Objectives This study explored the effects of short-term exposure to air pollution on hospital admissions for autism spectrum disorder (ASD), a proxy for symptom aggravation, among Korean children aged 5–14 years.
Design Time-series study.
Setting, participants and outcome measures We used data from the National Health Insurance Service (2011–2015). Daily concentrations of fine particulate matter (PM2.5), nitrogen dioxide (NO2) and ozone (O3) levels in each region were used as exposures. ASD cases were defined based on a principal admission diagnosis of the claims data. We applied distributed lag non-linear models and a generalised difference-in-differences method to the quasi-Poisson models to estimate the causal effects of air pollution for up to 6 days. We also performed weighted quantile sum regression analyses to assess the combined effects of air pollution mixtures.
Results PM2.5 levels at lag day 1, NO2 levels at lag day 5 and O3 levels at lag day 4 increased the risks of hospital admissions for ASD (relative risk (RR)=1.17, 95% CI 1.10 to 1.25 for PM2.5; RR=1.09, 95% CI 1.01 to 1.18 for NO2 and RR=1.03, 95% CI 1.00 to 1.06 for O3). The mean daily count of hospital admissions for ASD was 8.5, and it would be 7.3, 7.8 and 8.3 when the PM2.5 levels would be decreased by 10.0 µg/m3, NO2 by 10 ppb and O3 by 10 ppb, respectively. The weighted quantile sum index, constructed from PM2.5, NO2 and O3 levels, was associated with a higher risk of hospital admissions for ASD (RR 1.29, 95% CI 1.14 to 1.46), where NO2 was found to contribute to the effects most (the weight of 0.80).
Conclusions These results emphasise that reduction of air pollution exposure should be considered for ASD symptom management, with important implications for the quality of life and economic costs.
- child & adolescent psychiatry
- public health
- mental health
Data availability statement
Data may be obtained from a third party and are not publicly available. The datasets used in this study are not publicly available, but these can be provided on reasonable request after the approval of the National Health Insurance Service of the Republic of Korea.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Strengths and limitations of this study
This is the first study to directly investigate the effects of short-term exposure to air pollution on autism spectrum disorder (ASD)-related outcomes.
All cases of hospital admissions for ASD in Korean children aged 5–14 years were considered.
We estimated causal effects rather than observational associations by applying a causal inference method (ie, a difference-in-differences method).
We used regional air pollution levels not individual levels as exposures, leading to measurement error.
Due to the remaining social stigma, patients with ASD with mild symptoms might be less likely to receive psychiatric treatments and not be included in the analyses.
Autism spectrum disorder (ASD) is a heterogeneous neurodevelopmental disease with a wide range of symptoms and severity, characterised by deficits in social communication and interaction, restricted interests, and repetitive behaviours.1 Due to its high prevalence (1.5% in the USA and 2.2% in the Republic of Korea)2 3 and the high lifelong cost to support an individual with ASD (US$2.2 million in the USA and US$2.4 million in the UK),4 ASD is considered a major public health problem.
Neuroinflammation and systemic inflammation are often accompanied by ASD.5 Recent studies have also reported that ongoing inflammatory responses, represented by serum inflammatory cytokine levels, are associated with the severity of ASD symptoms,6 and the core symptoms of ASD can be improved by modulating the inflammatory status (eg, with drugs, supplements and dietary formulations), especially among patients with ASD with high serum proinflammatory cytokine levels.7 8 Collectively, these results suggest an association between the immune system and ASD symptoms and the presence of an immune subtype of ASD, which can potentially benefit from immune modulatory treatment and prevention strategies.9
Given the potential side effects from chronic application of immune modulatory agents, identification of risk factors for ASD symptom aggravation (which may lead to hospital admissions due to reasons, such as hyperactivity, aggression and self-injurious problems) is important with regard to the quality of life of patients and their family and economic costs due to treatment and care. Short-term exposure to air pollution (ie, days to weeks) can induce systemic inflammation and neuroinflammation (possibly due to the penetration of particulate air pollutants through the lungs and olfactory epithelium, changes in blood–brain barrier and activation of microglia) and may aggravate ASD symptoms.10–12 However, previous air pollution studies have only focused on the association between long-term exposure to air pollution (ie, months to years) during pregnancy (and early postnatal period in a few studies) and ASD development among children and have provided incomplete evidence for the association.13 To the best of our knowledge, no study has directly investigated the association between short-term exposure to air pollution and ASD symptom aggravation.
Therefore, we investigated the effects of short-term exposure to air pollution on hospital admissions for ASD as a proxy for ASD symptom aggravation.14 Because the developing nervous system is more susceptible to environmental exposures than the adult nervous system15 and because ASD cannot be reliably diagnosed until 3 years of age,16 we performed analyses using data on all hospitalisations for ASD in Korean children aged 5–14 years.
Study design and hospital admission data
We conducted a time-series study using data on daily counts of hospital admissions for ASD among children aged 5–14 years between 1 January 2011 and 31 December 2015. The data, which were provided by the National Health Insurance Service, the sole health insurance provider of the universal coverage system, were aggregated according to the 16 regions of the Republic of Korea (online supplemental table 1) and sex. The National Health Insurance Service had medical information on all the residents of the Republic of Korea, including the data of those covered by either the National Health Insurance (97%) or the Medical Aid programme (3%).17 Because the National Health Insurance Service data used in this study did not disaggregate according to emergency visit or follow-up visit, we considered hospital admissions regardless of the route of admission (eg, emergency visit, follow-up visit) and could not perform analyses excluding follow-up visits. We constructed a sex-combined time-series daily count dataset from sex-aggregated raw data and used the sex-combined dataset for further analyses, except for sex-stratified analyses.
Air pollution and meteorological factors
Previous studies have reported associations between long-term exposure to particulate matter with an aerodynamic diameter ≤2.5 µm (PM2.5), nitrogen dioxide (NO2), and ozone (O3) and ASD development.18 19 Thus, we selected PM2.5, NO2 (daily mean concentrations) and O3 (daily 8-hour maximum concentrations) as exposures of interest in this study.
Due to the lack of national monitoring data during the study period (2011–2015), PM2.5 levels were estimated using the Integrated Multi-Scale Air Quality System for Korea, as described in detail elsewhere.20 Briefly, we combined meteorological and chemical data and simulated hourly PM2.5 for each 3×3 km grid cell using the Community Multi-Scale Air Quality model (V.4.7.1). Region-specific daily mean PM2.5 levels were then estimated by averaging daily gridded PM2.5, calculated from predicted 24 hours PM2.5.
Meanwhile, data on NO2 and O3 levels in each region were obtained from 318 fixed-site monitoring stations of the National Ambient Air Monitoring Information System, which collects 24 hours air pollution monitoring data with stringent quality control.21 After excluding missing values of raw data from monitoring stations (<5%), we estimated region-specific daily mean NO2 levels and daily 8-hour maximum O3 levels by averaging daily mean NO2 levels and daily 8-hour maximum O3 levels of all monitoring stations in each region.
Additionally, data on temperature (°C) and relative humidity (%) in each region were obtained from the Korea National Meteorological Administration. The region-specific daily mean temperature and relative humidity were calculated by averaging the respective temperature and relative humidity levels from all weather stations in each region.
Autism spectrum disorder
ASD was defined as a principal admission diagnosis based on the International Statistical Classification of Diseases and Related Health Problems, 10th revision codes F84.0, 84.1, 84.5, 84.8 and F84.9. The accuracy of the diagnosis in the National Health Insurance Service data was assessed to be high, especially in inpatient settings.22
To evaluate the associations between short-term exposure to air pollution and hospital admissions for ASD considering delayed effects for up to 6 days, we constructed quasi-Poisson generalised linear regression models, which applied distributed lag non-linear models23 for air pollution exposures with lag structures (polynomial functions) and concentration-response curves (natural cubic splines) with 3 df, respectively. Although the df was determined according to a previous study,24 the consistency of the results was also assessed using different df as a sensitivity analysis.
To estimate the causal effects of short-term exposure to air pollution on hospital admissions for ASD that strictly controlled for potential confounders, including seasonality,19 we applied a generalised difference-in-differences method.25 By adjusting for spatial units (regions) and temporal units (days) as indicator variables, measured and unmeasured confounding factors related to these spatial and temporal units can be effectively controlled. This method can provide estimators for causal effects in the potential outcome framework, although it is time-intensive and resource-intensive and computationally exhaustive.25 Additionally, we adjusted the analyses for temperature and relative humidity, both of which were modelled with lag structures up to 6 days (3 df) and concentration-response curves (3 df), similar to air pollution. Finally, terms for the population of regions at each year were included as offsets. Therefore, main analytical models can be described as follows: log[E(Y s,t )] = β0 + β1Air Pollut s,t + β2I s + β3I t + β4Temp s,t + β5Humi s,t +log(Pop s,t ).
In this equation, Y s,t is the number of hospital admissions for ASD in region s at day t. Air Polluts,t, Temps,t and Humis,t are the cross-basis matrices of air pollutant levels, temperature and relative humidity in region ‘s’ at day ‘t’ with lag structures up to 6 days (3 df) and concentration-response curve (3 df), respectively. I s and I t are indicator variables for regions and days, respectively. Pop s,t is the population of region ‘s’ in the year of day ‘t’.
Previous animal and epidemiological studies have suggested stronger effects of air pollution exposure on ASD-related outcomes among men than among women.15 Therefore, to investigate the potential heterogeneity of the associations by sex, we evaluated the interactions between air pollution exposure and sex. We tested the interaction terms between a natural cubic spline of each air pollution exposure (3 df) and sex, added to the above-mentioned main models with the main effect term of sex by performing F-tests comparing models with and without the interaction terms.26 Analytical models for interaction analyses can be described as follows: log[E(Y s,t )] = β0 + β1Air Pollut s,t + β2I s + β3I t + β4Temp s,t + β5Humi s,t + β6Sex + β7Ns Air Pollut s,t : Sex +log(Pop s,t ).
In this equation, Sex is the indicator variable coded as 1 if the data are for boys and 0 if the data are for girls. Ns Air Pollut s,t is a natural cubic spline of air pollution exposure in region ‘s’ at day ‘t’ (3 df). We also performed stratified analyses according to sex and region characteristics (seven metropolitan cities (Seoul, Busan, Daegu, Incheon, Gwangju, Daejeon and Ulsan) versus nine non-metropolitan regions (Gyeonggi-do, Gangwon-do, Chungcheongbuk-do, Chungcheongnam-do, Jeollabuk-do, Jeollanam-do, Gyeongsangbuk-do, Gyeongsangnam-do and Jeju-do) (online supplemental table 1). We conducted the regional stratified analyses assuming that hospital visits due to ASD symptom aggravation might differ by region characteristics.
Because there may be confounding by other air pollutants due to substantial correlations among them, we constructed a multiple pollutant model that incorporated all three air pollution exposures (PM2.5, NO2 and O3) into the main model. Furthermore, to estimate the combined effect considering potential interactions among air pollutants and to identify the relative importance of individual exposure with respect to the effects on the outcome,18 19 we performed a weighted quantile sum regression analysis.27 In this analysis, we multiplied the quartiles of air pollution exposures identified to be associated with hospital admissions for ASD in the main model (ie, PM2.5 at lag day 1, NO2 at lag day 5 and O3 at lag day 4) by the magnitude of each effect of individual exposure (constrained between 0 and 1 and summed to 1) as weights, resulting in the weighted quantile sum index, which represents pollutant mixtures considered of relative importance. We then constructed an analytical model facilitating the weighted quantile sum index as an exposure and hospital admissions for ASD (modelled as a quasi-Poisson distribution) as an outcome. The model was adjusted for region, temperature (a natural cubic spline of 3 df) and relative humidity (a natural cubic spline of 3 df). The weighted quantile sum regression weights were estimated from 40% of 1000 bootstrap sample data, and the association between the weighted quantile sum index and hospital admissions for ASD was evaluated using the remaining 60% of the data. We did not constrain the parameter estimates as positive or negative.
We presented the results of all analyses, except for the weighted quantile sum regression analyses, by 10.0 µg/m3 increase for PM2.5 (reference of 10.0 µg/m3) and 10.0 ppb increase for NO2 (10.0 ppb) and O3 (30.0 ppb). In the weighted quantile sum regression analysis, we presented the result of a one-unit increase in the weighted quantile sum index, which is approximately interpreted as a one-quartile increase in pollutant mixtures. We also approximately calculated a daily mean count of outcomes in a hypothetical case when the PM2.5 levels would be decreased by 10.0 µg/m3, or NO2 or O3 levels would be decreased by 10.0 ppb, respectively, using the following equation: the daily mean counts of outcomes × 1/relative risk (RR).
All analyses were performed using R (V.4.0.5; R Foundation for Statistical Computing, Vienna, Austria).
Patient and public involvement
This time-series study using secondary claims data was designed and conducted without patient and public involvement. Our results will be disseminated to the public through publication in this journal.
The mean (SD) of the daily counts of hospital admissions for ASD during the study period (2011–2015) was 8.5 (8.2) for the total study population. The daily counts of hospital admissions for ASD were substantially higher among boys (7.0) than among girls (1.6). The means (SD) of daily mean PM2.5 and NO2 levels and daily 8 hours maximum O3 levels during the study period (2011–2015) were 19.3 (14.7) µg/m3, 20.7 (10.7) ppb and 37.2 (16.4) ppb, respectively, among which the daily mean PM2.5 and NO2 levels were higher than the levels of the USA but lower than the levels of China, and the daily 8 hours maximum O3 levels were lower than the levels of both the USA and China.28–30 The mean temperature was 13.2℃ (21.3℃ during warm seasons and 5.0℃ during cool seasons) (table 1).
In the Pearson’s correlation analysis for air pollution levels and meteorological factors, we found a moderate positive correlation between PM2.5 and NO2 levels (r=0.52, p<0.01), weak positive correlations between O3 levels and temperature (r=0.43, p<0.01) and between temperature and humidity (r=0.43, p<0.01), and a weak negative correlation between NO2 levels and temperature (r=−0.30, p<0.01). For other pairs, absolute values for correlation coefficients were smaller than 0.30, although correlations of all assessed pairs were statistically significant (p<0.05) (online supplemental figure 1).
PM2.5 levels at lag day 1, NO2 levels at lag day 5 and O3 levels at lag day 4 were associated with a higher risk of hospital admissions for ASD (RR=1.17, 95% CI 1.10 to 1.25 for PM2.5; RR=1.09, 95% CI 1.01 to 1.18 for NO2 and RR=1.03, 95% CI 1.00 to 1.06 for O3), whereas PM2.5 and NO2 levels at lag day 0 were associated with a lower risk (table 2). The observed mean daily count of hospital admissions for ASD was 8.5 in this study, and it would be 7.3, 7.8 and 8.3 when the PM2.5 levels would be decreased by 10.0 µg/m3, NO2 by 10 ppb and O3 by 10 ppb, respectively.
The associations between air pollution exposures and hospital admissions for ASD were different between boys and girls, especially for PM2.5 and NO2 (p-interactions: 0.03 for PM2.5, <0.01 for NO2 and 0.14 for O3). When the study population was stratified by sex, the associations between air pollution exposures and hospital admissions for ASD were more prominent among boys (RR 1.19, 95% CI 1.11 to 1.27 for PM2.5 levels at lag day 1; RR 1.07, 95% CI 1.00 to 1.15 for NO2 at lag day 4 and RR 1.10, 95% CI 1.02 to 1.20 for NO2 at lag day 5; and RR 1.04, 95% CI 1.00 to 1.07 for O3 at lag day 4 and RR 1.04, 95% CI 1.00 to 1.07 for O3 at lag day 5) than among girls (table 3). Among boys, the mean daily count of hospital admissions for ASD was 7.0 in this study, and it would be 5.9, 6.5 and 6.4 when the PM2.5 levels would be decreased by 10.0 µg/m3, NO2 by 10 ppb and O3 by 10 ppb, respectively. Among girls, the mean daily count of hospital admissions for ASD was 1.6, and it would be 1.5 when the PM2.5 levels would be decrease by 10.0 µg/m3 and O3 by 10 ppb.
In the analyses stratified by region characteristics, the associations between air pollution exposures and hospital admissions for ASD were generally more prominent in the nine non-metropolitan regions than those in the seven metropolitan cities, although the CIs overlapped (online supplemental table 2).
We found consistent associations between PM2.5, NO2 and O3 levels and hospital admissions for ASD in the multiple-pollutant model (online supplemental table 3) compared with the main analytical models. These results suggest that the concern for confounding by other air pollutants would be low. In the weighted quantile sum regression analysis, the weighted quantile sum index, constructed from PM2.5 at lag day 1, NO2 at lag day 5 and O3 at lag day 4, was associated with a higher risk of hospital admissions for ASD (RR 1.29, 95% CI 1.14 to 1.46). The weights for PM2.5 at lag day 1, NO2 at lag day 5 and O3 at lag day 4 were approximately 0.20, 0.80 and <0.01, respectively (online supplemental figure 2).
In the sensitivity analyses modelling air pollution exposures using df of 4–7 (instead of 3 df as in the main analytical models) for lag structures and concentration-response curves, the results did not change appreciably (data not shown).
Short-term exposure to PM2.5, NO2 and O3 was associated with a higher risk of hospital admissions for ASD. The associations were demonstrated to be more prominent among boys than among girls in sex-stratified analyses. In the weighted quantile sum regression analysis, the weighted quantile sum index was associated with a higher risk of hospital admissions for ASD with weights of 0.20, 0.80 and <0.01 for PM2.5, NO2 and O3, respectively.
A recent systematic review and meta-analysis concluded that there is evidence of an association between long-term exposure to air pollution (especially PM2.5 and NO2) during pregnancy and ASD development among children.31 A limited number of studies have been conducted on postnatal exposures, which also suggests an association between long-term air pollution exposure during early life after birth and ASD development. A case–control study in China reported the association between air pollution (PM2.5, particulate matter with an aerodynamic diameter ≤10.0 µm (PM10) and particulate matter with an aerodynamic diameter ≤1.0 µm) levels in the second and third year after birth and ASD.32 Another case–control study in Denmark reported an association between air pollution (PM2.5, PM10, NO2 and sulfur dioxide) levels for 9 months after birth and ASD.33
Short-term exposure to air pollution has been shown to be associated with a higher risk of hospital admission or emergency department visit for psychiatric disorders, such as mental disorder, depression, schizophrenia, suicide attempt, substance abuse disorder and panic attack.34 However, to the best of our knowledge, no direct evidence, except this study, exists on the association between short-term exposure to air pollution and ASD symptom aggravation.
Air pollution is known to impact the central nervous system by activating microglia and disrupting the blood–brain barrier through systemic inflammation, neuroinflammation, oxidative stress, cerebrovascular injury and neurodegenerative processes.11 Short-term exposure to air pollution is known to activate microglia, the resident immune cells of the central nervous system involved in the production and secretion of proinflammatory cytokines, such as interleukin (IL)-6, IL-1β and tumour necrosis factor (TNF)-α.12 Proinflammatory cytokine levels have been associated with the severity of communication impairment and aberrant behaviours among children with ASD,6 emphasising the importance of ongoing inflammatory responses with regard to ASD symptoms. Immune modulatory treatment targeting microglia has, therefore, been suggested and assessed to be effective for symptomatic therapy for ASD (at least the immune subtype of ASD). By administering dietary formulation of luteolin, a natural flavonoid, to children with ASD, IL-6 and TNF levels were lowered, and ASD symptoms substantially improved in the communication, daily living skills and social domains among children with higher baseline IL-6 and TNF levels.5 7 Risperidone and aripiprazole, second-generation antipsychotics approved for control of irritability symptoms among patients with ASD by the Food and Drug Administration, have also been suggested to affect ASD symptoms through anti-inflammatory properties, although their specific mechanisms remain unclear.5 9
This study unexpectedly found inverse associations between PM2.5 and NO2 levels at lag day 0 and hospital admissions for ASD. Because these inverse associations are clinically irrelevant (given that we considered hospital admissions as the outcome) and biologically implausible (considering the causal pathway via microglia activation and inflammatory status mentioned above), we assumed that the inverse associations between PM2.5 and NO2 levels at lag day 0 and hospital admissions for ASD may be explained by the harvesting effect (outcome displacement) induced by air pollution exposures at previous lag days (eg, PM2.5 at lag day 1 and NO2 levels at lag day 5).35 The findings of this study are further supported by the clear paralleled and lagged distribution patterns of the daily PM2.5 levels and counts of hospital admissions for ASD (online supplemental figure 3). Air pollution levels at previous lag days not only correlated with the air pollution levels at lag day 0 but also increased the risk of hospital admissions for ASD, leading to observed non-causal inverse associations.
Previous animal and epidemiological studies have shown male-specific and/or male-biased associations between air pollution exposure and ASD-related outcomes, consistent with the findings of this study. For example, male mice with perinatal exposures to PM2.5 have reduced anogenital and body sniffing behaviours (indicators of reciprocal social interaction).36 In epidemiological studies, PM2.5 levels during pregnancy and the first year after birth have been associated with ASD development only among boys.37 This sex difference may be explained by a larger number of microglia,38 activation of microglia due to testosterone39 and lower antioxidant (eg, glutathione and sulphate) levels among boys compared with that among girls.40
The association between O3 levels and hospital admissions for ASD remained after adjustment for PM2.5 and NO2 levels. However, the weight of O3 was estimated to be low (<0.01) in the weighted quantile sum regression analysis. It is notable that there is significantly less evidence for O3 than for PM2.5 and NO2 in previous epidemiological studies exploring the effects of long-term exposure to air pollution on ASD development (OR 1.00, 95% CI 1.00 to 1.01 for O3; OR 1.06, 95% CI 1.01 to 1.11 for PM2.5 and OR 1.02, 95% CI 1.01 to 1.04 for NO2 in random effects meta-analysis),31 although several animal studies have reported the effects of O3 exposure on neurobehavioral outcomes, such as social recognition memory.41 Collectively, the results of this study may be interpreted as the effect of O3 being independent, whereas the contribution of O3 might be relatively small with regard to ASD symptom aggravation in air pollution mixtures. Because this study is the first to investigate the effects of short-term exposure to O3 on ASD-related outcomes and it is not appropriate to determine the relative importance of exposures solely dependent on statistical models, further studies are warranted to confirm the results of this study regarding the contribution of O3.
This study has some limitations. First, some results may be spurious, possibly occurring by chance due to extensive analyses. However, we did not adjust for multiple comparisons, because each analysis was not independent and all of the results were presented without selection. Second, although the results of this study may be driven by subgroups of the study population (eg, patients with immune subtype of ASD, who have high proinflammatory cytokine levels and are also a target for immune modulatory treatment; those with morbidities, such as gestational diabetes mellitus and infections; and those with lower socioeconomic status),42 we could not evaluate this possibility due to a lack of essential information. Third, because this study was conducted in the Republic of Korea, a practically single-ethnic nation, caution should be exercised in generalising the results to other populations, given the potential heterogeneity of the results by race/ethnicity suggested by previous studies.43 Fourth, there is a concern of exposure misclassification attributable to large spatial units, although several time-series studies conducted in Korea used regional air pollution levels as exposures (instead of air pollution levels measured for a finer spatial unit).44–46 This type of error (ie, Berkson error) is likely to lead to imprecision in the estimation of associations rather than bias. Fifth, we only considered three criteria air pollutants (ie, PM2.5, NO2 and O3) as exposures according to previous studies.18 19 The next step in future studies is to analyse the effects of other hazardous air pollutants, such as lead, mercury and arsenic, on ASD symptom aggravation. Sixth, although hospital admissions are commonly used as a proxy for symptom aggravation in epidemiological studies,14 hospital admissions for ASD might disproportionately reflect aggravation of symptoms related to hyperactivity, aggression and self-injurious behaviour more than those related to deficits in social communication. In addition, private practice or community-based programmes are also potential options for providing care for ASD children with aggravated symptoms in Korea. This could have led to the underestimation of the effect of air pollution on ASD symptom aggravation observed in this study. We assume that the more prominent associations in the non-metropolitan regions than in metropolitan cities might also be explained at least in part by the fact that most resources for these treatment options other than those in hospitals were in metropolitan cities. Furthermore, predetermined scheduled hospital admissions can also be another source of outcome misclassification (in terms of ASD symptom aggravation) leading to imprecision in the estimation of associations. Therefore, further studies that directly evaluate ASD symptoms are necessary to address this issue. Seventh, due to the remaining social stigma for psychiatric treatments in the Republic of Korea,47 patients with ASD with mild symptoms (and not accompanying intellectual disability) might be less likely to receive psychiatric treatments than patients with ASD with more severe symptoms. The possibility that the outcome of this study might reflect more severe ASD cases, rather than all ASD cases, needs to be considered to correctly interpret the results.
However, this study also has several strengths. First, this is the first study to directly explore the association between short-term exposure to air pollution and ASD symptom aggravation, which has relevant implications for immune modulatory prevention (regarding air pollution exposures) as a possible ASD management strategy. Second, because all cases of hospital admissions for ASD in Korean children aged 5–14 years were considered, we could perform various analyses, such as sex-stratified analyses, with sufficient power. Third, we estimated causal effects rather than observational associations. To do this, we applied a generalised difference-in-differences method, a causal inference method. Fourth, in contrast to most previous studies that considered only one exposure at a time, we analysed multiple air pollution exposures together and explored the combined effects using the weighted quantile sum regression model.
This study suggests that short-term exposure to air pollution affects ASD symptom aggravation, which is more prominent among boys than among girls. Air pollution mixtures were also found to be associated with ASD symptom aggravation, mostly driven by PM2.5 and NO2. These results emphasise that reduction of air pollution exposure needs to be considered for successful ASD symptom management, which is important with regard to quality of life and economic costs. Because this is the first study on this subject, further studies, especially studies directly investigating ASD symptoms in more detail, are warranted to confirm the results and draw policy implications.
Data availability statement
Data may be obtained from a third party and are not publicly available. The datasets used in this study are not publicly available, but these can be provided on reasonable request after the approval of the National Health Insurance Service of the Republic of Korea.
Patient consent for publication
The study protocol was reviewed and approved by the Institutional Review Board of Seoul National University Hospital (E-1911-013-1076).
We are grateful to the National Health Insurance Service of the Republic of Korea for the provision of data through the Big Data Utilisation Specialist Programme (2020).
Contributors Conceptualisation: K-NK and Y-CH; Methodology: K-NK and Y-CH; Software: K-NK; Formal analysis: K-NK; Writing–original draft: K-NK and Y-CH; Writing–review and editing: JHS, SJC, HYS and SK; Supervision: Y-C.H; Project administration: Y-CH; Funding acquisition: K-NK and Y-CH; Guarantor for the overall content: Y-CH.
Funding This work was supported by the National Strategic Project-Fine Particle of the NRF funded by the Ministry of Science and ICT, Ministry of Environment, and Ministry of Health and Welfare of the Republic of Korea (Nos. NRF-2017M3D8A1092008 and NRF-2017M3D8A1092009).
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.