Association of particulate matter pollution and case fatality rate of COVID-19 in 49 Chinese cities


 Abstract
 
 The COVID-19 epidemic, caused by the SARS-CoV-2 virus, has resulted in 3352 deaths in China as of April 12, 2020. This study aimed to investigate the associations between particulate matter (PM) concentrations and the case fatality rate (CFR) of COVID-19 in 49 Chinese cities, including the epicenter of Wuhan. We used the Global Moran's I to analyze spatial distribution and autocorrelation of CFRs, and then we used multivariate linear regression to analyze the associations between PM2.5 and PM10 concentrations and COVID-19 CFR. We found positive associations between PM pollution and COVID-19 CFR in cities both inside and outside Hubei Province. For every 10 μg/m3 increase in PM2.5 and PM10 concentrations, the COVID-19 CFR increased by 0.24% (0.01%–0.48%) and 0.26% (0.00%–0.51%), respectively. PM pollution distribution and its association with COVID-19 CFR suggests that exposure to such may affect COVID-19 prognosis.
 



Introduction
SARS-CoV-2 is a newly emerged coronavirus that has posed immense challenges to global health and has caused unpredictable economic loss. First reported in December of 2019 in Wuhan, China, this novel coronavirus (COVID-19) epidemic exhibits human-to-human transmissibility and has spread rapidly across countries (Li et al., 2020). As of April 12, 2020, a total of 1,610,909 COVID-19 cases and 99,690 deaths have been confirmed in 211 countries, with China reporting a total of 82,214 J o u r n a l P r e -p r o o f 3 (Conticini et al., 2020). Exposure to ambient particulate matter (PM) pollution has been reported to increase the risks of mortality and morbidity from cardiopulmonary diseases worldwide (Atkinson et al., 2014;Chen et al., 2017;Samet et al., 2000;Samoli et al., 2005), with higher likelihood of adverse effects among elderly or people with underlying medical conditions (Zeka et al., 2006); elderly and people with underlying medical conditions also experienced higher fatality from COVID-19 Pan et al., 2020). Toxicological evidence showed that PM can cause pulmonary inflammation and affect the defense system against infection (Donaldson et al., 2001). In addition, exposure to PM could increase inflammation and oxide stress, aggravating respiratory symptoms resulting in increased hospital emergency visits of patients with asthma and chronic obstructive pulmonary disease (COPD) (EPA, 2020). Moreover, air pollution, especially PM pollution, is positively associated with case fatality from other coronavirus infection including severe acute respiratory syndrome (SARS) (Cui et al., 2003). Other studies have proposed that PM could carry viruses as carrier and spread viruses everywhere as a vector (Alonso et al., 2015;Yang et al., 2011).
To date, few studies have estimated and quantified the effects of air pollutants on case fatality rate (CFR) from COVID-19. Therefore, this study aims to investigate the associations between PM 2.5 and PM 10 and CFR of COVID-19 in Chinese cities.

Statistical analysis
Spatial auto-correlation statistics have been commonly used to examine spatial dependence or auto-correlation in spatial data. Spatial auto-correlation includes 1) the global spatial auto-correlation which is used for estimating the overall degree of spatial auto-correlation for spatial data, and 2) the local indicators of spatial association (LISA) which is used to assess the influence of individual locations on the magnitude of the global statistic and to identify the locations and types of clusters.
The spatial weights were created by rook contiguity rule, and applied to describe the spatial relationships among cities. We explored the spatial distribution of CFRs from 49 cities in China by calculating the Global Moran's I and LISA using ArcToolbox of ArcMap (version 10.2). The calculation formula of Global Moran's I is shown as following: LISA is computed as follows: J o u r n a l P r e -p r o o f 6 COVID-2019 CFRs of all cities, and W ij is the spatial weight matrix corresponding to the cities pair i and j. Global Moran's I index is between -1.0 and 1.0. Moran's I >0 represents positive spatial correlation, and the greater the value, the more obvious the spatial correlation. Moran's I <0 represents negative spatial correlation, and the smaller the value, the greater the spatial difference. Otherwise, Moran's I = 0, the space is in random mode. The positively spatial correlation means that the correlation becomes more and more significant with the location (distance) aggregation, and vice versa.
The risk estimates increased to 0.61% (0.09% -1.12%) and 0.33% (0.03% -0.64%) J o u r n a l P r e -p r o o f 10 The potential mechanism for PM exposure on respiratory outcomes might be the activation of inflammatory pathways in the small respiratory airways in response to PMs, leading to the recruitment of inflammatory cells (Kelly and Fussell, 2011).
These biological mechanisms might potentially influence the prognosis of COVID-19 patients.
In this study, PM-fatality associations are significantly positive no matter if using average PM levels during epidemics period or in 2015-2019. Considering that the patients died from COVID-19 are likely stay in hospital for treatment without being directly exposed to ambient air pollution prior to death. Therefore, we speculate that the effects of PM 2.5 and PM 10 on death mainly affect the progress of patients from mild to severe and prognosis, when the patients were not isolated from ambient air pollution. Additionally, our results showed positive association between long-term PM exposure and COVID-19 CFR, suggesting that long term PM exposure that prior to the epidemics period could have increased vulnerability of population to SARS-CoV-2. There is still a need to increase our efforts in the control of air pollutant emissions, critical for potential resurgence of COVID-19 epidemics in the future.
This ecological association study is limited by the duration of study period.
J o u r n a l P r e -p r o o f 11 per capita to ensure the credibility of the results. Also, the risk estimates stayed stable after adding other covariates including local LISA map values, city size and population or proportion of people older than 65 years. Previous studies reported that some chronic diseases, such as hypertension, diabetes and cardiopulmonary diseases were also potentially linked to the pathogenesis of COVID-19 (Yang et al., 2020); thus the co-morbidities of these diseases may serve as confounders to associations of PM exposure and COVID-19. The effects of prevalence of these diseases should be investigated further when data available. Limited by the resolution of COVID-19 data, which was only available at city level, exposure assessments of PM 2.5 and PM 10 in this study were done at city level based on ground monitoring data rather than using gridded predicting data of PM 2.5 and PM 10 at high spatial resolution (e.g. 1km×1km); using ground monitoring with limited spatial resolution might cause exposure misclassification. Nevertheless, we believe that this exposure misclassification would not substantially bias our findings since: 1) the ground monitoring sites were normally located at areas of high population density in China that could reflect the exposure levels of the majority of residents; 2) and this kind of nondifferential misclassification J o u r n a l P r e -p r o o f    Figure 3A: CFR was positively associated (Meta χ 2 =15.25, p=0.0042) with PM 2.5 in cities outside Hubei (blue points, r=0.56, p=0.005) and those inside Hubei except Wuhan (green points, r=0.33, p=0.26) pollution. GDP per capita and hospital beds per capita effects were removed during statistical analysis. Figure 3B: CFR was positively associated (Meta χ 2 =13.53, p=0.0090) with PM 10 in cities outside Hubei (blue points, r=0.48, p=0.019) and those inside Hubei except Wuhan (green points, r=0.47, p=0.11) pollution. GDP per capita and hospital beds per capita effects were removed during statistical analysis. Average PM concentrations were calculated from January 15, 2020 to February 29, 2020 to represent the mean PM levels during the main period of