Acute effects of ambient air pollution on lower respiratory infections in Guangzhou children of China

Background Daily concentrations of air pollution are associated with lower respiratory diseases. This study investigated the short-term association of ambient air pollution with daily hospital admissions due to pneumonia among children aged 0–17 in Guangzhou city of China. Methods Daily ambient air pollution concentrations were extracted from the databases of the Chinese Environmental Monitoring Center. Children hospital admission counts for pneumonia during 2013–2018 were sourced from the Guangdong Maternal and Child Healthcare Hospital. Associations between outdoor air pollution levels and hospital admissions were estimated for time lags of zero up to seven days using Quasi-Poisson regression models, adjusted for seasonal variations, meteorological variables, day of week and holidays. The associations between clinical pathogenic microorganism inspection results for pneumonia and air pollutants were calculated using Lasso regression models. Results Ambient air pollutants were all positively associated with children hospital admissions due to pneumonia of all ages. Signicant associations were found for air pollutants except for inhalable particulate matter (PM) 10 µm in aerodynamic diameter (PM 10 ) in children aged 0–17 years. Increments of an interquartile range (IQR) (279.10µg/m 3 and 28.42µg/m 3 , respectively) in the 7-day-average level of carbon monoxide (CO) and nitrogen dioxide (NO 2 ) were associated with a 26.17% (95% condence interval (CI) 1.40%-56.98%) and 25.09% (95%CI 0.54%-55.64%) increase in pneumonia hospitalizations for children aged 6–17, respectively. An IQR increase in CO concentrations (279.10µg/m 3 ) was associated with a 15.15% (95%CI 4.34%-27.08%) increase in pneumonia hospitalizations for children aged 1–5. Estimates for CO were statistically signicant among children aged 1–5 years in summer. The associations remained stable in two-pollutant models. Daily cases of microbial detection for pneumonia were positively associated with daily NO 2 concentration. The pneumonia hospitalizations due to Mycoplasma pneumonia, Flu A virus and Flu B virus, the predominant pathogenic microorganisms detected in children aged 0–5 are apparently associated with levels of PMs, CO, NO 2 and O 3 . Conclusions Strong associations detected cases of these two kind of pathogenic microorganisms accounted for 25.74%. RCs of daily mean PM 10 , NO 2 , CO and O 3 levels were positive to Mycoplasdema pneumonia which accounting for 24.89% in infants and 63.36% among the 1–5 year age group simultaneously. RCs of daily mean PM 2.5 and NO 2 levels were positive to Inuenza A and B virus simultaneously. RCs of daily mean O 3 , PM 2.5 and PM 10 levels were mainly positive for children less than 5 years-old. RCs of daily mean SO 2 , NO 2 and CO levels were mainly positive for children aged 6–17 years. The pneumonia hospitalizations due to Mycoplasma pneumonia, Flu A virus and Flu B virus in children aged 0–5 are apparently associated with the levels of air pollutants (i.e. PMs, NO 2 , CO and O 3 ). And pathogenic microorganisms, such as Mycoplasma pneumonia, Flu A virus and Flu B virus, might be possibly carried by PMs, which increased risks of the acute lower respiratory infections in children aged 0–5 years. We fortunately found a study from Urumqi city of Western China, which reported that the microorganisms responsible for human allergy and respiratory disease carried by PM 10 and PM 1 had been analyzed during winter. Their results showed that the bacterial community was mainly composed of Proteobacteria, Firmicutes and Actinobacteria. The sequences of several pathogenic bacteria and opportunistic pathogens were also detected, such as Acinetobacter, Delftia, Serratia, Chryseobacterium, which might impact on immunocompromised populations (elderly, children and postoperative convalescence patients) study and of for in


Background
Pneumonia in children refers to pulmonary in ammation in the population aged less than 18 years caused by different pathogenic or other factors. It easily occurs in all seasons, mainly in spring and winter [1]. Severe pneumonia is the major cause of death of children under 5 years old in China, in which, most of patient die of various kinds of pneumonia. Pneumonia pathogens include bacteria, virus and atypical pathogens such as Mycoplasma pneumoniae and Chlamydia pneumonia etc. The Western Paci c region had an estimated 0.11 pneumonia episodes per child-year with 61,900 pneumonia-related deaths in children less than 5 years of age in 2011 [2]. Furthermore, family jam [3] and malnutrition [4] also result in pulmonary in ammation. Time series analysis shows that pneumonia hospitalizations in children are associated with ambient air pollution level [5][6][7][8]. A child's respiratory system is susceptible to the adverse health effects of air pollution due to lower immunity.
The association between air pollution and hospitalization for acute respiratory infection (ARI) has been investigated worldwide [9]. Increased concentrations of nitrogen dioxide (NO 2 ) and sulfur dioxide (SO 2 ) were associated with increased admissions of pneumonia in Vietnamese children in the dry season (November-April), with excess risks of 8.50% (95% con dence interval (CI) 0.80-16.79) and 5.85% (95%CI 0.44-11.55), respectively. Daily concentrations of particulate matter (PM) 10 µm in aerodynamic diameter (PM 10 ) could be positively associated with increased pneumonia admissions in children in the dry season. Negative associations between air pollutants and pneumonia admissions in children were observed in the rainy season (May-October) in Vietnam [10]. Daily concentrations of PM 10 and ozone (O 3 ) were strongly associated with hospitalizations of lower respiratory tract infection in children aged 2-5 [11]. In the meantime, a report from Malaysia pointed out that increased acute lower respiratory tract infection admissions in children under 18 were associated with low rainfall but not PM 10 nor air pollutant index [12].
The air quality is getting worse in the progress of industrialization and urbanization process in China, which increases the disease burden of respiratory system [13]. The air pollutants are acutely and chronically associated with number of hospitalization, morbidity, mortality, clinical symptoms and pulmonary function change for various diseases in China [14]. If the total suspended particles (TSP) in the atmosphere reached to 184 µg/m 3 in Northern China, life time of local people would be reduced by 5.5 years [15]. Daily concentration of PM 2.5 µm in aerodynamic diameter (PM 2.5 ) was associated with all-cause mortality in Shanghai city (risk ratio (RR) 1.0068, for time lag0, 95%CI 1.0013-1.0123) [16]. PM 2.5 concentration was positively associated with daily hospital admissions for disease of respiratory system in Ji'nan, a provincial capital city in Northern China [17]. Elevated PM 2.5 and PM 10 were signi cantly associated with increased emergency department (ED) visits for pneumonia, respiratory tract infection (RTI) and coronary heart disease at both lag0 and lag0-3 in Guangzhou city. A 10 µg/m 3 increment of PM 2.5−10 (PMc) was estimated to increase ED visits for pneumonia by 6.32% (95% CI 4.19%-8.49%) and for RTI by 4.72% (95% CI 3.81%-5.63%), respectively. PMc showed stronger cumulative effects on asthma in children than elderly [18]. Daily concentration of SO 2 was apparently associated with respiratory mortality in Xi'an (RR 1.4200, 95%CI 0.8270-2.8540), a provincial capital city in Western China [19]. The air pollutant levels in Wuhan city of Central China and Lanzhou city of Western China were obviously and positively correlated to pneumonia hospitalizations in children [20].
Data of meteorological factors were collected for the same period from Reliable Prognosis website (http://rp5.ru/docs/about/cn) [24], which provides every three hours recording of temperature (in℃), relative humidity (in percent), wind speed (in m/s), rainfall (in mm), atmospheric pressure (in mmHg) and horizontal visibility (in km) at the Guangzhou Airport Meteorological station (113°30'97''E, 23°39'29''N). Data were averaged by calendar day to provide 24-h means of these meteorological data.
Children hospital admission records of pneumonia (International Classi cation Diseases 11th revision (ICD11) code CA40) sourced from the Guangdong Maternal and Child Healthcare Hospital (GDMCHH), covering a total of 17,149 anonymous cases from October 28, 2013 to June 30, 2018. The records include clinical diagnosis, admission number, gender, age and pathogenic microorganism test results of sputum cultivation (including Chlamydia pneumoniae, Haemophilus, Mycoplasma pneumoniae, In uenza A virus, In uenza B virus, Parain uenza virus, Adenovirus and Respiratory syncytial virus). GDMCHH is a large-scale tertiary hospital with healthcare, medical, educating, scienti c research, training and technical guidance. It is the educating hospital of eight universities including Sun Yat-Sen University, while it also establishes cooperation with Boston Children's Hospital of Harvard Medical School. Having two hospital districts with 1,500 beds and one branch of the Guangdong Cord Blood Bank. GDMCHH is one of the best maternal and child healthcare hospital in Southern China. Its clinical diagnosis records greatly re ect the children incidence of pneumonia in Guangzhou city. This study was conducted on October 1st, 2019.
All of the children patient records were anonymized and de-identi ed, the study underwent an ethical review by the ethical committee of GDMCHH (Approval number: 201901121, September 2019).

Data analysis
To study the association between ambient air pollution and daily counts of hospital admissions for pneumonia, we used the semi-parametric generalized additive quasi-Poisson regression models with a log-link function and adjustment for over-dispersion, adjusting for potential confounders.
Thin plate spline functions were used to capture the time trends and seasonal variations. The potential confounders were integrated into the models which including daily mean temperature, relative humidity, wind speed, rainfall, day of the week (DOW) and holidays. DOW was treated as a categorical variable with values ranging from 1 (Sunday) to 7 (Saturday). We had adjusted for meteorological factors averaged over the same day and the day before (lag 0-1) and over the ve days preceding the period (lag 2-6) and the interaction terms between them. Smoothing spline functions were used to non-parametric smooth of the meteorological data. The time nonlinear independent variables (time = 1-1707) in the time series data were tted and the most suitable smooth spline function degree of freedom was selected by the Generalized Cross-Validation (GCV). When the basic model was established, the air pollutant concentrations were added into the model to be the linear variables, we computed the associations between the two day moving average (lag 0-1), three-day moving average (lag 0-2), and seven-day moving average (lag 0-6) air pollutant concentrations and children hospital admissions. To facilitate comparing the effects of air pollutants, results were reported as RRs of hospital admissions with 95% CI for the one interquartile range (IQR) increment in the level of the respective pollutant variable. Meanwhile, analyses were performed for gender, ages (infants, children aged 1-5, aged 6-17 and children of all ages 0-17) and seasons (spring: March -May, summer: June -August, fall: September -November, winter: December -February). Statistical signi cance was de ned as two-tailed p-value < 0.05.
We also built two-pollutant models combining the pollutants which including PM 10 , PM 2.5 , SO 2 , NO 2 , CO and O 3 . Variance in ation factor (VIF) was used to evaluate the multicollinearity in these models. All pairs of pollutants had VIF-values maximum of 7.6 that is below the threshold of 10 indicating the strong multicollinearity [25].
However, investigations of associations among meteorological factors, air pollutants and the clinical pathogenic microorganism inspection results for pneumonia in children aged 0-17 in China remain sparse. Because of too many zero-values for the detected results of sputum microbial cultivation which obeyed the zero in ation Poisson distribution, these data were not applicable to the ordinary Poisson regression model. In order to nd interpretively independent variables as well as improve the explanatory power and accuracy of the models' prediction, we used Lasso regression model to explore the associations between ambient air pollutants and daily detected results of sputum microbial cultivation after considering the effects of meteorological factors, DOW, holidays and seasons. The initial Lasso regression model was derived for the daily cases of microbial detection for pneumonia, and air pollutants (lag 0-6) as well as meteorological factors. The cross-validation method was used to select the Lambda with the least average error to optimize the model. Finally, we calculated the root mean-square error (RMSE) which indicated the quality of the models. All statistical analyses were completed using the "mgcv", "spline" and "glmnet" packages of R language (version 3.5.3, http://www.r-project.org).

Results
Descriptive statistics for daily hospital admissions and detected results of sputum cultivation present in Table 1, while Table 2 presents descriptive statistics for daily air pollutant concentrations and meteorological variables. The investigation contained 17,149 hospital admissions for pneumonia (i.e., about average 10.04 cases a day). In daily hospital admissions, the counts of male were about 1.8 times than that of female. Out of all hospital admissions, infants, children aged 1-5 years and 6-17 years accounted for 67.30%, 27.70% and 5.03%, respectively. The daily hospital admissions were maximum in spring. The microbiological detection items for children hospitalizations with pneumonia included Chlamydia pneumoniae, Haemophilus, Mycoplasma pneumoniae, In uenza A virus, In uenza B virus, Parain uenza virus, Adenovirus and Respiratory syncytial virus. The daily hospital admissions accounted 46.70% for pulmonary Mycoplasma pneumoniae, 21.60% and 9.98% for Flu A and B virus respectively, but 1.99% for Haemophilus. There were seasonal preferences for daily hospital admission counts due to different pathogenic microorganisms.
Daily mean concentration of PM 2.5 was 39.80µg/m 3 , which exceeded the 24-h mean concentration of ne particles (25µg/m 3 ) by 37.19% set in the World Health Organization (WHO) guidelines for air quality. Daily mean concentration of PM 10 was 60.80µg/m 3 , which exceeded the 24-h mean concentration of inhalable particles (50µg/m 3 ) by 17.76% set in WHO guidelines for air quality. Daily mean concentrations of CO, NO 2 , O 3 and SO 2 were 935.00µg/m 3 , 37.30µg/m 3 , 46.90µg/m 3 and 12.90µg/m 3 , respectively, which were all lower than WHO standards. Daily concentrations of air pollutants were highest in winter and lowest in summer, except for O 3 . While, daily concentration of O 3 was highest in autumn and lowest in winter. It indicated that daily mean concentrations of air pollutants were associated with daily mean temperature, relative humidity and rainfall. Daily means of air pollutants were signi cantly correlated with each other. Daily means of PM 2.5 was strongly correlated with daily means of PM 10 (Spearman rank correlation coe cient, r = 0.97). Both daily means of PM 2.5 and PM 10 were correlated to daily means of SO 2 , NO 2 and CO (0.50≤ |r| ≤0.80). And as temperature decreased, the correlation coe cients between daily means of SO 2 and PM 2.5 or PM 10 gradually increased. Daily means of O 3 was remarkably correlated to daily means of PM 2.5 and PM 10 in summer and autumn (0.50≤ |r| ≤0.80) (see additional le 2). Daily mean temperature, relative humidity and rainfall in Guangzhou city were 21.70℃, 78.00% and 6.70mm, respectively.
The RRs of hospital admissions for pneumonia per IQR of the seven-day (lag 0-6) mean concentrations of the pollutants present in Table 3. Daily hospital admissions for pneumonia were positively associated with all pollutants for children of all ages. Statistically signi cant associations were observed for all pollutants except for PM 10  The RRs for pneumonia hospitalization were higher among children aged 1-5 and 6-17 years as compared to infants, except for PM 2.5 . For example, the RR for an IQR increase in CO was highest among children aged 6-17 years and lower in infants (RR = 0.97, 95% CI 0.89-1.05). Similar modes were found for lag 0, lag 0-1 and lag 0-2 (see additional le 3, additional le 4 and additional le 5). The RRs per IQR ranged from 0.97 to 1.05 for lag 0, from 0.99 to 1.05 for lag 0-1, from 1.00 to 1.06 for lag 0-3 and from 0.95 to 1.06 for lag 0-6 means for all pollutants in children under 17 years.
The season-speci c RRs of pneumonia hospitalizations per IQR increase in the seven-day (lag 0-6) moving average concentrations of ambient air pollutants present in Fig. 1. Daily hospital admissions for pneumonia were positively associated with PM 2.5 and PM 10 in gender and all age groups.
Daily hospital admissions for pneumonia were statistically signi cant for all pollutants in children aged 6-17 years, except for PM 2.5 and PM 10 .
Associations with pneumonia hospitalizations were signi cantly different for pollutants in different seasons. For example, the RR for an IQR increase in SO 2 was higher among children aged 6-17 years in the winter (RR = 1.43, 95% CI 1.09-1.87). The RR per IQR for NO 2 was higher in the summer (RR = 3.99, 95% CI 1.42-11.26). And the RR per IQR for CO was higher in the spring (RR = 1.60, 95% CI 1.08-2.37). Similar modes were found for lag 0, lag 0-1 and lag 0-2 (see additional le 3, additional le 4 and additional le 5).  Daily hospital admissions and interquartile range units presented in Table 2. Risk ratios (RR) estimated from Quasi-Poisson regression models, adjusting for secular trends and seasonal variation, day of the week, holiday, in uenza epidemic, and meteorological factors including temperature, relative humidity, pressure, horizontal visibility, precipitation and wind speed average. *p < 0.05. ** p < 0.01 (Wald χ 2 test).
Two-pollutant models presented in Fig. 2 and additional le 6. Estimates for NO 2 were the largest (per IQR) after inclusion of PM 2.5 and PM 10 , but stable after inclusion of SO 2 and CO. Effects of CO were higher after inclusion of PM 2.5 and PM 10 , but stable after inclusion of other pollutants. In contrast, effects of O 3 all dropped or lost statistical signi cance after inclusion of PM 2.5 , PM 10 , SO 2 or NO 2 in children aged 6-17 years. Results for pneumonia and O 3 were also insensitive to the inclusion of other pollutants in other age groups (see additional le 6).
The season-speci c two-pollutant models of NO 2 and CO for age groups presented in Fig. 3. Daily hospital admissions for pneumonia were statistically signi cant difference in different seasons. Effects of NO 2 were higher in summer and autumn after inclusion of PM 2.5 , PM 10 , SO 2 , CO and O 3 .
Furthermore, estimates for NO 2 were the larger with the increase of age. Estimates for CO were somewhat less sensitive to seasonal variation. Effects of CO were statistically signi cant among children under 5 years old in summer. But, effects of CO were signi cantly higher in spring after inclusion of PM 2.5 , PM 10 , SO 2 or NO 2 in children aged 6-17 years.
The descriptive statistics of the clinical pathogenic microorganism inspection results for pneumonia presents in additional le 7. The detected cases of pulmonary Mycoplasma pneumoniae were the most predominant. The positive rate of detection for female was higher than that for male, and the positive rate was the largest in the age group 1-5 year. The positive rate minimized in winter for all kinds of pathogenic microorganisms, except that for Chlamydia pneumoniae and Hemophilus. The cases of microbial detection due to pneumonia increased initially, but decreased during 2014 to 2018.
The Lasso regression model was derived for the daily cases of microbial detection due to pneumonia and air pollutants (lag 0-6), combined with meteorological factors. Regression coe cients (RCs) and root mean square errors (RMSEs) of models present in Table 4  Lots of epidemiological and clinical studies indicated that ambient particulate matter (PM) in air pollution was strongly associated with increased cardiovascular disease, respiratory disease, chronic obstructive pulmonary disease and heart disease in urban residents [27]. PM 10 could be positively associated with increased pneumonia admissions in children in the dry season [10]. Daily mean PM 10 levels were associated with prolonged hospitalizations in children aged 2-5 years in Vietnam [11]. Unexpectedly, we nd that the associations between daily mean PM 10 levels and pneumonia hospitalizations of children were not statistically signi cant in all age groups. Although, a study showed positive associations between ARI and both NO 2 and PM 10 during the dry season 2003-2005 [28], but the results were not statistically signi cant given the rather short series and limited statistical power [29]. Another report pointed out that increased acute lower respiratory tract infection admissions in children under 18 were associated with low rainfall but not PM 10 nor air pollutant index [12]. So, why are the effects of PM 10 completely different in different regions? Is it possible that the effect of PM 10 is mainly related to climate factor, immunity and age? Sure, the differences of the effect of PM 10 may also be determined by genetic predisposition. We interestingly nd that air pollutants might exacerbate genetic variations associated with asthma, including GLUTATHIONE-S-TRANSFERASE M1 (GSTM1) and GLUTATHIONE-S-TRANSFERASE P1 (GSTP1) gene. Among them, GSTP1 modi ed the delayed effects of PM 10 [33]. This may be an indication that smaller size PMs induce stronger in ammatory responses, particularly the ultra ne particles that can penetrate deeply into lung alveoli or be transported to other organs [34,35]. These results also suggested that high smaller size PMs exposures might adversely in uence both development of the innate immune system and development of lung function of infants. Hence, to protect infancy vulnerability of the rapid lung and immune system development from high levels of air pollution exposure is very important during the early months of life.
Our results demonstrated that the association of CO with hospital admissions due to pneumonia reached statistical signi cance in children aged 1-5 increased. An investigation showed that acute lower respiratory infection admissions among children under 5 years of age were generally positively associated with ambient levels of PM 10 , NO 2 , and SO 2 during the dry season, but not the rainy season, and negative results in the rainy season could be driven by residual confounding present from May to October in Ho Chi Minh City, Vietnam [10]. Daily concentration of SO 2 was apparently associated with respiratory mortality in Xi'an city [19]. Daily concentrations of CO, NO 2 , SO 2 , and PM 10 were signi cantly associated with increased risk of both cardiovascular and respiratory hospital admissions, whereas O 3 was associated with only respiratory hospital admission [36,37]. The observed seasonal difference in hospital admissions is larger than what one expects based on the difference in air pollution alone. The latter is though only one out of many determinants of hospital admissions.
The results of two pollutant models are presented in Fig. 2 and additional le 6. Basically, two-pollutant models could be used to evaluate the possible roles of single pollutants. Pneumonia related estimates for NO 2 were higher after inclusion of PM 2.5 and PM 10 in summer and autumn, but stable after inclusion of SO 2 and CO. Estimates for CO were statistically signi cant among children aged 1-5 years in summer. The two-pollutant models revealed consistent patterns across all outcomes luckily as shown in additional le 6.
Another feature of this study is to demonstrate associations between air pollutants and the clinical pathogenic microorganism inspection results in the cases of hospitalization due to pneumonia among Guangzhou children. might be possibly carried by PMs, which increased risks of the acute lower respiratory infections in children aged 0-5 years. We fortunately found a study from Urumqi city of Western China, which reported that the microorganisms responsible for human allergy and respiratory disease carried by PM 10 and PM 1 had been analyzed during winter. Their results showed that the bacterial community was mainly composed of Proteobacteria, Firmicutes and Actinobacteria. The sequences of several pathogenic bacteria and opportunistic pathogens were also detected, such as Acinetobacter, Delftia, Serratia, Chryseobacterium, which might impact on immunocompromised populations (elderly, children and postoperative convalescence patients) [38]. However, to our knowledge, no previous study has studied the association between outdoor air pollutants and daily cases of microbial detection for pneumonia in children. Nowadays, further investigations are warranted.
This study has some limitations. The GDMCHH is the tertiary hospital, children with severe diseases might make up a larger proportion than in other hospitals. No doubt, outpatients were excluded from our study. Therefore, to obtain a more convincing explanation of the effects of air pollutants among children aged 6-17 years, the data from other local tertiary hospitals would be used to analyze. The effect estimates in our model are based on the sample size. If our sample size was large enough, the conclusions should be more accurate.

Conclusion
In summary, we have found the positive associations between children's hospital admissions for lower respiratory infections and air pollutants as well as associations between air pollutants and the clinical pathogenic microorganism inspection results in the cases of hospitalization due to pneumonia among Guangzhou children. Signi cant associations were found for air pollutants except for PM 10 in children aged 0-17 years. Our study suggested that short-term exposures to air pollutions, especially to CO and NO 2 , were associated with increased risk of hospital admissions for pneumonia of children under 17 years old. The pneumonia hospitalizations due to Mycoplasma pneumonia, Flu A virus and Flu B virus, the predominant pathogenic microorganisms detected in children aged 0-5 are apparently associated with levels of smaller size particulate matters, carbon monoxide, nitrogen dioxide and ozone. We suggest the Guangzhou Municipal Government should make more efforts to improve the local air quality.