Warm season ambient ozone and children’s health in the USA

Abstract

Background

Over 120 million people in the USA live in areas with unsafe ozone (O₃) levels. Studies among adults have linked exposure to worse lung function and higher risk of asthma and chronic obstructive pulmonary disease (COPD). However, few studies have examined the effects of O₃ in children, and existing studies are limited in terms of their geographic scope or outcomes considered.

Methods

We leveraged a dataset of encounters at 42 US children’s hospitals from 2004–2015. We used a one-stage case-crossover design to quantify the association between daily maximum 8-hour O₃ in the county in which the hospital is located and risk of emergency department (ED) visits for any cause and for respiratory disorders, asthma, respiratory infections, allergies and ear disorders.

Results

Approximately 28 million visits were available during this period. Per 10 ppb increase, warm-season (May through September) O₃ levels over the past three days were associated with higher risk of ED visits for all causes (risk ratio [RR]: 0.3% [95% confidence interval (CI): 0.2%, 0.4%]), allergies (4.1% [2.5%, 5.7%]), ear disorders (0.8% [0.3%, 1.3%]) and asthma (1.3% [0.8%, 1.9%]). When restricting to levels below the current regulatory standard (70 ppb), O₃ was still associated with risk of ED visits for all-cause, allergies, ear disorders and asthma. Stratified analyses suggest that the risk of O₃-related all-cause ED visits may be higher in older children.

Conclusions

Results from this national study extend prior research on the impacts of daily O₃ on children’s health and reinforce the presence of important adverse health impacts even at levels below the current regulatory standard in the USA.

Key Messages.

• Ambient ozone levels are associated with higher rates of children’s emergency department visit for any cause

• An increase of 10 ppb in ambient ozone levels is associated with a higher risk of emergency department visits for ear disorders, allergies and asthma

• Importantly, these associations are observed even at levels below the current regulatory standard of 70 ppb, suggesting that the current standard may not be sufficient to protect children under 18 years of age

Background

Ozone (O₃), a secondary pollutant formed by compounds in the presence of heat and ultraviolet radiation, poses a significant health concern given widespread exposure across the globe.^1,2 O₃ is highly reactive and may cause significant harm to cells within the respiratory tract and other mucosal surfaces where it can lead to inflammation, hyperresponsiveness and increased mucous secretion.^3–9 Short-term exposure to O₃ is associated with higher risk of emergency department (ED) visits for asthma, respiratory infections and exacerbations of chronic obstructive pulmonary disease (COPD).^10–13

Despite lower O₃ concentrations in recent decades, a large number of US residents (120+ million) live in areas exceeding the US National Ambient Air Quality Standard (NAAQS) for daily maximum 8-hour O₃ of 70 parts per billion [ppb].¹⁴ Higher summertime temperatures under climate change combined with increased presence of volatile organic compounds (VOCs) may attenuate future pollution reduction, even if emissions remain constant.^15,16

Certain groups may be particularly susceptible to effects of O₃ exposures.^17–20 In particular, the US Environmental Protection Agency has identified children as a vulnerable group and has identified factors that may increase the risk of high O₃ exposure in children.²¹ Children’s respiratory and immune systems are not fully developed until 7–8 years of age, increasing susceptibility by limiting their ability to fight infections and recover from inflammatory damage.^22,23 At the same ambient levels, children may experience greater exposure to ozone compared to adults given greater time spent outdoors and faster respiratory rates.^24–26 Thus, the current NAAQS acceptable level of O₃ may not be protective for children.

Much of the research on children’s O₃ exposure has been limited to regional studies and to respiratory symptoms.^27–31 However, the ear, nose and eyes are important mucosal membranes in the body that may be susceptible to inflammation caused by O₃ exposure and warrant investigating related outcomes (e.g. allergies and ear disorders).³² Our study leverages a database from US paediatric hospitals, enabling a large-scale study to assess effects of O₃ on multiple health outcomes in children across the USA. We sought to describe associations between O₃ and ED visits from 2004–2015 for all causes and causes with mechanisms susceptible to O₃ damage for all exposure levels as well as exposure to concentrations below the NAAQS threshold (70 ppb). Selected causes of ED visits include allergies, ear disorders, respiratory disorders, asthma and respiratory infections. We further investigate effect modification by subgroups to identify potential heterogeneity. We selected outcomes based on published literature and potential for health response related to children’s physiology, especially mechanisms related to oxidative stress, inflammation and vulnerability of mucous membranes.^32–34

Methods

We obtained de-identified healthcare utilization claims data on individuals <18 years of age receiving care between 1 January 2004 and 30 September 2015 at US children’s hospitals reporting data to the Pediatric Health Information System (PHIS, https://www.childrenshospitals.org/phis). The PHIS database includes information regarding ED, surgical, inpatient and observation unit encounters in US children’s hospitals in 27 states. During this time period, data were available from 42 children’s hospitals. For each encounter, we extracted information on service date, hospital, diagnosis (based on the International Classification of Disease [ICD-9]), age, sex and insurance status. Patient address and race/ethnicity data were not available for this analysis.

Our analysis includes ED visits for all causes, allergies, ear disorders, respiratory disorders, asthma and respiratory infections (Table 1). These outcomes are mutually exclusive except for asthma (included in respiratory disorders with a subset in allergy [493.00]) respiratory infections (included in respiratory disorders), and a subset of allergies (477.XX, included in both allergy and respiratory categories). The number of daily visits was aggregated by specific cause and subgroups for age (0–5 y, 6–12 y, 13–17 y), sex (male, female) and insurance status (private, government, other insurance).

Table 1.

Children’s hospital emergency department (ED) visits for all-cause and specific outcomes, May–September, 2004–2015

Health outcome	ICD-9 codes	Number of ED visits	% Total visits (%)
All causes	001–999, E000–E999, V01–V91	11 171 779	—
Allergy	372.05, 372.14, 477.XX	57 938	0.52
Ear disorders	380–382	520 741	4.66
Respiratory disorders	460–519	1 693 358	15.16
Asthma	493	405 320	3.63
Respiratory infections	460–466	978 729	8.76

ICD-9, International Classification of Diseases, ninth revision.

We estimated exposure to air pollution and weather at county of treating hospital, applying estimates of daily maximum 8-hour O₃ described in prior work.³⁵ We calculated population-weighted average daily maximum 8-hour O₃ in each county using methods similar to those described by Spangler et al.³⁶ We analogously estimated population-weighted county daily mean temperature and relative humidity (RH) based on the 4-km PRISM model.³⁷ We define warm season as May-September and cold season as October-April.

We used a 1-stage time-stratified case-crossover design to estimate associations between daily maximum 8-hour O₃ and ED visits.³⁸ Cases were aggregated by calendar day for each admitting hospital and case days were matched to three or four referent days matching on hospital, calendar month, year and day of week. This design minimizes potential confounding by factors that are time-invariant or vary slowly over time.³⁹ We used conditional logistic regression in the context of a distributed lag non-linear model (DLNM) to quantify associations between O₃ and risk of ED visits.^40–42 The DLNM provides a flexible yet robust statistical framework to estimate associations with potentially non-linear exposure-response functions and/or lagged effects. We modelled O₃ exposure-response functions using a linear approach (based on previous literature⁴³) and lag-response functions using a natural cubic B-spline with two knots (33^rd and 66th percentiles) on the log scale of lags up to 3 days. We adjusted for lagged mean temperature and lagged RH using cubic b-splines with 3 degrees of freedom each to allow for non-linear exposure-response functions and lag-response functions with two knots (33rd and 66th percentiles) on the log scale of lags up to 3 days. We adjusted models for federal holidays as a potential confounder. We report results as risk ratios (RR) and 95% confidence intervals (CI) associated with a 10 parts per billion (ppb) increase in O₃. To evaluate whether current regulatory standards are protective for children, we repeated these analyses using O₃ levels below 70 ppb and tested for heterogeneity (total vs. <70 ppb) using the Wald test.

We performed a series of sensitivity analyses to assess the robustness of our findings to various analytic choices. To evaluate the influence of potential exposure misclassification, we restricted analyses to patients with a residential zip code with a 50-mile buffer around each hospital (the list of qualifying patients was provided by the PHIS data team). To evaluate the potential for residual confounding by meteorology, we refit the models: a) adjusting for maximum daily temperature or percentiles of daily temperature instead of mean daily temperature, b) excluding adjustment for temperature, c) excluding adjustment for RH and d) adjusting for temperature and RH only at lag 0 instead of across lags 0–3 days. Finally, we varied the number of knots in the lag-response functions for O₃, temperature and RH to assess the robustness of results to initial modelling choices.

We examined whether the association between O₃ and the risk of children’s ED visits varied by age, sex and insurance status as outlined above, assessing for heterogeneity using the Wald test. All analyses were performed using R version 4.1.1 and the ‘dlnm’ (version 2.4.7) and ‘survival’ (version 3.2.13) statistical packages.

Results

We identified ∼28 million ED visits among children <18 years of age across 42 hospitals between 1 January 2004 and 30 September 2015. Of these, 11.2 million ED visits occurred during the warm season, and 16.8 million during the cold season. Our analysis focused on impacts of warm season O₃; results presented here represent warm season while results for the cold season and entire year are included in the Supplementary Tables S1 and S2, and Supplementary Figures S1–S5 (available as Supplementary data at IJE online). Respiratory disorders accounted for approximately 15.2% of all ED visits, including 8.8% of all ED visits for respiratory infections and 3.6% for asthma (Table 1). Ear and allergy disorders accounted for 4.7 and 0.52% of all ED visits, respectively. Patients visiting the ED were predominantly male (54.0%), 0–5 years of age (58.0%) and carrying some form of government insurance (47.2%, Table 2).

Table 2.

Children’s emergency department (ED) visits by subgroup, May–September, 2004–2015

Subgroup	Number of ED visits	Total visits (%)
Sex
Female	5 135 923	45.97
Male	6 034 149	54.01
Age
0–5 years	6 477 832	57.98
6–12 years	3 043 883	27.25
13–17 years	1 650 064	14.77
Health insurance status
Private	2 489 243	22.28
Government	5 274 796	47.22
Other insurancea	556 224	4.98

^aIncludes no insurance, excludes missing.

Across the study area, daily maximum 8-hour O₃ concentrations ranged from 0.0 to 117.5 ppb, with an overall mean of 44.6 ppb (Supplementary Table S3, available as Supplementary data at IJE online). The highest mean O₃ concentrations were observed in the western and south-western USA, with localized high concentrations in major urban areas of the Midwest and eastern USA (Figure 1). Warm-season mean daily temperatures ranged from 15.6° to 103.4 F (Supplementary Table S3, available as Supplementary data at IJE online). Cold season and annual maps of the spatial distribution for O₃ indicate similar patterns (see Supplementary Figures S1 and S2, available as Supplementary data at IJE online).

Figure 1.

Distribution of warm-season ozone (O₃) concentrations. Map showing the spatial distribution of warm-season mean daily maximum 8-hour O₃ concentrations and the average number of daily ED visits at 42 US children’s hospitals, 1 May through 30 September 2004-2015. The highest mean ozone levels appear in the southwest and western US

We elected to use linear response functions when assessing the impact of O₃ on children’s health based on the relatively linear relationship of the exposure-response functions (besides allergy), especially at lower levels of exposure (Supplementary Figure S6, available as Supplementary data at IJE online). Daily 8-hour maximum O₃ was most strongly associated with ED visits for allergies (4.1% [2.5%, 5.7%]), followed by asthma (1.3% [0.8%, 1.9%] per 10 ppb increase), ear disorders (0.8% [0.3%, 1.3%]) and all-cause visits (0.3% [0.2%, 0.4%]) (Figure 2). We evaluated the lag response of the association between O₃ concentrations and ED visits for any cause and each cause separately (Figure 3). O₃ was most strongly associated with the risk of ED visits on the same day (lag 0) for all-cause visits, lag day 1 for asthma, lag days 1–3 for respiratory disorders and lag day 3 for allergies and ear disorders.

Figure 2.

Impact of ozone (O₃) exposure on emergency department (ED) visits at US children’s hospitals and effectiveness of current O₃ regulatory levels. Risk ratios and 95% confidence intervals of the association between a 10-ppb increase in warm-season 8-hour maximum daily O₃ exposure and emergency department visits for any cause or select specific causes among children seen in 42 US children’s hospitals, 1 May through 30 September 2004-2015. Values in black represent change in risk of all concentrations of exposure. Values in grey represent change in risk restricting to exposures <70 ppb. O₃ exposure is associated with increases in the risk of ED visits for all-cause, asthma, allergies and ear disorders. A similar relationship remains after restricting analyses to O₃ levels below the current National Ambient Air Quality Standards (NAAQS) standard

Figure 3.

Lagged exposures for warm season ozone (O₃) exposures. Lagged association between warm season daily 8-hour maximum O₃ exposure and all-cause and select cause-specific emergency department (ED) visits at US children’s at 42 US children’s hospitals, 1 May through 30 September 2004–2015

We repeated the analysis restricting to days where O₃ concentrations were below the NAAQS level and found little difference when comparing results using all exposure concentrations to results using only exposures below 70 ppb (Figure 2)_. When restricting to exposures below the NAAQS level, estimated risks for children’s health outcomes were not appreciably different (P-values = 0.915 to 0.999), pointing to significant increases in children’s ED visits at levels below the NAAQS standard.

We conducted sensitivity analyses to assess the robustness of our results (Supplementary Figures S7 and S8, available as Supplementary data at IJE online). Observed associations for ED visits for all causes, allergies, respiratory disorders, asthma and respiratory infections were sensitive to the inclusion or exclusion of RH and temperature. Adjusting the number of lag function knots, using a distance buffer and using maximum temperature had negligible effects on the results. To assess potential regional differences in the relationships between temperature and pollution and ED visits, we adjusted for temperature percentiles rather than absolute temperature, which also did not impact results.

We conducted stratified analyses across age, sex and health insurance subgroups (Supplementary Figures S9–S11, available as Supplementary data at IJE online). O₃ was associated with a higher risk of all-cause visits in children across strata of age, with increases in all-cause visits in 6- to 17-year-olds, allergy visits in 6- to 12-year-olds, asthma in 0- to 1-year-olds, ear disorders on 6- to 12-year-olds, and increases in respiratory disorders in 13- to 17-year-old children. In addition, O₃ was associated with a higher risk of all-cause ED visits for individuals with private insurance, asthma for individuals with government insurance, and allergies for individuals with government or other insurance. Differences were not materially different for all-cause or cause-specific ED visits by sex, except for a slightly higher relative risk of asthma observed in males.

Discussion

Despite extensive research on the effects of O₃ on health in adults, few large-scale studies exist in children. We leveraged a national dataset of ED visits at children’s hospitals in 27 US states to estimate the association between O₃ and rates of ED visits for all causes and specific causes. Our results indicate that higher levels of warm-season O₃ are associated with multiple children’s health outcomes and that these associations persist even at ambient ozone levels below the current regulatory standard.

Prior studies in children have primarily focused on asthma and demonstrate a link between O₃ and asthma events, agreeing with our findings.^{29,30,44–48} Sheffield et al.²⁹ conducted a study in New York City from 2005–2011 and found warm-season O₃ was associated with higher rates of asthma ED visits in 5- to 17-year-olds, ranging 2.9–8.4% in males and 5.4–6.5% in females, per 13 ppb increase in O₃. In Zheng et al., a positive association was still apparent after conducting pooled analyses of multiple studies, with asthma visits in children increasing by 0.9% (0.2%, 1.7%) per 10 ppb increase in O_3.⁴⁸

We observed a positive association between O₃ and ED visits for all causes. Few prior studies have considered all-cause ED visits and O₃ in children, with mixed results. In one such study, Fang et al.⁴⁹ found that O₃ was linked to a 10% (95% CI: 6%, 15%, per 88.6 ppb IQR) increase in all-cause ED visits in children 0–14 years. Another study failed to find an increase in all-cause ED visits among children with asthma (-0.1%, 95% CI: -1.3%, 12%, per 43.3 ppb).⁵⁰

Results from our analysis also suggest an association between O₃ and allergy ED visits. We found positive associations between O₃ and overall allergic disease, generally agreeing with previous findings regarding allergies such as conjunctivitis. Lu et al., 2019⁵¹ found that O₃ was related to a 0.7% (95% CI: 0.3%,1.0%, per 10 ppb) increase in outpatient visits. Chang et al. 2012⁵² also observed an association between O₃ and outpatient visits for conjunctivitis in Taiwan. Conversely, the association between O₃ and allergic rhinitis is less clear. In the USA, Parker et al., 2009⁵³ found a relationship between self-reported allergic rhinitis and O₃ in the USA (OR: 1.20, 95% CI: 1.15–1.26, per 10 ppb). However, in Taiwan, Hwang et al., 2006⁵⁴ observed null results for self-reported allergic rhinitis (OR: 1.05, 95% CI: 0.98–1.12, per 10 ppb). Some of this variation may be explained by the use of self-reported diagnoses over records and claims data.

Our results expand current knowledge regarding the impact of O₃ on ear disorders. Since the respiratory tract is anatomically connected to the Eustachian tubes of the inner ear, O₃ exposure could cause inflammation in both the lungs and ear.⁵⁵ Kousha et al.⁵⁶ investigated this link for O₃ in Ontario, Canada, and found a significant association between visits for otitis media and exposure at lag days 6 and 7 (16%, 95% CI: 2%, 31% and 20%, 95% CI: 5%, 34%, per 16.5 ppb, respectively). Zemek et al., 2010⁵⁷ looked at the effect of O₃ on ED visits for otitis media in Edmonton, Canada, and authors observed an association for single day lags during cold season O₃ (7.0%, 95% CI: 1.0%, 14%, increase per 14.0 ppb).

While our asthma results are comparable to other studies, we did not observe a strong link between O₃ and combined respiratory disorders or infections. Results for respiratory disorders were null (risk ratio (RR): 1.00, 95% CI: 1.00, 1.01), and we observed an inverse association between O₃ and respiratory infections. Previous research by Strosnider et al.⁵⁸ observed RRs of 1.017 (95% CI: 1.011, 1.023) for respiratory disorders in children 0–18 years per 20 ppb increase in O₃. Additionally, Darrow et al.²⁸ analysed acute respiratory visits in metro Atlanta hospitals and showed that visits for respiratory infections and pneumonia were related to O₃, increasing ED visits by 4.1% (95% CI: 1.9%, 6.4%) and 8.3% (95% CI: 3.8%, 13.1%) per 27.8 ppb, respectively. The results in our analysis point to the impact of cumulative lag 0–3 as the exposure metric in our main model. When considering lags on individual days, O₃ is minimally associated with respiratory infections at certain time slices on lag days 1 and 2 for the warm season, with a null association for lag day 3. Strong positive associations for lag days 1 and 2 are shown in our annual and cold season analyses (see Supplementary Figure S6, available as Supplementary data at IJE online). The negative results from lag day 0 and null results at lag day 3 suggest that there may be a delay in the onset of symptoms and that the effects of exposure are no longer important by lag day 3. Further assessment of individual lag days may provide clinically relevant information regarding the need for care following high O₃ exposure days.

Our findings also suggest that risk of ED visits may differ among subgroups of interest. For example, the risk of an all-cause ED visit appears to be lower in young children (0–5 years) versus older children and adolescents, potentially pointing to increased time spent outdoors and/or a higher risk of outcomes such as injuries in older children. We also found that privately insured individuals have a higher risk of all-cause ED visits versus those with government or other insurance, which could point to differential hospital use (general vs. children’s) among differently insured populations.

Concerning current policy, our results indicate that levels of O₃ below the existing NAAQS level of 70 ppb are similar to results found across the full range of O₃ Levels. If confirmed in additional studies, these results suggest that the current standard may need to be adjusted further to protect vulnerable populations, including children. More research is needed in determining appropriate regulatory threshold exposures for susceptible populations.

Data in this study were obtained from children’s hospitals, regardless of health insurance status. While this should aid in generalizability, we cannot rule out potential differences caused by preferences in health-seeking behaviour (e.g. choosing to go to the ED rather than urgent care or being seen in outpatient clinics). Additional uncertainty may also be introduced through exposure misclassification. To protect patient privacy, data were aggregated at the county level, leading to some loss of spatial granularity which likely contributed to some exposure misclassification. We also lack data on time-activity patterns of individuals, including amount of time spent at specific locations or the outdoor and indoor air quality at each location. While it would be ideal to have information on patient residential address, these data were not provided by PHIS. To evaluate the potential for exposure misclassification to have impacted our results, we included a sensitivity analysis restricted to patients residing within a 50-mile buffer around each hospital and performed selected comparisons of exposures for hospital counties versus adjacent counties.

Notwithstanding these limitations, this study represents one of the most extensive O₃ studies in US children. Our findings provide evidence that ambient O₃ is associated with a higher risk of ED visits in children and adolescents, even at levels below those deemed safe under current regulations. Additional research is needed to assess the effectiveness of current guidelines and determine whether interventions aimed at lowering ambient ozone levels would lead to improved health outcomes in children and other vulnerable groups.

Conclusions

In summary, we have shown that ambient O₃ is associated with adverse health impacts in children as evidenced by increased healthcare utilization. Specifically, we found that O₃ is associated with a higher risk of ED visits for all causes, allergies, asthma and ear disorders. These results were consistent, even is analyses restricted to O3 levels <70 ppb, the current regulatory standard in the USA. If confirmed in subsequent studies, these results suggest that current standards may not adequately protect children. In addition, our results suggest that older children and adolescents may be at higher risk of O3-related all-cause and respiratory ED visits. Impending changes in climate will likely increase exposures to O₃, thus potentially increasing the risk of adverse health outcomes in children.

Ethics approval

The Boston Children’s Hospital institutional review board determined that this study is not considered human subjects research and granted a waiver of informed consent on that basis.

Data availability

Under our data use agreement, the healthcare claims data from PHIS cannot be re-shared. Data on temperature metrics may be accessed via PRISM at https://prism.oregonstate.edu. The ozone exposure data used in this analysis can be requested from the authors of Réquia WJ, et al., ‘An Ensemble Learning Approach for Estimating High Spatiotemporal Resolution of Ground-Level Ozone in the Contiguous United States’.

Supplementary data

Supplementary data are available at IJE online.

Author contributions

JDS contributed to the design of the study, conducted the analysis and is the primary author of the text. YS contributed to the design of the study, data acquisition and editing of the text. ELG aided in data acquisition and editing of the text. KRS contributed to data acquisition, data curation and editing of the text. JS provided data and contributed to the study design, and editing of the text. AB contributed to study design, data acquisition, data curation and editing of the text. GAW contributed to study design and editing of the text. AN-S contributed to study design, data acquisition and editing of the text.

Funding

This work was supported by grant 5R01-ES032418-02 from the National Institutes of Health and grant 216033-Z-19-Z from the Wellcome Trust. The sponsors had no role in the design and conduct of the study: collection, management, analysis and interpretation of the data; preparation, review or approval of the manuscript; and the decision to submit the manuscript for publication.

Conflict of interest

None declared.