Table 1 Overview of selected publications studying associations between air quality and COVID-19 statistics
Study | Approach | Result | Area | Time |
|---|---|---|---|---|
Ogen [61] | Categorized NO2 measurements were compared | The results indicated a strong association between high values of the pollutant and high fatality cases | 66 administrative regions in Italy, Spain, France, and Germany | January to February 2020 |
Bashir et al. [62] | The individual correlation between risk factors and new infections, total infections, and mortality were measured on a daily basis. Kendall and Spearman rank correlation was calculated. It is not clear what measurement was used to determine air quality | Besides temperature, air quality was significantly correlated with the COVID-19 metrics | New York City, USA | March to April 2020 |
Accarino et al. [63] | The Spearman correlation between PM2.5, PM10, NO2 and COVID-19 incidence rate as well as mortality rate was measured | Significant associations between all of them were found | 107 Italian territorial areas | February and March 2020 |
Zhu et al. [64] | Daily infections, meteorological variables, and air pollution concentrations for PM2.5, PM10, SO2, CO, NO2, and O3 were collected. Generalized additive models were used to estimate the associations between lagged, moving average concentrations of air pollutants and daily infections | Significant positive associations for PM2.5, PM10, CO, NO2, and O3 and a negative association for SO2 were shown | 120 Chinese cities | January to February 2020 |
Stieb et al. [41] | A negative binomial model was used to measure the association between PM2.5 from 2000 to 2016 and infection count. The Akaike information criterion was used to some extent to select from the socio-demographic, health, time since peak incidence, and temperature variables | The multivariate model did not show a significant association for PM2.5 | 111 Canadian regions | Up to May 13, 2020 |
Wu et al. [65] | Negative binomial mixed models were used to regress on the mortality rate with PM2.5 and 20 other confounders as predictors. The particulate matter between 2000 and 2016 was considered | A notable association was found for PM2.5, population density, days since first reported case, household income, percent of owner-occupied housing, high school education, age, and percent of Black residents | 3089 US counties | Up to June 18, 2020 |
Rodriguez-Villamizar et al. [42] | A negative binomial hurdle model was used to analyze the effect of PM2.5 measured between 2014 and 2018 on COVID-19 mortality including socio-demographic, socio-economic and health confounders | PM2.5 did not show a significant association with mortality | 772 Colombian municipalities | Up to July 17, 2020 |
Adhikari et al. [43] | A negative binomial regression was applied on time-series data. Besides daily PM2.5 and ozone, meteorological confounders were included | Ozone was found to be significantly associated with the daily infections but not with deaths | Queens county, New York, USA | March to April 2020 |
Borro et al. [66] | Simple linear regressions were performed for cumulative COVID-19 incidence, mortality rate, and case-fatality rate with PM2.5 as predictor | Significant associations were found for all three metrics | 110 Italian provinces | February to March 2020 |
Travaglio et al. [44] | Negative binomial models were used to measure the association between PM2.5, PM10, NO, NO2, O3 and COVID-19 cases as well as deaths. Population density, average age, and mean earning were included as confounders. Air quality data prior to the pandemic were aggregated over one and five years | Both COVID-19 metrics showed significant associations with the air quality risk factors | England on regional and sub-regional level | February to May 2020 |
Tieskens et al. [67] | The incidence of five distinct time periods was analyzed via mixed-effect Poisson regression. Besides PM2.5, also 19 other socio-demographic, occupational, and mobility variables were incorporated. The variables were selected by excluding covariates with a variance inflation factor higher 2.5 in the regression of the first time period | PM2.5 was not selected, yet almost all selected socio-demographic and economic variables indicated strong variance of their association between the time periods | 351 cities in Massachusetts, USA | March to October 2020 |
Liang et al. [45] | Zero-inflated negative binomial models were used to determine the association between NO2, PM2.5, and O3 and case-fatality and mortality rates. Air quality measurements between 2010 and 2016 were considered. The models also included socio-demographic, socio-economic, health, and mobility variables | For NO2, a positive association with the COVID-19 metrics was found | 3122 US counties | January to July 2020 |