Impact of Air Quality on COVID-19 Confirmed and Death Cases: Evidence from China

doi:10.21203/rs.3.rs-1761728/v1

Download PDF

Research Article

Impact of Air Quality on COVID-19 Confirmed and Death Cases: Evidence from China

https://doi.org/10.21203/rs.3.rs-1761728/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Coronavirus disease 2019 (COVID-19) has triggered the most significant public health crisis in recent years. However, studies on the impact of air quality on COVID-19 confirmed cases and death cases are still limited. This study examines the impact of the air quality on daily COVID-19 confirmed cases and deaths in prefecture-level cities in China from January 22, 2020, to December 31, 2020. The zero-inflated Poisson regression is applied in this paper to explore the impact of air pollution levels on daily confirmed cases and daily death cases. We find significant evidence for air quality index (AQI) to impact daily confirmed COVID-19 cases positively. By further decomposing AQI to its components, we find a significant positive impact from SO₂, NO₂, O₃, and CO on daily confirmed COVID-19 cases. Results also show a positively significant impact from PM₁₀, SO₂, NO₂, O₃, and CO on daily dead COVID-19 cases. This study provides new insights into the relationship between air quality on daily COVID-19 confirmed cases and death cases in China and has direct implications for COVID-19 epidemic prevention and control.

COVID-19

Air pollution

China

Confirmed

Deaths

COVID-19 has caused the most significant public health crisis in recent years. To date, the COVID-19 pandemic has caused 386,548,962 confirmed cases worldwide, including 5,705,754 deaths, and has led to a reduction in global GDP. COVID-19 has seriously threatened the health of human beings and society, and the prevention and control of COVID-19 have become an important content of public health research.

Medical papers have examined a range of physiological mechanisms, examining whether acute air pollution can affect the likelihood of infection and the severity of respiratory infections such as SARS (Ciencewicki and Jaspers, 2007; Park et al., 2020). Among these studies, three mentionable explanations link air quality to the spread and development of infectious diseases are proposed. First, exposure to air pollution may exacerbate the progress of respiratory diseases (Analitis et al., 2006). Inhaled small virus particles can carry the virus deeper into the alveoli, increasing the risk of infection spreading (Yu et al., 2004). Second, exposure to air pollution increases the inflammatory response while reducing the immune response to new infections (Contini and Costabile, 2020; Martelletti and Martelletti, 2020). The higher the level of air pollution, the more people are prone to more severe symptoms and breathing difficulties. Oxidative pollutants in the air can also interfere with the immune system and impair the lungs' ability to clear viruses (Yousefi et al., 2021). Third, the higher the level of air pollution, the longer the virus stays in the air, and the greater the risk of infection (Frontera et al., 2020). These mechanisms may also play a role in COVID-19 cases (Contini and Costabile, 2020), which are tested in this study.

Several epidemiological studies have shown empirical links between air quality indicators and the number of respiratory virus infections and deaths. Most of these studies are based on cross-sectional or panel data containing regional and time-series variations. Few of the existing studies focus on causal effects.

For studies based on cross-sectional and panel data, there was a positive correlation between regional air pollution and local severity of the COVID-19 pandemic, as well as severe acute respiratory disease and influenza-like infection. Some studies attempted to link high levels of air pollution to confirmed cases of COVID-19 in Northern Italy and Dutch (Pansini and Fornacca, 2020; Andrée, 2020). This kind of evidence is also found in China, where Zhu et al. (2020) found a significantly positive spatial association between PM_2.5 and COVID-19 infection using the generalized additive model (GAM) based on case-based data between January 23, 2020, and February 29, 2020. Jiang and Xu (2021) also studied the association between COVID-19 deaths and air quality in Wuhan, China. Using Pearson and Poisson regression models, they found that daily COVID-19 deaths were positively correlated with AQI and PM_2.5. Vali et al. (2021) studied a similar topic based on the data of different countries around the world. However, the weather data as a control variable was not accurate enough due to the inconsistent territorial area of countries.

Studies focusing on other pandemics found similar relationships. A study studying influenza found a causality between PM_2.5 and weekly influenza cases in Taiwan using the generalized autoregressive conditional heteroscedastic (GARCH) model, where the effect is stronger on the elderly. A similar relationship is also found in the SARS pandemic in China (Cui et al., 2003). They use the air pollution index (API) to measure air quality and find a positive correlation between air pollution and the fatality rate of SARS in China.

Few studies focus on the causal relationship between air pollution and cases. Persico and Johnson (2020) showed a positive impact of API on COVID-19 confirmed cases and death cases in the U.S. A similar result is also found by Deryugina et al. (2019), where their study is also conducted in the U.S. and found the contribution of PM_2.5 to elderly mortality.

Studies on the impact of air quality on COVID-19 are still limited. To the best of my knowledge, there is not any literature studying the impact of air quality on COVID-19 in China based on evidence from all prefecture-level cities. Moreover, existing literatures are studying similar impact in China using Poisson regression model, whereas the excess zeros exhibit in the data indicates zero-inflated Poisson regression model is required to deliver better and more reliable result. The travel intensity is also not considered in previous literature, however, the decrease of travel intensity due to travel restrictions may inversely impact the air quality, leading to potential bias in previous regression results. Effective control and prevention need more information, and our paper provides more insights into the current research.

2.1 Data

We obtain the COVID-19 confirmed case and death data from AkShare, an open-source financial data interface library based on Python, which summarized the data interface of Baidu and NetEase and provided me the data related to COVID-19. Daily air quality data are obtained from National real-time release platform for urban air quality, and weather condition data are obtained from National Oceanic and Atmospheric Administration (NOAA). With-in city travel intensity data comes from Baidu Huiyan.

2.1.1 Sample construction

The sample contains data from China and is studied from the prefecture city level. All the prefecture-level cities are covered at the beginning, where after data processing, we have a total of 202 cities. The study period starts from 2020.01.22 to 2020.12.31. Observations with missing data are excluded. Our working sample contains a total of 12139 observations.

Table 1 provides descriptive statistics for all substantive variables included in our analyses. The average numbers of daily confirmed cases and death cases are 4.55 and 0.21, respectively. The average AQI is 66.75, which is classified as the second pollution level according to the Chinese Ambient air quality standard. At this level, some pollutants will have a weak impact on the health of a very small number of sensitive people.

Table 1

Descriptive statistics result
Variable	Mean	Std. Dev.	Min	Max
PM_2.5	43.60	39.67	1	825
PM₁₀	68.91	84.61	1	3228
SO₂	10.17	10.06	1	443
NO₂	26.15	19.61	1	142
CO	0.79	0.42	0.1	6.45
O₃	51.25	25.30	1	160
AQI	66.75	51.74	7	500
Temperature	90.57	90.05	-311	325.8
Dew_temp	3.39	115.11	-800	262.5
Pressure	10210.85	75.73	9896.20	10521.75
Wind_Direc	178.24	80.13	0	360
Wind_Speed	27.98	16.56	0	160.4
Sky_Coverage	4.65	2.95	0	9
Travel	3.61	1.33	0.30	8.88
Dead_daily	0.21	3.86	0	216
Confirmed_daily	4.55	129.18	0	13436

2.1.2 Confirm and death cases

Outcome variables confirmed and death cases are defined as the daily increased number of confirmed and death cases.

2.1.3 AQI

The air quality index (AQI) describes how clean or polluted the air is. A higher AQI value indicates that air pollution is more severe and harms to human health is greater. The pollutants involved in the AQI evaluation are SO₂, NO₂, PM₁₀, PM_2.5 O₃, and CO, which are independently tested from the AQI index to get their impact on COVID-19 cases.

2.1.4 Covariates

As suggested by other studies, weather conditions are included as covariates in our study (Xie and Zhu, 2020; Islam et al., 2021; Fareed et al., 2020). More specifically, air temperature, dew point temperature, sea level pressure, wind direction, wind speed rate, and sky condition total coverage code are included. Dew point temperature is defined as the temperature to which a given parcel of air must be cooled at constant pressure and water vapor content for saturation to occur. Sea level pressure is defined as the air pressure relative to the mean sea level. Wind direction is the angle, measured in a clockwise direction, between true north and the direction from which the wind is blowing. Wind speed rate is the rate of horizontal travel of air past a fixed point. Sky condition total coverage code is the code that denotes the fraction of the total celestial dome covered by clouds or other obscuring phenomena. Air temperature, dew point temperature, sea level pressure, and wind speed rate are scaled by a factor of 10.

There is an endogenous problem that when the COVID-19 infection number is high, the government will post regulations, including quarantine and travel restrictions, which decrease the travel intensity and thus improve the air quality because of re-suspension and emission are decreased. This impact from the outcome variable to the independent variable would lead to the endogenous problem caused by the reciprocal causation. Therefore, the model includes travel intensity as a control variable to address this endogenous problem.

2.2 Method

This study examines the impact of air quality on daily COVID-19 cases and death cases in China at the prefecture city level from January 22, 2020, to December 31, 2020. We use the zero-inflated Poisson regression model (ZIPR) to explore the impact of air pollution levels on COVID-19 cases and deaths.

The discrete Poisson regression model is often used to fit count data, such as the number of events and items. However, if the dataset contains too much zero, the dataset distribution will deviate greatly from the standard Poisson distribution. The ZIPR theory suggests that the extra zeros are generated by a separate process and can be modeled independently. It can be effectively used when the observed data is presented as non-negative integers with extra zeros, which leads to an excessive dispersion of dependent variables (Omer et al., 2021). The model can be divided into two sub-models: the Logit model, which is used to fit the excess zeros, and the Poisson counting model, which is used to fit other values. The ZIPR model assumes the dependent variable follows the mix distribution

$$\left\{ \begin{gathered} P({y_i}{\text{=0 | }}{x_i}{\text{)=}}\theta \hfill \\ P({y_i}{\text{=}}j{\text{ | }}{x_i}{\text{)=}}\frac{{(1 - \theta ){{\text{e}}^{-{\lambda _i}}}\lambda _{i}^{j}}}{{j!(1 - {{\text{e}}^{-{\lambda _i}}}{\text{)}}}}{\text{ (}}j{\text{=1,2,}}...{\text{)}} \hfill \\ \end{gathered} \right.$$

where ${\lambda _i}=\exp (x_{i}^{'}\beta )$. Therefore, we could let θ depend on the independent variable z_i (z_i could equal to x_i or overlap with x_i) and use the Logit model to estimate this binary selection problem (y_i=0 or y_i>0). We then use maximum likelihood estimation to estimate the above model and get ZIPR.

The histograms of our independent variables are shown in Fig. 1. We could see that all independent variables, including daily confirmed and dead cases, are all with excess zeros. Therefore, the ZIPR model is a good fit for this study. Since work by Wilson (2015) has shown that the Vuong test is inappropriate for testing zero inflation, we do not use the Vuong test as support for the adaption of the ZIPR model.

For the zero-inflated (Logistic) regression part, variables are selected as follows: remove the model term (other than the intercepts) with the largest p-value over 0.2 and rerun. Repeat until all p-values are less than the threshold of 0.2.

Lag terms for air quality variables (including AQI and its components) are created because the effects of these variables on the outcome variables may not be real-time effects but could be intertemporal effects that develop over time. To address the potential multi-collinearity problem, we conduct the uncentered variance inflation factor (VIF) test for independent variables and eliminate variables with a VIF value higher than five. Finally, we include lag terms for each variable except PM_2.5, because PM_2.5 shows a VIF value larger than the threshold of five. This could be caused by PM_2.5 having a lower deposition value and is easier to be re-suspended, and therefore shows a more severe multi-collinearity problem. ALL these variables are lagged for seven terms.

3.1 Contribution from AQI

The result of the impacts of AQI on COVID-19 confirmed and death cases are shown in Table 2. The impact of AQI is significant on all the outcome variables, wherein the parentheses under each regression coefficient are the t-statistics value. The impact of lagged AQI is also significant from the result. As stated in Section 3, we test for multi-collinearity to ensure the lagged terms do not cause severe problems. This result indicates that the contribution from both AQI and its lagged terms are significant to daily COVID-19 cases in China. A positive impact from AQI is found in daily confirmed cases, whereas a negative impact is found in daily death cases. Contributions from all added weather controls are significant, which would provide insights for future studies.

Table 2

ZIPR results of AQI on COVID-19 cases
	(1)	(2)
	Confirmed_daily	Dead_daily
AQI	0.0043^***	-0.0026^**
	(25.3466)	(-2.5605)
AQI_lag1	-0.0084^***	-0.0026^**
	(-42.3126)	(-2.4905)
AQI_lag2	0.0062^***	-0.0022^***
	(41.0494)	(-2.7256)
AQI_lag3	-0.0017^***	-0.0060^***
	(-11.0902)	(-7.4555)
AQI_lag4	0.0024^***	0.0021^***
	(16.2630)	(2.8074)
AQI_lag5	-0.0036^***	-0.0022^***
	(-23.1398)	(-2.8393)
AQI_lag6	-0.0087^***	-0.0039^***
	(-62.6341)	(-4.6644)
AQI_lag7	0.0138^***	0.0080^***
	(157.6503)	(12.5438)
travel	-3.7708^***	-3.3852^***
	(-3.3e + 02)	(-44.8120)
Temperature	-0.0056^***	-0.0018^***
	(-35.0348)	(-2.6612)
dew_temp	0.0120^***	0.0060^***
	(73.4644)	(7.4914)
Pressure	-0.0007^***	0.0031^***
	(-7.8570)	(6.8839)
Wind_Speed	-0.0130^***	0.0147^***
	(-29.9359)	(5.7928)
Wind_Direction	0.0015^***	0.0007^**
	(24.5114)	(2.3999)
Sky_Coverage	-0.0966^***	-0.0670^***
	(-45.1282)	(-6.7673)
N	12139	12139
Notes: , , ** indicates significance at the 10%, 5%, and 1% level; t statistics in parentheses.

3.2 Contributions from components of AQI

The results of the impact of AQI components on COVID-19 confirmed and death cases are shown in Table 3. As stated in the previous section, PM_2.5 lagged terms have VIF values higher than five, so they are not included in the ZIPR model. We find a significant positive impact from SO₂, NO₂, O₃, and CO on daily confirmed cases. All contributions from these variables are significant at a 1% level. We also find a significant positive impact from PM₁₀, NO₂, and CO on daily dead cases, where PM₁₀ is significant at a 10% level as well as NO₂ and CO significant at a 1% level. The complete regression result with the behavior of lagged term can be seen in Online Resource 1.

Table 3

ZIPR results of AQI components on COVID-19 cases
	(1)	(2)
	Confirmed_daily	Dead_daily
PM_2.5	-0.0038^***	-0.0159^***
	(-6.5158)	(-6.2882)
PM₁₀	-0.0077^***	0.0031^*
	(-12.8149)	(1.7498)
SO₂	0.0070^***	0.0027
	(4.0884)	(0.3167)
NO₂	0.0133^***	0.0160^***
	(24.2809)	(4.8740)
O₃	0.0041^***	0.0009
	(9.9005)	(0.3683)
CO	1.3575^***	1.0521^***
	(57.7328)	(5.8340)
Travel	-3.3678^***	-2.9605^***
	(-2.7e + 02)	(-34.8105)
Temperature	-0.0052^***	0.0010
	(-27.2648)	(1.0737)
Dew_temp	0.0122^***	0.0083^***
	(65.5812)	(7.8730)
Pressure	0.0042^***	0.0055^***
	(39.8692)	(9.8596)
Wind_Speed	-0.0121^***	0.0028
	(-25.6321)	(0.9405)
Wind_Direction	0.0007^***	0.0009^**
	(10.0906)	(2.4315)
Sky_Coverage	-0.1382^***	-0.0650^***
	(-54.1654)	(-5.4120)
Lagged terms for PM₁₀, SO₂, NO₂, O₃, and CO	Yes	Yes
N	12043	12043
Notes: , , ** indicates significance at the 10%, 5%, and 1% level. t statistics in parentheses.

COVID-19 has triggered the most significant public health crisis in recent years. This study examines the impact of air pollutants and air quality index (AQI) on daily COVID-19 confirmed cases and deaths in prefecture-level cities of China from 2020.01.22 to 2020.12.31. The epidemic data comes from AkShare, an open-source financial data interface library based on Python that summarizes the data interface of Baidu and NetEase. The zero-inflated Poisson regression is applied in this paper to explore the impact of air pollution levels on daily confirmed cases and daily dead cases. We find significant evidence for AQI to positively impact daily confirmed COVID-19 cases. By further decomposing AQI to its components, we find a significant positive impact from SO₂, NO₂, O₃, and CO on daily confirmed COVID-19 cases. All contributions from these variables are significant at a 1% level. We also find a significant positive impact from PM₁₀, NO₂, and CO on daily dead cases, where PM₁₀ is significant at a 10% level as well as NO₂ and CO significant at a 1% level. This study provides new insights into the relationship between air quality on daily COVID-19 confirmed cases and death cases in China and has direct implications for COVID-19 epidemic prevention and control.

Competing interests The authors declare no competing interests.

Author contribution Yifei Wang proposed the topic, conduct research, and write the paper. Mingshan Lu provide instructions for the whole paper. Meiling Jin help with literature summarization. Chunyan Zhu help with result discussion.

Funding The authors declare no funding was received for conducting this study.

Data availability The COVID-19 confirmed case and death data analyzed during the current study are available in the AKShare, https://www.akshare.xyz/. Daily air quality data are obtained from National real-time release platform for urban air quality, https://air.cnemc.cn:18007/. Weather condition data are obtained from NOAA, https://www.ngdc.noaa.gov/ftp.html. With-in city travel intensity data comes from Baidu Huiyan, https://qianxi.baidu.com/#/.

Analitis, A., Katsouyanni, K., Dimakopoulou, K., Samoli, E., Nikoloulopoulos, A. K., Petasakis, Y., Touloumi, G., Schwartz, J., Anderson, H. R., Cambra, K., et al. (2006). Short-term effects of ambient particles on cardiovascular and respiratory mortality. Epidemiology, 17(2):230–233. https://doi.org/10.1097/01.ede.0000199439.57655.6b
Andrée, B. P. J. (2020). Incidence of COVID-19 and connections with air pollution exposure: evidence from the Netherlands. World Bank Policy Research Working Paper, (9221). https://ssrn.com/abstract=3584842
Ciencewicki, J. and Jaspers, I. (2007). Air pollution and respiratory viral infection. Inhalation toxicology, 19(14):1135–1146. https://doi.org/10.1080/08958370701665434
Contini, D. and Costabile, F. (2020). Does air pollution influence COVID-19 outbreaks?. Atmosphere, 11(4), 377. https://doi.org/10.3390/atmos11040377
Cui, Y., Zhang, Z.-F., Froines, J., Zhao, J.,Wang, H., Yu, S.-Z., and Detels, R. (2003). Air pollution and case fatality of SARS in the People's Republic of China: an ecologic study. Environmental Health, 2(1):1–5. https://doi.org/10.1186/1476-069X-2-15
Deryugina, T., Heutel, G., Miller, N. H., Molitor, D., and Reif, J. (2019). The mortality and medical costs of air pollution: Evidence from changes in wind direction. American Economic Review, 109(12):4178–4219. https://doi.org/10.1257/aer.20180279
Fareed Z, Iqbal N, Shahzad F, Shah SGM, Zulfiqar B, Shahzad K, Hashmi SH, Shahzad U (2020). Co-variance nexus between COVID-19 mortality, humidity, and air quality index in Wuhan, China: new insights from partial and multiple wavelet coherence. Air Qual Atmos Health 13:673–682. https://doi.org/10.1007/s11869-020-00847-1
Frontera, A., Martin, C., Vlachos, K., and Sgubin, G. (2020). Regional air pollution persistence links to COVID-19 infection zoning. Journal of infection, 81(2):318–356. https://doi.org/10.1016/j.jinf.2020.03.045
Islam, S., Alam, G., Ibrahim, S. M., et al. (2021). Effect of meteorological factors on COVID-19 cases in Bangladesh. Environment, Development and Sustainability, 23(6):9139–9162. https://doi.org/10.1007/s10668-020-01016-1
Jiang, Y. and Xu, J. (2021). The association between COVID-19 deaths and short-term ambient air pollution/meteorological condition exposure: a retrospective study from Wuhan, China. Air Quality, Atmosphere & Health, 14(1):1–5. https://doi.org/10.1007/s11869-020-00906-7
Martelletti, L. and Martelletti, P. (2020). Air pollution and the novel COVID-19 disease: a putative disease risk factor. SN comprehensive clinical medicine, 2(4):383–387. https://doi.org/10.1007/s42399-020-00274-4
Omer, T., Sjölander, P., Månsson, K., and Kibria, B. G. (2021). Improved estimators for the zero-inflated Poisson regression model in the presence of multicollinearity: simulation and application of maternal death data. Communications in Statistics: Case Studies, Data Analysis and Applications, 7(3):394–412. https://doi.org/10.1080/23737484.2021.1952493
Pansini, R. and Fornacca, D. (2020). Initial evidence of higher morbidity and mortality due to SARS-CoV-2 in regions with lower air quality MedRxiv. http://conacyt.repositorioinstitucional.mx/jspui/handle/1000/2260
Park S, Allen R, Lim C (2020) A likely increase in fine particulate matter and premature mortality under future climate change. Air Qual Atmos Health 13:143–151. https://doi.org/10.1007/s11869-019-00785-7
Persico, C. and Johnson, K. R. (2020). Deregulation in a time of pandemic: Does pollution increase coronavirus cases or deaths?. IZA Discussion Paper No. 13231. http://dx.doi.org/10.2139/ssrn.3596698
Vali, M., Hassanzadeh, J., Mirahmadizadeh, A., Hoseini, M., Dehghani, S., Maleki, Z., Méndez-Arriaga, F., and Ghaem, H. (2021). Effect of meteorological factors and air quality index on the COVID-19 epidemiological characteristics: an ecological study among 210 countries. Environmental Science and Pollution Research, 28(38):53116–53126. https://doi.org/10.1007/s11356-021-14322-6
Wilson, P. 2015. The misuse of the Vuong test for non-nested models to test for zero-inflation. Economics Letters 127: 51–53. https://doi.org/10.1016/j.econlet.2014.12.029
Xie, J. and Zhu, Y. (2020). Association between ambient temperature and COVID-19 infection in 122 cities from China. Science of the Total Environment, 724:138201. https://doi.org/10.1016/j.scitotenv.2020.138201
Yousefi, M., Oskoei, V., Jonidi Jafari, A., Farzadkia, M., Hasham Firooz, M., Abdollahinejad, B., and Torkashvand, J. (2021). Municipal solid waste management during COVID-19 pandemic: effects and repercussions. Environmental Science and Pollution Research, 28(25):32200–32209. https://doi.org/10.1007/s11356-021-14214-9
Yu, I. T., Li, Y., Wong, T. W., Tam, W., Chan, A. T., Lee, J. H., Leung, D. Y., and Ho, T. (2004). Evidence of airborne transmission of the severe acute respiratory syndrome virus. New England Journal of Medicine, 350(17):1731–1739. https://doi.org/10.1056/NEJMoa032867
Zhu, Y., Xie, J., Huang, F., and Cao, L. (2020). Association between short-term exposure to air pollution and COVID-19 infection: Evidence from China. Science of the total environment, 727:138704. https://doi.org/10.1016/j.scitotenv.2020.138704

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Impact of Air Quality on COVID-19 Confirmed and Death Cases: Evidence from China

Status:

Version 1

Abstract

Figures

1. Introduction

2. Materials And Methods

2.1 Data

2.1.1 Sample construction

2.1.2 Confirm and death cases

2.1.3 AQI

2.1.4 Covariates

2.2 Method

3. Results And Discussion

3.1 Contribution from AQI

3.2 Contributions from components of AQI

4. Conclusions

Declarations

References

Additional Declarations

Status:

Version 1