Short-term ozone exposure significantly elevates paediatric respiratory disease risk in Foshan, Southern China

Abstract

Background

Although an association between ozone (O₃) exposure and respiratory disease (RD) in children has been established, there is significant heterogeneity in global epidemiological evidence. The aim of this study was to fill a regional scientific gap by evaluating the impact of short-term O₃ exposure on hospitalisations for RD among children in Foshan, a city in southern China.

Methods

This ecological time-series study used electronic medical records from a hospital in Foshan between 1 July 2021 and 30 June 2024, which included hospitalised children aged 0–14 years with a primary diagnosis of all RD classified according to the International Classification of Diseases 10th Revision (J00–J99). The distributed lag nonlinear model was applied to evaluate the association between O₃ exposure and RD in children.

Results

A total of 3944 hospitalisations for paediatric RD were enrolled, including 2218 boys (56.2%) and 1726 girls (43.8%). With the 2.5th percentile of O₃ concentration as the reference, a 10 μg/m³ increase in O₃ was associated with a peak relative risk at lag days 6 and 7, with relative risk of 1.034 (lag day 6: 95% CI 1.014 to 1.054; lag day 7: 95% CI 1.016 to 1.051). Stratified analyses revealed no significant effect modification by season or sex in the association between O₃ exposure and paediatric RD.

Conclusion

Short-term O₃ exposure significantly increased the risk of paediatric RD hospitalisations in Foshan. These results support the need for region-specific interventions to mitigate health risks from O₃ in children.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Ozone exposure has been associated with respiratory problems in children; however, regional variations contribute to uncertainty, particularly in southern China.

WHAT THIS STUDY ADDS

• This study indicates that short-term ozone exposure significantly increases the risk of hospitalisation for respiratory diseases among children in southern China and reveals distinct lag effects in this association between boys and girls.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This study highlights the importance of developing region-specific public health strategies to reduce paediatric ozone exposure and may inform the development of prevention policies and the refinement of air quality standards for childhood respiratory diseases in urban areas.

Background

Childhood respiratory diseases (RDs) pose a critical global health challenge. Lower respiratory tract infections remain the leading cause of death among children under 5 years,¹ and asthma affects millions worldwide with marked geographic variation.² In China, paediatric hospitalisations for RD surged by 23% after the 2023 pandemic.³ Collectively, these findings underscore the substantial and persistent burden of paediatric RD.

Numerous epidemiological studies have examined the association between ozone (O₃) exposure and childhood RD,^{4 5} but findings are inconsistent. Recent systematic reviews have consistently shown that ambient O₃ exposure is a risk factor for childhood RD even at concentrations below the current WHO air quality standards.^{6 7} For instance, a 10 μg/m³ increase in O₃ was associated with a 2% higher risk of paediatric respiratory hospitalisations in southern Europe⁸ and a 31% increased risk of childhood asthma in a 14-city Chinese study.⁵ However, other studies reported no significant association with lung function growth⁹ or respiratory symptoms.¹⁰ In addition, a few experimental studies have reported protective effects of low-dose, short-term O₃ exposure in animal or therapeutic settings via hormesis.^{11 12} Given these epidemiological inconsistencies and the divergence between experimental hormetic findings and population-based evidence, further region-specific investigations are imperative to inform local public health strategies.

This study focuses on Foshan, a city that remains among the five lowest-ranked in Guangdong Province for air quality despite recent improvements. Although a previous multicentre study across 14 Chinese cities provided valuable insights into the association between O₃ and asthma,⁵ it was limited to a single disease outcome and did not assess the impact of O₃ exposure on other RD. To address these gaps, we investigated the association between O₃ exposure and all respiratory system diseases (International Classification of Diseases, 10th Revision (ICD-10) codes J00–J99) as a composite outcome. Using a distributed lag non-linear model (DLNM), we assessed whether O₃ exposure increases the overall risk of paediatric hospitalisations for RDs in Foshan and identified critical exposure windows.¹³ This study provides region-specific evidence for Foshan, thereby addressing the limitations of the existing literature.

Methods

Study location

The study was conducted in Foshan city (23°02’N, 113°06’E), which is located in central-southern Guangdong Province, within the Pearl River Delta and serves as a vital hub in the Guangdong-Hong Kong-Macao Greater Bay Area. Foshan encompasses a total area of 3797.72 km² and has a resident population of approximately 9.557 million. Foshan is a major industrial hub in the subtropical monsoon Guangdong-Hong Kong-Macao Greater Bay Area, where O₃ predominates in autumn and summer.¹⁴ These features make it an ideal setting to examine region-specific O₃ related health effects.

Data collection

This ecological time-series study used electronic medical records from the Department of Paediatric Inpatient Services at Guangdong General Hospital’s Nanhai Hospital. The unit of analysis was the day, with exposure variables assigned at the population level. Children aged 0–14 years admitted between 1 July 2021 and 30 June 2024 with a primary diagnosis of RD (ICD-10 codes J00–J99) were included. Patients were excluded if they had incomplete data, non-respiratory primary diagnoses or were transferred to intensive care units within 24 hours of admission to avoid potential confounding by critical illness severity. Daily aggregates of admissions were calculated for analysis. The study was approved by the Ethics Committee of Nanhai Hospital of Guangdong Provincial People’s Hospital (no. 2023302H). Informed consent was waived due to the use of anonymised aggregate data.

Ambient 24-hour average concentrations of fine particulate matter (PM_2.5), inhalable particulate matter (PM₁₀), sulfur dioxide (SO₂), carbon monoxide (CO), nitrogen dioxide (NO₂) and daily maximum 8-hour average concentration of O₃ for the Foshan city area were obtained from the National Urban Air Quality Real-time Release Platform, administered by the China National Environmental Monitoring Centre. During this study period, a total of eight monitoring stations compliant with national standards were operational within the city’s administrative boundaries. City-level exposure metrics were calculated as the arithmetic mean of daily concentrations across all available stations, in accordance with the Ambient Air Quality Standards (GB 3095–2012). This approach reflects the regional average exposure level, consistent with the study’s spatial scale. A publicly accessible preprocessed database is available at https://quotsoft.net/air/, with daily updates. Additionally, meteorological data (mean wind speed, relative humidity and mean temperature) for the same period were obtained from the Xihe Energy Meteorology Big Data Platform (https://www.xihe-energy.com). These variables were derived as the regional average across Foshan’s administrative boundaries, based on the NASA (National Aeronautics and Space Administration) historical reanalysis dataset.

Statistical analysis

We employed a DLNM framework to examine the non-linear and lagged effects of O₃ exposure on paediatric respiratory morbidity. The model is formalised as follows:

$${\displaystyle {\displaystyle {\displaystyle \begin{array}{rl}\log \left(E\left[Y_t\right]\right)& =\alpha +\sum _{l=0}^{14}cb\left(o_{3,t-l};\beta \right)\\ +ns\left(\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e},7\times df/\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}\right)+{\mathrm{D}\mathrm{O}\mathrm{W}}_t+{\mathrm{H}\mathrm{o}\mathrm{l}\mathrm{i}\mathrm{d}\mathrm{a}\mathrm{y}}_t\\ +\sum _{k=1}^{K}ns\left(z_{k,t},3\right)+{ϵ}_t\end{array}}}}$$

where $E\left[Yt\right]$ represents the expected daily count of paediatric respiratory admissions on day t, $\mathrm{\alpha }$ is the intercept and $\mathrm{c}\mathrm{b}\left({\mathrm{o}}_{3,\mathrm{t}-\mathrm{l}};\mathrm{\beta }\right)$ denotes the exposure-lag-response function constructed via natural cubic splines with 3 df for the exposure dimension and 4 df for the lag dimension (maximum lag=14 days). Sensitivity analyses were performed by varying the exposure df (2–4) and lag df (3–5), with model fit assessed using the quasi-Akaike information criterion. The combination of exposure df=3 and lag df=4 was selected as the main model, balancing model fit with parsimony (detailed results are provided in online supplemental table S1). Adjustments included long-term temporal trends via a natural spline of time (21 df, ie, 7 df/year×3 years), categorical variables for day of the week (DOW) and holidays and non-linear effects of meteorological factors (temperature, relative humidity and wind speed) with natural splines (3 df).

All analyses used the 2.5th percentile of O₃ concentration as the reference for estimating the exposure-response function. The primary effect estimate was the relative risk (RR) associated with per 10 μg/m³ increase in O₃ concentration from this reference value. The RR for a specific O₃ concentration ‘c’ was derived as:

$${\displaystyle {\displaystyle {\displaystyle \mathrm{R}\mathrm{R}(c)=\mathrm{e}\mathrm{x}\mathrm{p}\left(\sum _{l=0}^{14}{\hat{\beta }}_l\cdot cb\left(c;l\right)\right)}}}$$

where ${\widehat{\mathrm{\beta }}}_{\mathrm{l}}$ are the estimated coefficients from the cross-basis matrix. Overdispersion was assessed via Pearson residuals. Based on previous studies,¹⁵ May–September was defined as the warm period and October–April as the cold period. Stratified analyses were conducted to examine the associations separately by season (warm vs cold) and by sex (boys vs girls). To test whether the effect modification by season or sex was statistically significant, we used likelihood ratio F-tests comparing models with and without the cross-basis × season interaction (season main effect included in both) and similarly for cross-basis × sex interaction.

To assess robustness, we performed three sensitivity analyses all referenced to the same 2.5th percentile: (1) the RR at the 97.5th percentile; (2) two-pollutant models separately adding PM_2.5, PM₁₀, SO₂ or NO₂, as well as a full model including all co-pollutants (PM_2.5, PM₁₀, SO₂ and NO₂) and (3) the RR per IQR increase in O₃, with the IQR calculated over the entire study period. All analyses were performed using R (V.4.5.0) with the dlnm and splines packages.

Patient and public involvement

Members of the public were not involved in the design and conduct of the surveys because we conducted this study directly analysing existing data.

Results

A total of 3944 hospitalisations for paediatric RD were enrolled, including 2218 boys (56.2%) and 1726 girls (43.8%). The daily mean number of hospitalisations for RD, O₃, PM_2.5, PM₁₀, SO₂, NO₂, temperature, relative humidity and wind speed were 3.6, 60.9 µg/m³, 21.94 µg/m³, 39.26 µg/m³, 6.76 µg/m³, 29.42 µg/m³, 23.2 °C, 77.73% and 3.38 m/s, respectively (table 1). Online supplemental figure S2 shows the monthly trends of O₃ concentrations and RD admissions, revealing distinct temporal patterns between the two variables.

Table 1. Summary of daily respiratory disease-related hospital admissions, atmospheric pollutants and meteorological variables in Foshan, 2021–2024.

Variable	Mean	SD	Min	P 2.5	P 25	P 50	P 75	P 97.5	Max	IQR
RD	3.6	2.81	0	0	1	3	5	10	13	4
O3	60.9	29.45	6	14.06	39.58	56.25	78.69	128.35	168.53	39.1
PM2.5	21.94	12.36	1.74	6.27	12.06	19.5	29.26	53.6	69.5	17.2
PM10	39.26	21.35	3.52	11.11	22.39	34.77	51.93	93.1	124.17	29.5
SO2	6.76	1.61	3.54	4.14	5.67	6.46	7.54	10.67	13.5	1.88
NO2	29.42	13.99	6.5	10.7	19.38	26.53	35.71	67.97	86.71	16.3
Temp	23.2	5.82	5.21	10.12	19.08	24.97	28.2	30.05	31.37	9.12
RH	77.73	10.21	40.32	54.11	72.14	79.37	85.39	92.64	95.23	13.2
WS	3.38	1.55	0.71	1.32	2.27	3.04	4.14	7.25	10.85	1.87

Children aged 0–14 years were included in this study.

IQR, interquartile range; Max, maximum; Min, minimum; NO₂, nitrogen dioxide; O₃, ozone; P_2.5, 2.5th percentile; P₂₅, 25th percentile; P₅₀, 50th percentile; P₇₅, 75th percentile; P_97.5, 97.5th percentile; PM₁₀, particulate matter with aerodynamic diameter of ≤10μm; PM_2.5, particulate matter with aerodynamic diameter of ≤2.5μm; RD, respiratory disease; RH, relative humidity; SD, standard deviation; SO₂, sulfur dioxide; Temp, temperature; WS, wind speed.

As shown in online supplemental figure S1, O₃ exhibited weak to moderate positive correlations with PM_2.5 (r=0.45) and PM₁₀ (r=0.50), whereas PM_2.5 was strongly positively correlated with PM₁₀ (r=0.95, p<0.05) and moderately associated with SO₂ (r=0.67) and NO₂ (r=0.66). Temperature and relative humidity were negatively associated with PM_2.5 (r = -0.28 and r = -0.52, respectively) and PM₁₀ (r=−0.26 and r=−0.57, respectively), with humidity also showing a negative correlation with SO₂ (r=−0.55). Wind speed exhibited weak negative correlations with multiple pollutants.

Using the 2.5th percentile of O₃ concentration as the reference, the cumulative risk of RD over lag 0–14 days increased with increasing O₃ concentration, with a particularly sharp increase observed at concentrations exceeding 100 μg/m³ (figure 1a). A 10 µg/m³ increase in O₃ concentration was associated with a significantly elevated risk of RD from lag day 3 to lag day 10. The risk peaked at lag days 6 and 7, with RR of 1.034 (lag day 6: 95% CI 1.014 to 1.054; lag day 7: 95% CI 1.016 to 1.051) (figure 1b).

Figure 1. Associations between ambient O₃ exposure and the risk of hospitalisation for respiratory disease. (a) Exposure-response curve for the cumulative effect of O₃ exposure over lags 0–14 days on the risk of hospitalisation for respiratory disease. (b) Relative risk of hospitalisation for paediatric respiratory disease associated with a 10 μg/m³ increase in O₃, with the 2.5th percentile as the reference. Children aged 0–14 years were included in this study. O₃, ozone; RR, relative risk.

Online supplemental table S2 shows that both O₃ concentrations and RD admission counts differed significantly between the warm and cold seasons. With the 2.5th percentile O₃ concentration as the reference, a 10 μg/m³ increase in O₃ was not significantly associated with RD risk in both seasons (figure 2). A quasi-likelihood ratio F test for seasonal interaction confirmed no significant modification of the O₃–RD association (F = 1.114, p =0.344). Furthermore, for both seasons, the cumulative RRs at the 25th, 50th, 75th and 97.5th percentiles showed no statistically significant association with RD risk (online supplemental table S3).

Figure 2. Seasonal variation in the RR of hospitalisation for paediatric respiratory disease per 10 μg/m³ increase in O₃, with the 2.5th percentile as the reference. Error bars indicate 95% CIs. Children aged 0–14 years were included in this study. O₃, ozone; RR, relative risk.

Sex-stratified analyses showed that per 10 μg/m³ increase in O₃, the highest RR for boys occurred at lag day 5 (RR 1.039, 95% CI 1.008 to 1.071), whereas for girls the peak RR was observed at both lag days 8 and 9, with identical RR of 1.044 (95% CI 1.016 to 1.074 for lag day 8 and 1.011 to 1.078 for lag day 9) (figure 3). However, a quasi-likelihood ratio F test for sex interaction confirmed no significant modification (F = 0.705, p = 0.748).

Figure 3. Sex-stratified relative risk of hospitalisation for paediatric respiratory disease per 10 μg/m³ increase in O₃, with the 2.5th percentile as the reference. Error bars indicate 95% CIs. Children aged 0–14 years were included in this study. O₃, ozone; RR, relative risk.

To assess the robustness of the primary findings sensitivity analyses were conducted using the 2.5th percentile O₃ concentration as the reference. The on-the-day RR at the 97.5th percentile of O₃ peaked at lag day 7 (RR 1.083, 95% CI 1.036 to 1.132) (online supplemental figure S3a), and the cumulative RR at the 97.5th percentile over lag 0–14 days became significantly elevated from lag day 6 to lag day 14, with a maximum of 1.958 (95% CI 1.299 to 2.952) at lag day 12 (online supplemental figure S3b). Across different adjustment strategies, this cumulative RR ranged from 1.733 to 2.131 (online supplemental figure S4a). Using the IQR (IQR=39.1 µg/m³) as an alternative exposure metric (reference: the 2.5th percentile), the lag-specific risks per IQR increase peaked at lag days 6 (RR 1.080, 95% CI 1.034 to 1.127) (online supplemental figure S4b).

Discussion

Summary of main findings

This study provides the first systematic assessment of the association between O₃ exposure and paediatric RD hospitalisations in Foshan. The findings indicate a non-linear exposure-response relationship between short-term O₃ exposure and the risk of RD admissions. Per 10 µg/m³ increase in O₃ concentration was associated with an RR of 1.034 for RD admissions at the highest risk lag days 6 and 7 (lag day 6: 95% CI 1.014 to 1.054; lag day 7: 95% CI 1.016 to 1.051). Moreover, the cumulative risk at high O₃ concentration (97.5th percentile) reached 1.958 times (95% CI: 1.299 to 2.952) that at low concentration (2.5th percentile) by lag day 12. Stratified analyses revealed no significant effect modification by season or sex in the association between O₃ exposure and paediatric RD.

Potential biological mechanisms

O₃ exposure affects childhood RD through multiple interrelated biological mechanisms. As a potent oxidant, O₃ directly damages airway epithelial cells by inducing membrane lipid peroxidation and DNA damage, leading to oxidative stress.⁴ This damage activates signalling pathways and triggers the release of inflammatory mediators, thereby amplifying airway inflammation.^{16 17} Both short-term and long-term O₃ exposure promote infiltration of inflammatory cells (eg, neutrophils and eosinophils), whose subsequent release of cytokines and chemokines further exacerbates inflammation.^{18 19} O₃ exposure also induces airway smooth muscle contraction and remodelling, increasing airway hyperresponsiveness, sensitivity and reactivity, which aggravates asthma symptoms and respiratory distress.^{20 21} Genetic susceptibility modulates disease severity, with specific variants increasing vulnerability to O₃-induced oxidative stress and inflammation.²² Finally, O₃ compromises immune function by damaging epithelial cells, impairing defence mechanisms and amplifying allergic responses, thereby raising RD risk.²³

Comparison with other studies

Earlier studies have highlighted the adverse effects of O₃ exposure on children’s respiratory health. Long-term O₃ exposure has been associated with reduced lung function growth in Mexican schoolchildren,²⁴ as well as airway inflammation, increased respiratory symptoms and decreased lung function in Greek children.²⁵ A Dutch study on physically active children found a slight negative correlation between postexercise peak expiratory flow and the previous day’s maximum O₃ concentration, suggesting a delayed effect.²⁶ In Portugal, O₃ concentration increases over a 30-day lag period were linked to higher cumulative risks for paediatric emergency visits and hospital admissions.⁸ Similarly, a Canadian study reported that same-day O₃ increases were associated with higher emergency department visit rates among children with lower RD.²⁷ These findings support a positive association between O₃ exposure and the risk of paediatric RD, which is consistent with our results. This concordance likely reflects shared biological mechanisms driven by O₃’s potent oxidising properties. However, discrepancies exist with respect to the lag interval: some studies observed effects on the day of exposure or the following day, whereas in the present study the RR peaked at approximately 1 week of lag, suggesting a longer delayed effect. These differences may stem from heterogeneity in paediatric populations (eg, age, health status, susceptibility, lifestyle), climatic conditions, as well as differences in exposure assessment and statistical methodologies.

Studies have shown that the association between O₃ and RD exhibits seasonal variation, with generally stronger effects in warmer seasons.²⁸ A meta-analysis found a 0.84% (95% CI 0.09 to 1.59%) increase in chronic obstructive pulmonary disease hospitalisations per 10 µg/m³ increase in O₃, particularly during warmer seasons.²⁹ In Hanoi, each 10 µg/m³ increase in O₃ was associated with a 6.2% increase in paediatric respiratory hospitalisations in the cold season versus 1.2% in the warm season.³⁰ By contrast, this study showed that O₃ exposure was not significantly associated with RD hospitalisation risk in either the warm or cold season individually. This discrepancy may be attributable to several factors. First, the definition of the warm season and the cold season adopted in the present study differs from that used in previous investigations. Second, the outcome of interest was hospitalisation for all RD, whereas earlier studies predominantly focused on specific conditions; different disease subtypes may exhibit varying seasonal responses to O₃ exposure. Accordingly, future research should define seasons based on local climatic characteristics and conduct stratified analyses by specific subtypes of RD, in order to more accurately assess seasonal variations in the health effects of O₃.

A study in Taiwan region of China revealed that long-term O₃ exposure more significantly impaired lung function development in boys than in girls.³¹ Similarly, research in Maine revealed that each 10 ppb O₃ increase was associated with a 7% (95% CI 4 to 11) higher asthma-related emergency department visit risk, particularly among males under 15 years of age.³² These findings appear to contrast with the present study, in which short-term O₃ exposure significantly increased hospitalisation risks for RD in both sexes, but the effect did not differ significantly between boys and girls. This discrepancy may be attributed to two factors. First, outcome definitions varied: the present study assessed overall childhood RD, whereas previous studies focused on lung function development or asthma-related emergency visits. Second, variations in regional population characteristics may influence the manifestation of sex-specific effects. Consequently, future research should investigate sex differences in the effects of O₃ across specific RD subtypes to clarify the nature of sex modification. In addition, sensitivity analyses using alternative exposure metrics and different pollutant adjustment strategies consistently supported the robustness of these associations.

Limitations

This study has several limitations. First, its single-city focus constrains generalisability, although Foshan represents a typical Pearl River Delta industrial city. Variations in climate, pollution profiles, demographics, healthcare access and socioeconomic factors across regions may modify O₃ health effects. Second, fixed-site monitoring cannot capture individual actual exposure (activity patterns, microenvironments, indoor O₃ sources), likely causing nondifferential misclassification and attenuating effect estimates. Third, despite adjusting for multiple copollutants and meteorological factors, complex pollutant interactions (synergistic/antagonistic) and residual confounding cannot be fully excluded. Fourth, data are lacking on individual protective behaviours, allergen exposure, indoor environment (dampness, mould) and sociodemographic factors, which may act as confounders or effect modifiers. Fifth, due to insufficient sample size, exploratory subtype analyses cannot generate stable estimates for specific RD categories, precluding conclusions on effect heterogeneity (upper vs lower respiratory tract). Future studies should expand geographic scope, incorporate personalised exposure assessment, multipollutant interaction analysis and collect key covariates to enhance generalisability and accuracy.

Conclusions

This study demonstrates that O₃ exposure significantly increases the risk of paediatric hospitalisations for RD among children in Foshan. These findings provide new evidence of the adverse effects of O₃ on children’s respiratory health and underscore the importance of considering temporal patterns in environmental health assessments. However, given the study’s limitations, future research should validate these findings in broader regions and populations and explore additional influencing factors and underlying mechanisms.

Data availability statement

Data are available upon reasonable request.