COVID-19 Fatality: Statistical Evidence to Engender the Need for Focal Shift from Air Pollutants to Multi-dimensional Intervention

Since the outbreak of COVID-19 pandemic, several researchers worldwide have suggested chronic exposure to air pollutants viz. PM2.5, NO2, and O3 as one of the influencing factors for the increased rate of fatality. However, most of these studies lacked a comprehensive international outlook. A strong correlation on a regional scale might require further investigations to evaluate the transboundary validity. Therefore, the current study aims to explore the statistical soundness of association of COVID-19 fatality with PM2.5, NO2, and O3 concentration levels across 463 air quality monitoring stations located in 35 selected cities from USA, India and European Regions (France, Germany and Italy). An aggregated open-source air quality data source was used to download the PM2.5, NO2, and O3 concentration for > 900 days in the selected cities. The median of this long-term exposure was tested against the COVID-19 fatality rates. A strength of association parameter, ω2, and the coefficient of determination, R2, were used to evaluate the transboundary association. ω2 results indicated that only 24.6%, 0.03% and 15.4% of the variation in COVID-19 fatality rates could be explained using PM2.5, NO2, and O3 concentrations respectively for all the analyzed cities. Further, low values of R2 between pollutant concentrations and COVID-19 fatality rates corroborated the results (0.27 for PM2.5, 0.00038 for NO2 and 0.18 for O3). These observations strongly suggest a focal shift towards the inclusion of more explanatory variables and an extensive multi-disciplinary work is required in order to understand the cause of COVID-19 fatality.


Corona virus disease 2019 (COVID-19) caused by Severe Acute Respiratory Syndrome Corona
Virus 2 (SARS-CoV-2) was first identified in Wuhan city of China in December 2019 (Chen et al., 2020) and later became a pandemic. Since its outbreak, the contagious infection spread across 216 countries, affecting more than 10 million people and claiming over 555,642 lives worldwide as on 11 th July 2020 (https://covid19.who.int/). The course of COVID-19 infection gradually starts with mild flu-like symptoms or no symptoms and further progress to infect the respiratory system of the infected individual. In certain severe cases, however, it might lead to death due to Acute Respiratory Distress Syndrome (ARDS) and pneumonia (Li and Ma, 2020). The widespread magnitude of this catastrophe has, therefore, made it critical for people to understand various aspects of COVID-19, its spread and prevention as well as its impact on health, economy, etc. Since, COVID-19 is primarily a respiratory syndrome, and several epidemiological studies suggests that chronic exposure to air pollutants viz. PM2.5, NO2 and O3 can increase the risk for cardio respiratory mortality and morbidity (Russell and Brunekreef, 2009;Pope, 2000;Sicard et al., 2019), recently, some studies tried to statistically explore whether pollutant exposure can influence the likelihood of COVID-19 fatality.
In this context, Frontera et al. (2020) reported a correlation of 0.53 between mean PM2.5 and COVID-19 deaths in worst COVID-19 affected regions in Italy. Similarly, Ogen, 2020 has also reported higher correlation between long-term exposure to NO2 and COVID-19 fatality in Italy and Spain. Along the same lines, Wu et al. (2020), Pansini and Fornacca (2020), and Yao et al. (2020) reported a strong association between COVID-19 fatality rate and PM concentrations for USA and China. Table 1 summarizes some of the previously published study relating the fatality rate of COVID-19 with chronic exposure to air pollutants. These studies primarily examined the significant association of the geo-environmental determinants with COVID-19 fatality cases on a regional scale or within a country. However, a strong correlation on a regional scale might require further investigations to evaluate its transboundary validity. Therefore, in this pursuit, the primary objective of this study is to assess the statistical soundness of association of chronic air pollutant exposure viz. NO2, O3, PM2.5 with COVID-19 fatality rate over a broader perspective. The statistical analysis was performed for 35 different cities located across five different countries (USA, India, France, Germany and Italy). The corresponding data collection and its processing are discussed in the methodology, and its results are discussed in the results and discussion section.

Data Collection and Processing
The hourly surface NO2, O3, and PM2.5 concentration levels were obtained for 463 air quality It has an archive of air quality data collected from more than 12,000 air qualitymonitoring stations situated in 93 countries across the globe. There are reports that researchers have previously used this platform to understand the global pollutant patterns (Manning et al., 2018). Fig. 1 shows the number of monitoring stations analysed from each of the selected city. The selection of pollutant parameters (PM2.5, NO2, and O3) was inspired from previous studies, which revealed that chronic exposure to these pollutants might lead to increase in respiratory stress (Pope, 2000;Sicard et al., 2019). Furthermore, the choice of cities and countries are based on availability of ground-based PM2.5, NO2 and O3 data, and the high COVID-19 infection rates. Since the effectiveness of pollutants on fatality depends upon the exposure time, we have considered median values of concentrations calculated from the collected data (2018)(2019)(2020). Also, as the OpenAQ website does not provide explicit information on errors in measurement, prior to the calculation of median, data were filtered for the outliers. The outliers were detected in a moving window of 10 data points where anything above Eq. (1) and below Eq. (2) were removed from the dataset (Benkert et al., 2008).
Third Quartile + 1.5 × Inter Quartile range (1) First Quartile -1.5 × Inter Quartile range (2) Record of number of COVID-19 confirmed and deceased cases were obtained from different credible government and local websites for each selected city until 1 st July 2020. Globally, almost 74,630 COVID-19 death cases were accounted in this analysis. The city wise fatality rate was determined using the Eq. (3).

Number of deceased COVID-19 cases in the city COVID-19 Fatality rate
Number of confirmed COVID-19 cases in the city (in 100 s) = (3)

Statistical Analysis
The goal of the current study is to explore the association of COVID-19 fatality rate relative to long-term exposure of NO2, O3, and PM2.5 concentrations for the selected cities. To quantify this association, a strength of association statistic, ω 2 , is used. Omega squared, ω 2 , is a descriptive statistic that determines the strength of association between an explanatory variable and a response variable (Stephanie, 2016). A detailed methodology to calculate ω 2 is explained elsewhere (Salkind, 2010).
The values of ω 2 lies between 0 and 1, where 0 indicates no association, and 1 indicates that all the variation in the response variable can be explained using the explanatory variable. Here, the processed pollutant (NO2, O3, and PM2.5) concentrations were considered as the explanatory variable, and the COVID-19 fatality rate formed the response variable. The present study basically aims to investigate what amount of variability of COVID-19 fatalities that could be explained by the pollutants (NO2, O3, and PM2.5), but not to establish the sign of the correlation or the significance of correlation. Therefore, such a statistic, ω 2 , was selected. Commenting on sign of correlation and its significance would require epidemiological interventions rather than only statistical analysis (Bontempi et al., 2020). However, to bolster the ω 2 results, a common association parameter, coefficient of determination, R 2 , was also calculated for each pollutant and COVID-19 fatality rates.

RESULTS AND DISCUSSION
The median PM2.5, NO2 and O3 concentrations of all the 35 selected cities along with their COVID-19 fatality rates is listed and presented in the supplement section (Table S1). To understand the pattern of median PM2.5, NO2 and O3 concentrations with COVID-19 fatality rate, scatter plots between them were drawn for the 35 selected cities (Fig. 2). Fig. 2 depicts no clear association between concentrations of PM2.5, NO2 and O3, and the fatality rate of COVID-19. Cities in India experiencing higher PM2.5 concentrations are marked with low fatality rates, whereas the distribution of fatality rates in low PM polluted cities is arbitrary. Furthermore, the data points on O3 and NO2 plots indicate that their concentration is widely spread, following no apparent trend, with respect to COVID-19 fatalities.
To understand the amount of variability in COVID-19 fatality rates that can be explained using pollutant concentration, ω 2 and R 2 values were calculated. The results of statistical analysis are summarized in Table 2. In the analyzed cities, ω 2 values indicate that PM2.5, NO2, and O3 could explain only 24.6 %, 0.03 % and 15.4 % of variation in COVID-19 fatality rates, respectively. In addition, the observed R 2 values were 0.27, 0.00038 and 0.18 for PM2.5, NO2, and O3, respectively. Such lower values of ω 2 and R 2 shows that the pollutant concentration data is noisy relative to COVID-19 fatality rate, and not statistically sufficient to explain the variability in the data. This is also reflected in higher values of sum of squares of residuals. Therefore, the results indicate that when the association between the COVID-19 fatality rates and long-term pollutant concentrations were explored over different worst-affected COVID-19 countries as a whole, they could explain only a fraction of COVID-19 fatality variation. This suggests that COVID-19 fatality might be influenced by a number of other co-factors, like age, co-morbidities, cross-immunities etc. (Chakrabarti et al., 2020). Concentration of air pollutants might have a secondary effect on some of these variables. However, to completely understand the COVID-19 fatality function, cumulative impact of all the potentially influencing other variables should be considered. This engenders the need for a multidimensional and multi-disciplinary investigation to better comprehend the fatality levels of COVID-19 globally.

CONCLUSIONS
This cohort study evaluated the association of PM2.5, NO2 and O3 concentration with COVID-19 fatality rate across 35 different cities located in 5 different countries. For the same, ground level pollutant concentrations (PM2.5, NO2 and O3) for > 900 days were downloaded from OpenAQ website and processed for the present investigation. A strength of association statistic, ω 2 , and the coefficient of determination, R 2 , were used to assess the relationship. Upon analysis, the ω 2