatient visits for ASCVD in the tropical city of Haikou, and to specifically examine whether and how temperature modifies this association.

2. Data Collection and Analysis Methods

2.1. Data Collection

This time-series study utilized outpatient visit records for atherosclerotic cardiovascular diseases (ASCVD; ICD-10 codes: I20–I25, I63, I73.9, G45, Z95.1, Z95.5) from public hospitals in Haikou, covering the period from 1 January 2020 to 31 December 2022. The data were provided by the Statistics Information Center of Hainan Provincial Health Commission. Records from individuals residing outside Haikou were excluded. Demographic variables, including sex, age, and visit date, were extracted for analysis.

Daily ambient air pollutant concentrations—including 24 h average levels of fine particulate matter (PM_2.5), inhalable particles (PM₁₀), sulfur dioxide (SO₂), nitrogen dioxide (NO₂), carbon monoxide (CO), and the daily maximum 8 h average O₃—were obtained from the China Air Quality Online Monitoring and Analysis Platform (https://www.aqistudy.cn, accessed on 1 December 2023). Concurrent meteorological data, such as daily temperature (℃) and relative humidity (%), were sourced from the Qingyue Open Environmental Data Center (https://data.epmap.org, accessed on 26 December 2023).

2.2. Statistical Analyses

The association between environmental exposures and cardiovascular outcomes was commonly evaluated using time-series analysis [13,25,26]. Spearman correlation analysis was first conducted to assess the relationships among environmental variables. A three-stage analytical approach was implemented to comprehensively assess the association between ambient ozone (measured as the daily maximum 8 h average) and ASCVD outpatient visits, and to evaluate the potential modifying effect of temperature.

In the first stage, an over-dispersed Poisson generalized additive model (GAM) was employed to estimate the overall association between O₃ and ASCVD outpatient visits. The model incorporated multiple covariates to control for temporal patterns, meteorological influences, and other potential confounders, including calendar time, daily mean temperature, relative humidity, public holidays, and day of the week (DOW). The model structure was specified as follows:

$$\begin{array}{cc}\hfill \mathrm{Log}\left(\mathrm{E}\left[{\mathrm{Y}}_{\mathrm{t}}\right]\right)=\mathsf{\alpha }+& \mathsf{\beta }{\mathrm{O}}_{3,\mathrm{t}}+\mathrm{n}\mathrm{s}(\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e},{\mathrm{d}\mathrm{f}}_1)+\mathrm{n}\mathrm{s}(\mathrm{t}\mathrm{e}\mathrm{m}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{e},{\mathrm{d}\mathrm{f}}_2)\hfill \\ \hfill & +\mathrm{n}\mathrm{s}(\mathrm{r}\mathrm{e}\mathrm{l}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{h}\mathrm{u}\mathrm{m}\mathrm{i}\mathrm{d}\mathrm{i}\mathrm{t}\mathrm{y},{\mathrm{d}\mathrm{f}}_3)+\mathrm{D}\mathrm{O}\mathrm{W}+\mathrm{H}\mathrm{o}\mathrm{l}\mathrm{i}\mathrm{d}\mathrm{a}\mathrm{y}\hfill \\ \hfill & =\mathsf{\beta }{\mathrm{O}}_{3,\mathrm{t}}+\mathrm{n}\mathrm{s}(\mathrm{t}\mathrm{e}\mathrm{m}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{e},{\mathrm{d}\mathrm{f}}_2)+\mathrm{C}\mathrm{O}\mathrm{V}\mathrm{s}\hfill \end{array}$$

(1)

Here, E(Yt) denoted the expected number of ASCVD outpatient visits on day t. The regression model included an intercept (α) and a coefficient (β) that represented the log of the relative risk (RR) of daily ASCVD outpatient visits associated with a 10 μg/m³ increase in ambient 8 h maximum O₃ concentration. The natural spline function, ns(), was used to model nonlinear covariates. Degrees of freedom (df) for time trends (8 per year), temperature, and relative humidity (both 5 per year) were selected based on previous studies [13,27] and the Akaike Information Criterion. COVs encompassed other adjusting variables included in the model.

To assess delayed effects, both single-day lagged model from the current day up to the previous 7 days (lag 0–lag 7) and multi-day lagged model (moving averages of the current and previous days (lag 01–lag07) were constructed. Stratified analyses were performed by gender (male/female), age group (<65/≥65 years), and season (cold: October-March; warm: April–September). Several sensitivity analyses were conducted to evaluate the robustness of the primary findings, including varying the df values of calendar time (df = 7–10), temperature (df = 3–6), and relative humidity(df = 3–6); implementing two-pollutant models to assess confounding effects; and visualizing the exposure–response relationship between O₃ and ASCVD visits using GAM-derived curves.

In the second stage, a distributed lag nonlinear model (DLNM) with cross-basis functions was applied to capture the potentially nonlinear and lagged effects of temperature on outpatient visits over a maximum lag of 30 days, accounting for prolonged meteorological influences [28]. The model was expressed as:

$$\mathrm{Log}\left(\mathrm{E}\left[{\mathrm{Y}}_{\mathrm{t}}\right]\right)=\mathsf{\beta }{\mathrm{O}}_{3,\mathrm{t}}+{\mathsf{\beta }}_1\mathrm{C}\mathrm{b}.\mathrm{T}\mathrm{e}\mathrm{m}\mathrm{p}+{\mathrm{C}\mathrm{O}\mathrm{V}}_{\mathrm{S}}$$

(2)

where β1 represented the regression coefficient of the continuous temperature, Cb.Temp referred to the cross-basis matrix of temperature. The COVs matched those in Equation (1).

In the third stage, a nonparametric bivariate response surface model was applied to visually and statistically explore the potential interaction between O₃ and temperature on ASCVD visits, avoiding strong assumptions regarding the shape of these associations [25]. The model was specified as:

$$\mathrm{Log}\left(\mathrm{E}\left[{\mathrm{Y}}_{\mathrm{t}}\right]\right)=\mathrm{S}\mathrm{T}\left(\mathrm{t}\mathrm{e}\mathrm{m}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{e},{\mathrm{O}}_3\right)+{\mathrm{C}\mathrm{O}\mathrm{V}}_{\mathrm{S}}$$

(3)

where ST() indicated a thin-plate spline function, and COVs remained consistent with model 1.

All statistical analyses were performed using R software (version 4.3.2) with the MGCV and DLNM packages. The effect estimates and 95% confidence interval (CI) were showed as percentage changes converted from relative risks in daily hospital outpatient visits associated with each 10 µg/m³ increment in ambient O₃ concentration. p value < 0.05 was considered statistically significant.

3. Results

3.1. Study Population and Environmental Factor Characteristics

A total of 163,348 outpatient visits for atherosclerotic cardiovascular disease (ASCVD) were recorded in public hospitals in Haikou between 1 January 2020, and 31 December 2022. Among these, 96,137 (58.90%) were male and 102,421 (62.7%) were aged 65 years or older. The mean daily number of outpatient visits was 149 (standard deviation [SD]: 76) (

Table 1. Descriptive statistics on daily hospital outpatient visits and environmental factors during 2020–2022.

Variables	Mean	SD	Min	Median	Max	IQR
Daily ASCVD Outpatient Visits
Total	149	76	6	138	349	124
Male	88	47	3	79	229	77
Female	61	30	2	57	142	46
<65 years	56	30	3	52	139	50
≥65 years	93	48	2	85	221	76
Cold season	161	82	6	155	349	138
Warm season	137	68	11	122	325	99
O3(μg/m3)
Total	80.0	31.3	20.0	74.0	192.0	40.3
Cold season	92.8	31.7	28.0	90.0	192.0	42.5
Warm season	67.3	25.2	20.0	63.0	180.0	30.0
Temperature (℃)
Total	24.7	4.6	7.4	25.5	34.0	7.0
Cold season	21.5	3.8	7.4	21.8	29.5	5.1
Warm season	28.0	2.6	15.3	28.4	34.0	2.6
Relative humidity (%)
Total	81.4	9.1	47	82.2	100	12
Cold season	82.4	9.8	47	84.1	100	12
Warm season	80.5	8.3	53	80.1	100	12

Total: summary statistics for the entire study period.

). The ambient O₃ concentration was 80 μg/m³, among them exceeded the World Health Organization (WHO) global air quality guideline (AQG) level (100 μg/m³, Global update 2021) in 250 days (22.8%) during 2020–2022. Other ambient air pollutants showed only minor elevations and remained below global air quality standards (Table S1). Both ASCVD visits and O₃ concentrations were higher during the cold season compared to the warm season. The mean ambient temperature throughout the study period was 24.7 °C, averaging 21.5 °C in the cold season and 28 °C in the warm season. Notably, 102 days in the warm season had average temperatures exceeding 30 °C. The mean relative humidity was 81.4%, with minimal seasonal variation.

Figure 1A displayed daily ASCVD outpatient visits and environmental factors from 2020 to 2022. Generally, there appeared to be a correlation between the variations in concentrations of air pollutants and the number of outpatient visits for ASCVD. All air pollutants exhibited obvious seasonal dependence and the concentrations were higher in the cold season. The number of daily outpatient visits for ASCVD showed a wide range but, in general, remained steady until 2022 when a slight increase started to occur.

As shown in Figure 1B, the ambient O₃ was positively correlated with other pollutants, but negatively correlated with temperature and relative humidity. A significant negative correlation was observed between daily mean temperature and O₃ concentration (r = −0.44, p < 0.001). This correlation remained weakly negative in the cold season (r = −0.19, p < 0.001) but was not significant in the warm season (r = −0.07, p = 0.099).

3.2. Association of O₃ Exposure with ASCVD Outpatient Visits

Single-pollutant models revealed significant positive associations between O₃ exposure and ASCVD outpatient visits across all lag days (Figure 2A). For single-day lags, the peak values occurred at lag0 with an increase of 1.35% (95% CI: 0.63, 2.07). For multi-day lags, a 10 μg/m³ increment in O₃ levels was associated with an increment in ASCVD outpatient visits up to 1.64% (95% CI: 0.77, 2.53) for lag06 (the 7 d moving average, i.e., the moving average of the same day and previous 6 days exposure to O₃), 1.68% (95% CI: 0.79, 2.58) for lag07 (the 8 d moving average, i.e., the moving average of the same day and previous 7 day exposure to O₃).

Specifically, a curved shape correlation was observed between the ambient O₃ on lag0 and the number of outpatient visits for ASCVD, indicated that higher concentrations of O₃ were associated with a greater number of outpatient visits. Meanwhile, the curve shape reached a peak at O₃ concentration of 135 μg/m³ (Figure 2B).

To ensure the robustness of the results, a series of analyses were performed. After sensitivity analyses, the results remained consistent by adjusting degrees of freedom for temporal trends (7–10 per year), temperature (3–6 per year), and relative humidity (3–6 per year) (Table S2). Furthermore, the daily average concentrations of PM_2.5, PM₁₀, SO₂, NO₂ or CO in the two-pollutant model were adjusted, and the estimated effect of O₃ on ASCVD was still statistically significant (

Table 2. Relative risk (%) and 95% CI of outpatient visits for ASCVD in two-pollutant models (O₃, lag0).

Model	RR (95%CI)
O3	1.014 (1.007, 1.021)
O3 + PM2.5	1.014 (1.005, 1.022)
O3 + PM10	1.014 (1.005, 1.022)
O3 + NO2	1.012 (1.005, 1.019)
O3 + SO2	1.013 (1.005, 1.020)
O3 + CO	1.011 (1.004, 1.019)

), indicating that the relationship between ambient O₃ and ASCVD outpatient visits was stable and robust in the main model.

Subgroup analyses by sex, age, and season were presented in Figure 2C. Estimated risks were higher among males than females. Short-term O₃ exposure significantly affected individuals aged ≥65 years at lag0, while stronger effects at lag06 and lag07 were observed in those under 65 years. Additionally, a significant association was found only during the warm season, suggesting a potential modifying effect of temperature. Consequently, we further examined the lagged and interactive effects of temperature and O₃.

3.3. Relationships Between Ambient Temperature and ASCVD Outpatient Visits

Significant exposure–lag–response relationships were observed between daily mean temperature and ASCVD visits (Figure 3). Temperature exhibited both immediate and delayed health effects, which varied across temperature ranges. At high temperatures (above 30 °C), the relative risk of ASCVD visits increased after a 4-day lag, peaking at 34 °C with a 6-day lag. In contrast, cold temperatures (below 16 °C, recorded on 41 days over the three years) showed weaker but prolonged effects, lasting up to 27 days.

3.4. The Impact of Combined O₃-Temperature on ASCVD Outpatient Visits

Figure 4 illustrated the combined effects between ambient O₃ and temperatures on ASCVD outpatient visits using three-dimensional visualization graphs. Concurrent exposure to high temperature and high O₃ levels was associated with a pronounced increase in ASCVD outpatient visits. Notably, with high temperature exposure, O₃ levels even below WHO standards were still associated with escalated ASCVD outpatient visits.

4. Discussion

This time-series study provided evidence that exposure to ambient O₃ was associated with increased risk of ASCVD outpatient visits in a tropical coastal city. Notably, ambient temperature might play a crucial modification role in O₃-associated elevation of ASCVD.

Haikou, a tropical coastal city in southern China with abundant sunlight, has experienced increasing ambient O₃ concentrations in recent years. During the study period, O₃ emerged as the dominant air pollutant in the region, exceeding the WHO air quality guideline (100 μg/m³) on more than 250 days. The city exhibited a distinct seasonal O₃ pattern influenced by its monsoonal climate. In contrast to the typical summer peaks observed in inland regions, this city experienced elevated O₃ concentrations from October to December [29]. This unique pollution and seasonal profile established Haikou as an ideal setting for investigating temperature-O₃ interactions on health outcomes. We observed a positive association between O₃ exposure and ASCVD outpatient visits. An increase of 10 µg/m³ in O₃ concentration was associated with a 1.35% (95% CI: 0.63, 2.07) increase in visits at lag0, while the strongest cumulative effect was observed at lag07, with a 1.68% (95% CI: 0.79, 2.58). The exposure–response relationship exhibited a curved shape, indicating increasing risk with higher O₃ concentrations. When compared with previous studies, the effect estimates showed certain consistencies as well as variations in magnitude and lag structure. For example, a multi-city analysis in China reported risk increases for cardiovascular diseases ranging from 0.40% to 0.75% at lag01 [13], while another study reported an adjusted odds ratio of 1.16 (95% CI: 1.08, 1.25) for high 10-year ASCVD risk per 1 μg/m³ O₃ increase [30]. In contrast, some studies have reported null or even negative associations between O₃ and specific cardiovascular outcomes such as stable ischemic heart disease or acute coronary syndrome [11,27]. These discrepancies might reflect differences in regional pollution profiles, study populations, or methodological approaches.

Epidemiological evidence has consistently linked high temperatures to increased risks of cardiovascular morbidity and mortality, including elevated rates of emergency department visits, hospital admissions, and fatalities [31,32,33]. For instance, a meta-analyses indicated a 3.44% rise in cardiovascular mortality per 1 °C increase in ambient temperature above the optimal temperature threshold for a certain region [34]. Another meta-analysis of 23 studies showed a relative risk of 1.016 for myocardial infarction hospitalization per 1 °C increase in ambient temperature [35]. In the present study, the association between O₃ and ASCVD outpatient visits was significantly stronger only during the warm season. This finding was consistent with a nationwide cohort study (n = 96,955), which reported a hazard ratio of 1.093 (95% CI: 1.046, 1.142) for overall cardiovascular diseases per 10 μg/m³ increase in warm-season O₃ concentrations [17]. In contrast, some studies yielded opposite results showing that O₃ had an effect on CVDs during the cold period (October-March) [36]. These discrepancies might arise from regional differences in climate, pollutant mixtures, population susceptibility, or healthcare-seeking behavior, warranting further investigation. Although the independent effects of heat and O₃ were well-documented, their joint effects remained inadequately explored. Most existing studies have focused on mortality endpoints rather than morbidity, and the limited available evidence is inconsistent [25,37,38,39,40,41,42,43]. For example, Shi et al. found that high temperatures amplified the effect of O₃ on cardiovascular mortality by 0.42% (95% CI: 0.32, 0.51) per 10 μg/m³ [42]. Our findings extended this evidence to outpatient morbidity, demonstrating synergistic effects under concurrent exposure to high temperature and O₃. Notably, even at low O₃ pollution, high temperatures were associated with increased ASCVD visits, underscoring a growing climate-related health burden.

Several mechanisms may underlie the modification of O₃-related ASCVD risk by temperature. O₃ exposure induces airway inflammation and oxidative stress, leading to systemic inflammation, endothelial dysfunction, and autonomic nervous system imbalance [44,45]. These pathways collectively contribute to atherosclerotic progression and plaque instability. High temperatures may exacerbate these mechanisms by increasing cardiac workload, promoting dehydration, and altering vascular tone [24]. The combined exposure likely accelerates pathological processes, particularly in vulnerable individuals with pre-existing cardiovascular conditions.

Several limitations should be considered. Firstly, the use of aggregated data limited our ability to control for individual-level confounders such as socioeconomic status, behavioral factors, or medication use. Secondly, exposure misclassification may exist due to the use of fixed-site monitoring data rather than personal measurements. Thirdly, although we adjusted for major temporal and meteorological factors, residual confounding may persist, particularly from unmeasured variables during the COVID-19 pandemic (2020–2022), which could have influenced both air quality patterns and healthcare-seeking behavior. Thus, valuable future directions would include panel studies, spatially resolved exposure assessments, the use of individual-level data, and multi-city comparisons across diverse climatic zones to further validate and generalize these findings.

Despite these limitations, our study had several notable strengths. First, rigorous data cleaning and verification procedures were implemented to ensure high data quality. Second, both single-pollutant and multi-pollutant models were employed, along with sensitivity analyses, to compare and validate the results, thereby strengthening the robustness of the observed associations. Third, as a pioneering investigation conducted under a unique climatic conditions where O₃ was the sole air pollutant that consistently exceeded WHO standards, this study provided novel insights into the combined adverse effects of ambient O₃ and temperature on cardiovascular diseases.

5. Conclusions

In summary, our study demonstrated a positive association between ambient ozone pollution and an increased risk of outpatient visits for ASCVD during the warm season. Furthermore, concurrent exposure to high temperatures and O₃ pollution was found to exert a synergistic effect, leading to a greater increase in ASCVD-related outpatient visits. These findings provide crucial evidence for the temperature-dependent nature of ozone-related cardiovascular morbidity, particularly relevant for tropical urban populations. From a public health perspective, our results underscore the urgent need for integrated heat-ozone health warning systems and combined mitigation strategies. As climate change continues to intensify both air pollution challenges and extreme heat events, coordinated policies addressing these interacting environmental stressors will be essential for protecting cardiovascular health in vulnerable communities. Future public health interventions should consider these compound risks through early warning systems that simultaneously address heat and ozone exposure, while emission control strategies should account for their potential climate co-effects.