Ambient Air Pollution and Kawasaki Disease in Korean Children: A Study of the National Health Insurance Claim Data.

Background Kawasaki disease (KD) is a systemic vasculitis of unknown etiology that primarily affects children under 5 years of age. Some researchers suggested a potential triggering effect of air pollution on KD, but the findings are inconsistent and limited by small sample size. We investigated the association between ambient air pollution and KD among the population of South Korea younger than 5 years using the National Health Insurance claim data between 2007 and 2019. Methods and Results We obtained the data regarding particulate matter ≤10 or 2.5 µm in diameter, nitrogen dioxide, sulfur dioxide, carbon monoxide, and ozone from 235 regulatory monitoring stations. Using a time-stratified case-crossover design, we performed conditional logistic regression to estimate odds ratios (OR) of KD according to interquartile range increases in each air pollutant concentration on the day of fever onset after adjusting for temperature and relative humidity. We identified 51 486 children treated for KD during the study period. An interquartile range increase (14.67 μg/m3) of particulate matter ≤2.5 µm was positively associated with KD at lag 1 (OR, 1.016; 95% CI, 1.004-1.029). An interquartile range increase (2.79 ppb) of sulfur dioxide concentration was associated with KD at all lag days (OR, 1.018; 95% CI, 1.002-1.034 at lag 0; OR, 1.022; 95% CI, 1.005-1.038 at lag 1; OR, 1.017; 95% CI, 1.001-1.033 at lag 2). Results were qualitatively similar in the second scenario of different fever onset, 2-pollutant model and sensitivity analyses. Conclusions In a KD-focused national cohort of children, exposure to particulate matter ≤2.5 µm and sulfur dioxide was positively associated with the risk of KD. This finding supports the triggering role of ambient air pollution in the development of KD.

carbon monoxide

Kawasaki disease

nitrogen dioxide

ozone

particulate matter ≤10 or 2.5 µm in diameter, respectively

sulfur dioxide

Clinical Perspective

What Is New?

This study investigated the association between ambient air pollution and development of Kawasaki disease using data from a Kawasaki disease‐focused national cohort of children.

In children aged under 5 years, short‐term exposure to particulate matter ≤2.5 µm and sulfur dioxide was positively associated with the risk of Kawasaki disease.

What Are the Clinical Implications?

Ambient air pollution may have a triggering role in the development of Kawasaki disease among children under 5 years of age.

Kawasaki disease (KD) is a systemic vasculitis of unknown etiology that primarily affects children under 5 years of age. , The common clinical signs are fever, skin rash, red eyes, red lips, enlarged anterior cervical lymph nodes, and joint pain. , KD can result in coronary artery aneurysms and is the leading cause of acquired heart disease among pediatric patients in high‐income countries. , For children with acute KD, timely initiation of intravenous immunoglobulin (IVIG) treatment reduces the risk of coronary complications. ,

Incidence of KD varies considerably between different racial and ethnic groups, with the highest rates among Asian/Pacific Islander children and the lowest among White children in the United States. , , In Asian countries, the incidence of KD is as high as 191 in South Korea, 184.6 in Japan, 68.8 to 107.3 in China, 69 in Taiwan, and 2.1 to 3.4 in Thailand per 100 000 children under 5 years of age. Prior epidemiological studies observed that the development of KD is associated with several respiratory viral infections, early childhood environment, and winter season. , , Although the cause of KD remains unknown, clinical findings suggest that a respiratory agent may trigger KD in susceptible children causing inflammation in the upper airway. , , Some research reported a potential triggering effect of wind‐borne desert dust and extreme temperature on KD, , but the findings are inconsistent and limited by small sample size. Therefore, we investigated the association between ambient air pollution and KD using the data of all identified KD cases from the population aged under 5 years in South Korea.

METHODS

Because of the sensitive nature of the data constructed for this study, a subset of the data generated for this study are available from the corresponding author upon reasonable request.

Data Source

The National Health Insurance System is a public, single‐payer system covering over 99% of all citizens and residents in South Korea. Following medical services, the provider sends reimbursement claims for medical expenses incurred to the Health Insurance Review & Assessment Service, which reviews the claims, assesses the quality of care provided, and evaluates adequacy for healthcare services. Based on the results of the review by the Health Insurance Review & Assessment Service, the National Health Insurance System reimburses the service providers for the cost of supplying the medical services to patients. Using the Health Insurance Review & Assessment Service data, we constructed a KD‐focused national cohort of children aged 0 to 4 years who had the International Classification of Diseases, Tenth Revision (ICD‐10) code for Kawasaki disease (M30.3) and received IVIG between November 2007 and October 2019. We identified all children with a first episode of KD (n=53 837) and further restricted the population of children to those who lived in the districts with available ambient air pollution monitors (n=51 486). Because the Health Insurance Review & Assessment Service data were anonymized, informed consent was not required. The protocol of this study was approved by the institutional review board (GCI 2020‐06‐006).

Estimation of Ambient Air Pollution Before the Onset of Disease

We used hourly concentrations of particulate matter ≤10 or 2.5 µm in diameter (PM10 and PM2.5, respectively), nitrogen dioxide (NO2), sulfur dioxide (SO2), carbon monoxide (CO), and ozone (O3) measured at 235 air quality monitoring sites located throughout the country. We obtained the air pollution data between 2007 and 2019 from the National Institute of Environmental Research (https://www.nier.go.kr/). The following daily representative concentrations of the 6 pollutants were determined for each child: 24‐hour averages for PM10, PM2.5, NO2, and SO2 and a maximum of 8‐hour moving averages for CO and O3. The maximum concentrations of CO and O3 were used because most of their emission is affected by commuter traffic and sunlight, respectively. We used the air pollutant measurements collected from urban background sites to monitor population exposures. When 2 or more urban background sites were present in 1 district, we averaged their daily means. Because the residential addresses were available at the district level, we assigned district averages of air pollution to children living in the same district as individual exposures as applied in previous studies. , The urban background monitoring sites are mostly located at the community‐service centers in largely populated residential areas. Thus, these monitoring sites may well represent air pollution exposure to residents of the district. Majority of the children under 5 years old lived Seoul Metropolitan Area including Seoul and its neighboring metropolitan cities (Incheon and Gyeonggi‐do), and there are 66 districts with relatively small sizes (median size=44 km2 in 2007). In addition, because temporal variability is much larger than spatial variability in daily averages of air pollution exposures, the difference in air pollution concentrations of participants assessed by area averages and concentrations from the nearest monitor would be small and exposure measurement error of spatial misclassification in our health analysis may not be substantial. Given that the diagnosis of KD requires the presence of fever lasting at least 5 days, the onset of fever can be supposed to be almost 1 week before the IVIG administration. According to the American Heart Association, IVIG treatments to prevent a coronary event are recommended by day 7 of KD, and some patients might have received the IVIG treatment at day 5 or earlier, which is suggested to result in a lower risk of cardiac sequelae of KD. Therefore, we presented a second scenario where the alleged onset of KD is 4 days before IVIG administration. We assumed that the children's addresses remained the same within a month of IVIG treatment.

Weather conditions were extracted from 510 automatic weather stations across the country (https://data.kma.go.kr/) and used to adjust the association between air pollutants and KD. The average daily values of ambient temperature and relative humidity were assigned to individuals according to their districts. When daily average weather conditions for the monitoring station were not available, we imputed the data using a random forest‐based method from the missForest function in R package missForest. ,

Statistical Analysis

We used a time‐stratified case‐crossover design to control short‐term time‐varying confounders and individual baseline risks. In this design, cases serve as their own controls using exposure on the days before or after the case day. Owing to the consistency of the subject across days, this design can control all known and unknown confounding factors that are likely to vary nonrandomly—for example, age, body mass index, socioeconomic factors, health‐related behaviors, and genetic predisposition. The time‐stratified approach avoids bias from the day of week and seasonality by matching the control days on the same day of the week within the same month and year as the case day. For example, if a KD case was identified on the last Monday of January 2019 (January 28, 2019), all other Mondays within January 2019 were assigned as the control days for that case (January 7, 14, and 21, 2019). This approach resulted in 3 control days for each case.

Six pollutants were introduced into the model one at a time. We estimated the effects of air pollutants on the day of disease onset (lag 0), the day before (lag 1), and the 2 days before (lag 2), assuming the immediate effect of air pollution (lags of 1–5 days for the second scenario, Figure 1). , , Within‐individual differences in the air pollution level of case days (district‐average on lag 0–2 days) and control days were assessed using paired t tests. We used Spearman's correlation to examine the pairwise relationships of air pollutants, ambient temperature, and relative humidity. Although PM2.5 is the subset of PM10 particles and thus both are likely to show a high correlation, we analyzed both separately to compare our findings with those of prior studies. We performed conditional logistic regression to estimate odds ratios (OR) of KD according to interquartile range (IQR) increases in each air pollutant concentration on the day of fever onset. We also explored the lag pattern in the effects using single‐day exposure lags of 1 and 2 days to identify the critical exposure period. A spline of daily mean temperature and relative humidity was included in the adjusted models to control potential nonlinear relationships with KD. The most appropriate degrees of freedom of the temperature and relative humidity were determined by minimizing Akaike Information Criterion. Lags for temperature and relative humidity paralleled the modeled air pollution lags. For pollutants that showed significant association with the KD episode, we applied multipollutant models. We conducted sensitivity analyses to explore the robustness of the findings. First, we restricted the data to the Seoul region, where the individual raw data were complete. Second, we changed the degree of freedom for temperature and relative humidity in the adjusted analyses. All statistical analyses were conducted using R version 4.0.3.

Figure 1

Estimation of the onset of Kawasaki disease (KD) based on the day of intravenous immunoglobulin (IVIG) administration in the National Health Insurance claim data.

Lag 1 and 2 are 1 and 2 days before the alleged onset of KD, respectively. A, 7 days before IVIG with lag 0 to 2 days; B, 4 days before IVIG with lag 0 to 5 days.

RESULTS

Among the total of 51 486 patients with KD, 58.2% were male. 48.7% were <24 months old, and 74.4% lived outside of Seoul. The KD episodes were the highest in winter (Table 1). Annual average PM10, PM2.5, NO2, SO2, and CO concentrations decreased from 2007 to 2019 for all days included in the study (Table S1 and Figure S1). Annual average O3 concentration increased. The means of daily PM10 and PM2.5 concentrations over the case and control days during the study period were 47.99 and 25.04 μg/m3, respectively (Table 2). The mean NO2, SO2, and O3 were 26.34, 4.92, and 37.40 ppb, respectively. The mean CO was 0.64 ppm.

Table 1

Demographic Data of Kawasaki Disease Cases (n=51 486) From November 2007 to October 2019

Variable	Frequency (%)
Sex
Male	29 982 (58.2)
Female	21 504 (41.8)
Age group
Less than 12 mo	12 925 (25.1)
12–23 mo	12 166 (23.6)
24–35 mo	11 062 (21.5)
36–47 mo	8999 (17.5)
48–59 mo	6334 (12.3)
Region
Seoul	13 170 (25.6)
Outside of Seoul	38 316 (74.4)
Year of KD onset
2007–2011	15 616 (30.3)
2012–2015	19 009 (36.9)
2016–2019	16 143 (31.4)
Month of Kawasaki disease onset
March–May	12 821 (24.9)
June–August	13 024 (25.3)
September–November	12 026 (23.4)
December–February	13 615 (26.4)

Table 2

Daily Concentration of Air Pollutants, Temperature, and Relative Humidity Measures on Case and Control Days Among 51 486 Patients With Kawasaki Disease Aged Under 5 Years of South Korea, 2007 to 2019

Daily environmental variable	Mean (SD)	Min	Quartile 1	Quartile 2	Quartile 3	Max	Interquartile range
Air pollutants
PM10, μg/m3	47.99 (27.92)	1.33	30.24	42.72	58.92	708.75	28.68
PM2.5, μg/m3	25.04 (12.84)	0.04	16.37	22.86	31.04	161.53	14.67
Nitrogen dioxide, ppb	26.34 (13.21)	0.12	16.42	23.88	33.83	136.00	17.42
Sulfur dioxide, ppb	4.92 (2.52)	0.08	3.21	4.46	6.00	81.83	2.79
Carbon monoxide, ppm	0.64 (0.31)	0.07	0.44	0.57	0.76	5.47	0.32
Ozone, ppb	37.40 (18.94)	0.25	23.62	34.75	49.00	163.79	25.38
Weather conditions
Temperature, °C	12.69 (6.62)	−19.70	10.13	12.90	15.60	34.30	5.47
Relative humidity (%)	64.03 (12.88)	0.10	57.63	65.51	72.28	99.90	14.66

PM10 and PM2.5 indicates particulate matter ≤10 and 2.5 µm in diameter.

In within‐individual comparison of 6 air pollutants, case days showed higher daily mean concentration than control days for PM2.5 (lag 1 day) and SO2 (lag 0–2 days, Table S2). In general, the means of 6 air pollutants showed a low‐to‐moderate pairwise correlation (Table S3). The daily average concentration of CO was highly and positively correlated with NO2 (r=0.62) and moderately correlated with SO2 (r=0.47). The concentration of PM10 was moderately and positively correlated with NO2 (r=0.50) and CO (r=0.54). In general, the daily air pollutant concentrations were negatively correlated with temperature (r ranged from −0.11 to −0.34) and relative humidity (r ranged from −0.02 to −0.09). The mean concentration of air pollution in a given district for a given month or year was relatively stable, for example, in Seoul (Figure S2).

In the single pollutant model adjusting for temperature and relative humidity, the direction of association between air pollutant concentration and KD was positive for PM10, PM2.5, SO2, and CO, specifically at lag 1 (Figure 2A). For example, an IQR increase (per 14.67 μg/m3 change) of PM2.5 concentration was associated with KD at lag 1 (OR, 1.016; 95% CI, 1.004–1.029) (Table S4). An IQR increase (2.79 ppb) in SO2 concentration was associated with KD at all lag days (lag 0: 1.018; 95% CI, 1.002–1.034; lag 1: 1.022;, 95% CI, 1.005–1.038; lag 2: 1.017; 95% CI, 1.001–1.033). The association between KD and an IQR increase (per 25.38 ppb) in O3 was negative at lag 2 (0.979; 95% CI, 0.959–0.999). When we assigned the onset of KD as 4 days before IVIG treatment, the associations of SO2 and O3 with KD remain the same. An IQR increase in SO2 concentration was associated with KD at lag 1 to 2 (lag 1: 1.021; 95% CI, 1.004–1.038; lag 2: 1.017; 95% CI, 1.001–1.033 at lag 2) (Figure 2B). The association between KD and an IQR increase in O3 was also negative at lag 0 to 1 (lag 0: 0.951; 95% CI, 0.953–1.011; lag 1: 0.967; 95% CI, 0.947–0.988).

Figure 2

Associations between Kawasaki disease (KD) and air pollution at different lag days among 51 486 patients in South Korea aged under 5 years with KD, 2007 to 2019.

The associations are shown as odds ratio (OR) with 95% CI for interquartile range (IQR) increases in exposure. All models were adjusted for ambient temperature and relative humidity. A, 7 days before IVIG; B, 4 days before IVIG. CO indicates carbon monoxide; IVIG, intravenous immunoglobulin; NO2, nitrogen dioxide; O3, ozone; PM10 and PM2.5, particulate matter ≤10 and 2.5 µm in diameter; and SO2, sulfur dioxide.

In the 2‐pollutant model that simultaneously included terms for SO2 or O3, which showed significant association with KD, the point estimates did not change compared with the single pollutant model for all lag periods (Figure S3), indicating the independent effects of both pollutants. In the subgroup analysis using only data from Seoul to address exposure misclassification due to imputation, effect estimates were not meaningfully altered, and CIs widened with losses of subjects (Table S5). The change in the degree of freedom for temperature and relative humidity did not substantially affect the estimated effects of all air pollutants within the range of 2 to 10 degrees of freedom. Different smoothing approaches do not change the ORs in the magnitude and direction (result not shown).

DISCUSSION

Using national data of all KD events from the population under age 5 years in South Korea, we observed a positive association between daily variations in ambient air pollution and KD. More specifically, an IQR increase in SO2 and PM2.5 concentrations before the onset of KD were associated with higher odds of KD. Although the effect estimates were generally positive, the association was not statistically significant for PM10, NO2, and CO. The association between KD and O3 was in a negative direction at lag 1 to 2. This association did not change in the 2‐pollutant model for lag 0 to 2 days where SO2 and O3 were concurrently incorporated.

Previous studies showed that air pollution might be positively associated with the risk of KD but did not always reach statistical significance. Similar to our findings, a study of 224 cases in Japan reported that SO2 was significantly associated with an increased KD incidence. A study of 2344 KD cases in Shanghai, China, found a positive but statistically insignificant association between KD incidence and PM10, NO2, and SO2. Another study of 3009 KD cases in 7 metropolitan regions of North America observed some positive effect estimates of PM2.5 exposure on KD, which were statistically insignificant at lagged periods varying from 2 to 14 days. Similarly, a recent single‐center cohort study of 711 KD cases in South Korea showed the null association between KD and ambient PM2.5 without adjusting for weather conditions. Inconsistencies between the previous studies and our current findings could result from a difference in sample size, variation in the concentration of air pollutants, and adjustment of confounders. For example, the study in North America had the mean concentration of PM2.5 ranged from 7.0 to 17.2 μg/m3, which is much lower than those of our study population.

The underlying pathogenesis behind the association between KD and short‐term exposure to air pollutants is not yet fully elucidated. Acute exposure to a high concentration of SO2 is known to cause airway irritation and inflammation, aggravating bronchoconstriction in patients with asthma. When exposed to a high level of PM2.5 and SO2, proinflammatory cytokines and enzymes from injured bronchoepithelial cells may induce a rapid systematic inflammatory reaction, damaging the arterial wall and possibly leading to KD symptoms. , , The other air pollutants are also respiratory irritants causing inflammatory responses. The close‐to‐null association of KD with the other air pollutants may reflect some specific role of SO2 in the pathogenesis of KD, and further investigation is needed to evaluate this potential biological mechanism.

Notably, the inverse association of KD with O3 exposure in the 2 days before the onset of KD (lag 2) of our study was also observed in a study of 695 KD cases in Taiwan, whereas the ambient O3 level decreased during the research period of their study. This finding may be explained by the ozone alert, which is adopted in many countries. , Starting from 2005, this ozone warning system advises children, senior citizens, and patients with respiratory diseases to limit their outdoor activities when the ozone level is considered high (Annual O3 Alert Trend, www.airkorea.or.kr). Except ozone, the studied air pollutants are largely produced from fossil fuel combustion. Imprecise risk estimates for NO2 and CO in our study may be because of different pattern of spatial variation of these pollutants.

Our study had some potential limitations. First, air pollutant data were accessed from ambient air pollution monitoring stations rather than measured at personal exposure levels and could not represent the indoor concentration of air pollutants. Second, data on the potential confounders, including genetic, infectious, and socioeconomic factors, were not available in this study. However, we attempted to account for personal invariant factors such as genetic susceptibilities by applying the case‐crossover design. Third, the date of IVIG treatment was the only available information on the timing of the KD event in our data source. The definition of KD onset based on the day of IVIG administration may lead to potential misclassification of exposure because the date does not necessarily represent the actual timing of the disease. Given that the mean duration of fever before IVIG administration was 6 to 7 days among children under‐5 years diagnosed with KD, , we believe that the possibility of misclassification would be minimal. In addition, laboratory findings were not included in our analysis because they are not available in the National Health Insurance System claim database, which is constructed for administrative purpose. To confirm our findings, future studies may benefit from using clinical inflammatory indicators such as white blood cell count or C‐reactive protein, as intermediate markers for KD.

CONCLUSIONS

This study adds to the growing body of evidence implicating the positive association between short‐term variations in air pollution and the development of KD, supporting the hypothesis that the short‐term elevations in PM2.5 and SO2 may trigger the onset of KD among children under 5 years of age. This finding needs to be confirmed in future studies with different study designs and among different populations.

Sources of Funding

This research was supported by the National Research Foundation of Korea grant (NRF‐2016R1D1A1B03933410 and 2018R1D1A1B07048821), which is funded by the Korean Government. This funding source had no role in study design, in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the article for publication. The contents of this report are solely the responsibility of the authors and do not necessarily represent the official views of the sponsoring organizations.

Disclosures

None.

Tables S1–S5

Figures S1–S3

Click here for additional data file.