Abstract
Background
Early ecological studies have suggested links between air pollution and risk of coronavirus disease 2019 (COVID-19), but evidence from individual-level cohort studies is still sparse. We examined whether long-term exposure to air pollution is associated with risk of COVID-19 and who is most susceptible.
Methods
We followed 3 721 810 Danish residents aged ≥30 years on 1 March 2020 in the National COVID-19 Surveillance System until the date of first positive test (incidence), COVID-19 hospitalisation or death until 26 April 2021. We estimated residential annual mean particulate matter with diameter ≤2.5 μm (PM2.5), nitrogen dioxide (NO2), black carbon (BC) and ozone (O3) in 2019 by the Danish DEHM/UBM model, and used Cox proportional hazards regression models to estimate the associations of air pollutants with COVID-19 outcomes, adjusting for age, sex, individual- and area-level socioeconomic status, and population density.
Results
138 742 individuals were infected, 11 270 were hospitalised and 2557 died from COVID-19 during 14 months. We detected associations of PM2.5 (per 0.53 μg·m−3) and NO2 (per 3.59 μg·m−3) with COVID-19 incidence (hazard ratio (HR) 1.10 (95% CI 1.05–1.14) and HR 1.18 (95% CI 1.14–1.23), respectively), hospitalisations (HR 1.09 (95% CI 1.01–1.17) and HR 1.19 (95% CI 1.12–1.27), respectively) and death (HR 1.23 (95% CI 1.04–1.44) and HR 1.18 (95% CI 1.03–1.34), respectively), which were strongest in the lowest socioeconomic groups and among patients with chronic respiratory, cardiometabolic and neurodegenerative diseases. We found positive associations with BC and negative associations with O3.
Conclusion
Long-term exposure to air pollution may contribute to increased risk of contracting severe acute respiratory syndrome coronavirus 2 infection as well as developing severe COVID-19 disease requiring hospitalisation or resulting in death.
Tweetable abstract
Long-term exposure to air pollution may contribute to increased risk of contracting SARS-CoV-2 infection as well as developing severe COVID-19 disease requiring hospitalisation or resulting in death https://bit.ly/3O2vrJI
Introduction
Identifying key modifiable factors that could contribute to increased risk of contracting severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and developing coronavirus disease 2019 (COVID-19) or worsen the severity of health outcomes among individuals with COVID-19 is crucial for the management of the current COVID-19 pandemic and preparing for future similar pandemics. Air pollution is the major environmental stressor and fourth global risk factor for morbidity and mortality, after smoking, high blood pressure and diet, causing 6.7 million deaths in 2019 [1]. Several biologically plausible mechanisms explain how air pollution may increase the risk of COVID-19. Long-term exposure to air pollution may increase vulnerability to SARS-CoV-2 indirectly, by increasing the risk of respiratory and cardiometabolic diseases [1], which increase the risk of severe COVID-19 [2–5], or by directly compromising immune responses [6]. The link between air pollution and COVID-19 is further supported by air pollution links with other respiratory infections including pneumonia [7–10].
Early ecological studies [11] raised headlines postulating that air pollution increases the risk of COVID-19, despite the potential fallacies of this design [12]. Epidemiological evidence based on cohort studies with individual-level data on exposure to air pollution and later onset of COVID-19 is still sparse and mixed, yet crucial in understanding the impact of air pollution on contracting SARS-CoV-2 and developing COVID-19 disease. Only three studies examined the incidence of SARS-CoV-2 infection in the general population. A study from Varese (Italy) detected an association of long-term exposure to particulate matter with diameter ≤2.5 μm (PM2.5) with COVID-19 incidence [2], a study from Rome (Italy) found no association of PM2.5 or nitrogen dioxide (NO2) with incidence, but detected associations with COVID-19 mortality [13], and a study in UK Biobank detected associations of PM2.5 and NO2 with incidence, but none with COVID-19 hospitalisations or death [14]. In addition, a study from Barcelona (Spain) found no association of PM2.5 or NO2 with a SARS-CoV-2-positive serum test, but detected associations with self-reported COVID-19, which were strongest with severe outcomes (hospitalisation or death) [15]. Furthermore, nine studies with data on SARS-CoV-2 cases only [16–24] reported associations between PM2.5 and increased risk of severe COVID-19 outcome in terms of hospitalisations [16–18, 20–22], intensive care unit admission [17, 19, 20] or death [17, 19, 20, 23]. One study showed associations between PM2.5 and increased risk of dying in a population of patients hospitalised with COVID-19 [24].
In this large nationwide study from Denmark, we examined whether long-term exposure to air pollution is associated with risk of SARS-CoV-2 infection, hospitalisation and death, and identified those who were most susceptible by sex, age, socioeconomic status (SES) and comorbidities.
Methods
Study population and COVID-19 outcome definitions
We created a population-based nationwide cohort (AIRCODEN) by including all Danish residents aged ≥30 years on 1 March 2020 and who had lived in Denmark for at least 1 year prior to this date. Using the unique personal identification number, we linked AIRCODEN to the National COVID-19 Surveillance System, with information on SARS-CoV-2 PCR test date, test result, hospital admission date and death date. PCR testing, and in the second pandemic wave after 1 August 2020 lateral flow testing, in Denmark was offered to all citizens free of charge with easy access and self-booking opportunities. A positive lateral flow test had to be confirmed by PCR in order to be registered in the National COVID-19 Surveillance System, so only PCR-confirmed COVID-19 cases are considered in this study. We defined three COVID-19 outcomes, as defined by the National COVID-19 Surveillance System: incidence (first positive test), hospitalisation (inpatient admission, including emergency room admission from any cause, for >12 h within 14 days after the first positive test) and death from any cause (within 30 days of the first positive test). We also evaluated all-cause and non-COVID-19 mortality, to examine whether air pollution–all-cause mortality association during the pandemic period is comparable to that observed before the pandemic [20]. We defined two pandemic waves reflecting two distinct periods, with respect to testing capacity and preventive (lockdown) measures: the first pandemic wave (1 March 2020 to 1 August 2020) reflects the period with very limited testing capacity and full lockdown, and the second wave (starting on 1 August 2020) reflects the period with full opening of society along with significant improvement of testing capacity and opening of testing centres. We extracted individual-level SES information in 2019, including employment, education, income, wealth, marital status and household size, from socioeconomic registers at Denmark Statistics, and defined parish- and municipality-level SES. There are 2163 parishes and 96 municipalities in Denmark. In addition, there are five administrative regions in Denmark, which run secondary healthcare systems (e.g. set strategies for handling COVID-19 in the hospitals, hospital capacity, etc.) and COVID-19 testing strategy and testing capacity. Thus, in our current analyses, we have adjusted for regions in COVID-19 analyses, to account for different COVID-19 strategies in the five regions, but not in all-cause mortality analyses, which is not affected by regional handling of secondary healthcare, and consistent with our previous analyses on air pollution and all-cause mortality [25]. Comorbidities were defined as any hospital contact (inpatient admission, emergency room admission or outpatient visit) in the Danish National Patient Register prior to 1 March 2020 (supplementary table S1).
Air pollution exposure
We used the Danish DEHM/UBM model estimates of annual means of NO2, PM2.5, particulate matter with diameter ≤10 μm (PM10), BC and O3 at 1×1 km resolution in the period 1979–2019. This validated model consists of the chemistry transport model, the Danish Eulerian Hemispheric Model (DEHM) and the Urban Background Model (UBM), including several domains with different spatial resolutions to calculate intercontinental, regional and local transport of air pollution [26–29]. The models were successfully validated against measurements. In addition, we used the European-wide hybrid land use regression (LUR) model, developed within the Effects of Low-Level Air Pollution: A Study in Europe (ELAPSE) project, which provides annual means of PM2.5, NO2, BC and O3 (warm season) in 2010 at 100×100 m resolution [30], recently linked to all-cause mortality in Europe [31, 32] and Denmark [25]. In brief, the LUR model utilised routine monitoring data from the European Environment Agency AirBase for PM2.5, NO2 and O3, and the European Study of Cohorts for Air Pollution Effects (ESCAPE) monitoring data for BC. Satellite data, chemistry transport model estimates, land use and traffic variables were predictors to estimate annual mean pollutant concentrations. The models performed well in five-fold hold-out validation [30].
Statistical analysis
We used Cox proportional hazards models (for time-to-event data) with calendar time as the underlying timescale to examine the associations of air pollutants with COVID-19 incidence, hospitalisation, death or all-cause mortality (separately), censoring at the date of death from other causes (except for all-cause mortality where all deaths are outcome), emigration or the end of follow-up on 26 April 2021, whichever came first. We fitted four models: Model 1: adjusted for age (strata, 5-year band), sex (strata) and region (strata, five administrative regions in Denmark; not included in all-cause mortality analyses); Model 2: additionally adjusted for individual-level SES, including marital status, highest completed education, occupational status, individual wealth (tertile), family income (tertile) and household size; Model 3 (main model): additionally adjusted for population-level SES including parish-level population density, municipality-level access to healthcare (the number of general practitioners/citizens), parish-level SES factors (mean income, median wealth, unemployment rate, primary or low education rate), and the SES difference between municipality and parish; and Model 4: additionally adjusted for monthly municipality-level SARS-CoV-2-positive rate (PCR tests), using time-varying Cox models, in order to adjust for spatiotemporal COVID-19 pandemic development.
We fitted single-pollutant models using mean exposure in 2019 (main analyses), 3-year (2017–2019) and 10-year (2010–2019) means of pollutants estimated by the Danish DEHM/UBM model, and 2010 mean exposure estimated by the ELAPSE model. We estimated exposure–response functions using natural cubic splines with three degrees of freedom. We estimated associations in a subsample of the population below predefined levels of pollutants, to examine associations at the lower end of exposure. Effect modification of an association of PM2.5 and NO2 with COVID-19 by sex, age, individual-level SES, ethnicity and comorbidities with cardiovascular disease (CVD), respiratory disease, acute lower respiratory infections (ALRIs), diabetes, lung cancer, dementia and diabetes was evaluated by entering interaction terms into the model and tested by the Wald test, a multiplicative scale test for the difference between the subgroups of subjects defined by these effect modifiers. We fitted two-pollutant models for pollutant combinations with a Pearson correlation coefficient <0.7.
We performed several sensitivity analyses. To evaluate confounding by missing information on smoking and body mass index (BMI), we additionally adjusted for parish-level prevalence rates of COPD or lung cancer (proxies for smoking) and diabetes (obesity proxy), and applied the Shin indirect adjustment method [33], using the associations of air pollution with smoking and BMI based on the Danish National Survey. We estimated associations in a subsample of the population tested for SARS-CoV-2, to explore whether associations with air pollution are affected by a selection bias in who gets tested. We estimated associations in a subsample of the population, as this is a selected sample, and not directly comparable to the entire population. In addition to all-cause mortality, we estimated associations of air pollutants with non-COVID-19 all-cause mortality. We examined the stability of the associations over time and whether they were affected by changes in testing capacity and stringency of and compliance with pandemic measures (lockdown, wearing mask, distancing, etc.) by estimating associations separately in two pandemic waves (1 March 2020 to 31 July 2020 and 1 August 2020 to 26 April 2021).
Subjects with complete information for Model 3 variables were included in the analyses. We conducted analyses using R version 4.1.2 (www.r-project.org) and presented hazard ratios (HRs) and 95% confidence intervals per interquartile range increase in pollutant.
Results
Of the 3 743 013 subjects aged ≥30 years and Danish resident on 1 March 2020, we excluded 8397 with missing information on air pollution, 676 on individual SES, 12 127 on parish SES and three with a positive SARS-CoV-2 test before 1 March 2020, leaving 3 721 810 subjects for the final analyses. During 14 months of follow-up (411–417 days) and two pandemic waves (figure 1), 138 742 individuals tested positive for SARS-CoV-2, 11 270 were hospitalised and 2557 died from COVID-19, whereas 62 359 individuals died in total. Compared with the total population, subjects who died or were hospitalised with COVID-19, or died from any cause, were less likely to be women, highly educated, employed, married or have a high income (table 1). Similar patterns, but with less pronounced differences, were observed with incident COVID-19 cases. The strongest positive correlation was observed between NO2 and BC (0.75), reflecting the same source (traffic), followed by NO2 and PM2.5 (0.61) (supplementary figure S1). The strongest negative correlation was observed between NO2 and O3 (−0.86) (supplementary figure S1). Traffic-related NO2 and BC are highest in urban areas of Denmark, whereas O3 is highest in rural areas. PM2.5 in Denmark is, apart from urban areas, also high in Southeastern Denmark due to long-range transported particles from Central/Eastern Europe (supplementary figure S2).
FIGURE 1.
The COVID-19 pandemic in Denmark between 1 March 2020 and 26 April 2021: daily numbers of COVID-19-positive cases, hospital admissions and deaths in Denmark.
TABLE 1.
Characteristics of the 3 721 813 participants of the AIRCODEN cohort at the study baseline on 1 March 2020
| Total population | COVID-19 incidence | COVID-19 hospitalisation | COVID-19 mortality | All-cause mortality | |
| Participants | 3 721 813 | 138 742 | 11 270 | 2557 | 62 359 |
| Person-days at risk | 1 531 385 032 | 1 549 500 678 | 1 551 066 930 | 1 551 066 930 | |
| Follow-up time, days | 411 | 416 | 417 | 417 | |
| Individual level | |||||
| Age, years | 56.3±15.6 | 52.0±14.6 | 68.1±15.6 | 81.1±10.3 | 78.6±12.2 |
| Age >65 years | 1 157 323 (31.1) | 25 207 (18.2) | 6859 (60.9) | 2383 (93.2) | 53 873 (86.4) |
| Female | 1 904 171 (51.2) | 72 250 (52.1) | 5024 (44.6) | 1162 (45.4) | 30 520 (48.9) |
| Employed | 2 124 059 (57.1) | 96 036 (69.2) | 3042 (27.0) | 0119 (4.65) | 4247 (6.81) |
| Married/partner | 2 069 552 (55.6) | 84 220 (60.7) | 6019 (53.4) | 0991 (38.8) | 23 548 (37.8) |
| Income | |||||
| Low | 1 082 427 (29.1) | 38 286 (27.6) | 5266 (46.7) | 1454 (56.9) | 36 205 (58.1) |
| Middle | 1 245 348 (33.5) | 45 530 (32.8) | 3408 (30.2) | 0815 (31.9) | 18 717 (30.0) |
| High | 1 394 038 (37.5) | 54 926 (39.6) | 2596 (23.0) | 0288 (11.3) | 7437 (11.9) |
| Danish origin | 3 263 925 (87.7) | 103 383 (74.5) | 8739 (77.5) | 2316 (90.6) | 59 186 (94.9) |
| Higher education | 477 065 (12.8) | 19 288 (13.9) | 0742 (6.6) | 0099 (3.9) | 2436 (3.9) |
| Wealth | |||||
| Low | 1 072 333 (28.8) | 47 463 (34.2) | 2175 (19.3) | 0202 (7.9) | 6307 (10.1) |
| Middle | 1 122 312 (30.2) | 44 976 (32.4) | 4670 (41.4) | 1255 (49.1) | 27 735 (44.5) |
| High | 1 527 168 (41.0) | 46 303 (33.4) | 4425 (39.3) | 1100 (43.0) | 28 317 (45.4) |
| Household size n≤2 | 3 313 067 (89.0) | 117 557 (84.7) | 9856 (87.5) | 2399 (93.8) | 58 909 (94.5) |
| Area level# | |||||
| Mean income, DKK | 287 915±67 593 | 289 951±74 259 | 287 678±75 397 | 291 937±74 262 | 281 695±64 536 |
| Median wealth, DKK | 120 780±169 063 | 110 713±176 085 | 108 098±176 085 | 115 528±180 480 | 114 052±158 615 |
| Unemployment rate, % | 1.0±0.5 | 1.2±0.5 | 1.2±0.5 | 1.1±0.5 | 1.0±0.5 |
| Low education rate, % | 22.6±7.6 | 21.5±7.6 | 22.3±7.7 | 21.7±7.6 | 23.9±7.5 |
| Population density, km−2 | 20.8±42.3 | 30.2±51.4 | 27.8±45.8 | 28.2±44.6 | 17.1±33.5 |
| GP visit rate, % | 77.3±2.0 | 76.8±2.2 | 76.9±2.1 | 76.9±2.0 | 77.5±1.8 |
| Air pollution in 2019 | |||||
| PM2.5, μg·m−3 | 7.4±0.5 | 7.5±0.4 | 7.5±0.4 | 7.5±0.4 | 7.4±0.5 |
| NO2, μg·m−3 | 10.7±2.4 | 11.5±2.4 | 11.4±2.3 | 11.5±2.3 | 10.6±2.3 |
| BC, μg·m−3 | 0.3±0.1 | 0.4±0.1 | 0.4±0.1 | 0.4±0.1 | 0.3±0.1 |
| PM10, μg·m−3 | 12.7±0.9 | 12.6±0.8 | 12.5±0.8 | 12.5±0.7 | 12.8±0.9 |
| O3, μg·m−3 | 54.5±2.2 | 54.0±2.0 | 54.1±2.0 | 54.1±1.9 | 54.7±2.2 |
Data are presented as n, mean±sd or n (%). GP: general practice; PM2.5: particulate matter with diameter ≤2.5 μm; NO2: nitrogen dioxide; BC: black carbon; PM10: particulate matter with diameter ≤10 μm; O3: ozone. #: area-level variables were based on the parish, the smallest administrative unit in Denmark.
We detected strong and significant positive associations of PM2.5, NO2 and BC with all three COVID-19 outcomes (table 2), which were strongest for mortality (23% and 18% higher risk of COVID-19 death for each 0.53 and 3.59 μg·m−3 increase in PM2.5 and NO2, respectively). Weaker (half of those with mortality) associations were found for PM2.5 and PM10 and COVID-19 incidence and hospitalisations, whereas associations with NO2 and BC were almost identical for the three outcomes. We found negative associations of O3 with the three COVID-19 outcomes and no association of PM10 or O3 with all-cause mortality. The hazard ratios attenuated most after area-level covariate adjustment, whereas additional adjustment for municipality-level SARS-CoV-2 positivity rates did not affect the estimates.
TABLE 2.
Associations between long-term exposure to air pollution and COVID-19 incidence, hospitalisation and mortality, as well as all-cause mortality among the 3 721 813 participants of the AIRCODEN cohort
| Model 1 | Model 2 | Model 3# | Model 4 | |
| COVID-19 incidence# (n=138 742) | ||||
| PM2.5 | 1.16 (1.12–1.21) | 1.23 (1.18–1.27) | 1.10 (1.05–1.14) | 1.09 (1.05–1.12) |
| NO2 | 1.25 (1.20–1.30) | 1.32 (1.27–1.37) | 1.18 (1.14–1.23) | 1.15 (1.11–1.18) |
| BC | 1.07 (1.03–1.10) | 1.07 (1.04–1.11) | 1.05 (1.01–1.08) | 1.04 (1.01–1.07) |
| O3 | 0.82 (0.80–0.85) | 0.79 (0.77–0.81) | 0.86 (0.84–0.89) | 0.90 (0.88–0.92) |
| PM10 | 1.05 (1.01–1.08) | 1.07 (1.04–1.11) | 1.09 (1.06–1.12) | 1.06 (1.03–1.09) |
| COVID-19 hospitalisation# (n=11 270) | ||||
| PM2.5 | 1.22 (1.15–1.30) | 1.24 (1.17–1.32) | 1.09 (1.01–1.17) | 1.09 (1.02–1.17) |
| NO2 | 1.35 (1.27–1.44) | 1.35 (1.28–1.43) | 1.19 (1.12–1.27) | 1.15 (1.08–1.22) |
| BC | 1.08 (1.04–1.12) | 1.08 (1.04–1.12) | 1.05 (1.01–1.08) | 1.04 (1.01–1.08) |
| O3 | 0.77 (0.74–0.81) | 0.77 (0.74–0.81) | 0.86 (0.82–0.91) | 0.89 (0.85–0.94) |
| PM10 | 1.10 (1.03–1.17) | 1.10 (1.04–1.17) | 1.14 (1.07–1.20) | 1.08 (1.03–1.14) |
| COVID-19 mortality (n=2557) | ||||
| PM2.5 | 1.33 (1.17–1.50) | 1.31 (1.16–1.48) | 1.23 (1.04–1.44) | 1.22 (1.04–1.43) |
| NO2 | 1.34 (1.21–1.49) | 1.29 (1.16–1.42) | 1.18 (1.03–1.34) | 1.12 (0.98–1.28) |
| BC | 1.09 (1.05–1.14) | 1.09 (1.05–1.13) | 1.06 (1.02–1.10) | 1.05 (1.00–1.09) |
| O3 | 0.78 (0.72–0.84) | 0.80 (0.74–0.87) | 0.87 (0.78–0.96) | 0.92 (0.82–1.02) |
| PM10 | 1.15 (1.04–1.27) | 1.13 (1.02–1.25) | 1.19 (1.07–1.33) | 1.13 (1.01–1.26) |
| All-cause mortality (n=62 359) | ||||
| PM2.5 | 1.02 (1.00–1.03) | 1.02 (1.01–1.03) | 1.02 (1.01–1.03) | 1.02 (1.01–1.02) |
| NO2 | 1.05 (1.02–1.08) | 1.04 (1.02–1.06) | 1.04 (1.01–1.07) | 1.03 (1.02–1.05) |
| BC | 1.02 (1.01–1.04) | 1.02 (1.01–1.03) | 1.01 (1.00–1.02) | 1.01 (1.00–1.02) |
| O3 | 0.98 (0.96–1.00) | 0.99 (0.97–1.00) | 1.00 (0.98–1.02) | 1.00 (0.99–1.02) |
| PM10 | 0.99 (0.97–1.00) | 0.96 (0.95–0.98) | 0.98 (0.96–0.99) | 0.98 (0.97–0.99) |
Data are presented as hazard ratio (95% CI). PM2.5: particulate matter with diameter ≤2.5 μm; NO2: nitrogen dioxide; BC: black carbon; O3: ozone; PM10: particulate matter with diameter ≤10 μm. Results are presented per interquartile range increase: 0.53 μg·m−3 for PM2.5, 3.59 μg·m−3 for NO2, 0.09 μg·m−3 for BC, 2.79 μg·m−3 for O3 and 1.14 μg·m−3 for PM10. #: n=3 721 810 (three people excluded due to COVID-19 infection before baseline on 1 March 2020). Model 1 adjusted for calendar time (time axis), sex (strata), age at baseline (strata) and region (strata); Model 2 additionally adjusted for marital status, household size, individual wealth, family income, education and occupational status; Model 3 (Main model) further adjusted for parish-level population density, mean income, median wealth, unemployment rate, primary or low education rate, the difference of those variables between parish and municipality, and municipality-level access to healthcare; Model 4 (time-varying Cox) additionally adjusted for municipality-level monthly COVID-19-positive rates as a proxy for spatial and temporal pandemic development (analysis for all-cause mortality was not stratified by region).
Compared with 1-year exposure, hazard ratios were almost identical with 3- and 10-year exposure windows of all air pollutants (supplementary figure S3). Associations with air pollution estimated by the ELAPSE model (supplementary table S2), which showed moderate correlation with the Danish model (e.g. correlation 0.51 between PM2.5 from the two models and 0.63 for NO2 from the two models) (supplementary figure S4), were comparable to those observed with the Danish model (supplementary figure S5): slightly weaker for PM2.5, NO2 and O3, and stronger for BC. We also observed considerably wider confidence intervals with the Danish model than with the ELAPSE model (supplementary figure S5). Exposure–response functions were linear or curvilinear for most of the pollutants and COVID-19 outcomes, although notably with weaker associations for PM2.5 and COVID-19 incidence <7 μg·m−3, and none for NO2 and COVID-19 incidence <8 μg·m−3. We also note the limitation of estimating associations in these lower exposure ranges where data are more sparse. We found generally stronger associations at lower exposure levels (supplementary figure S6 and supplementary table S3).
In the two-pollutant models, associations with NO2 and BC were robust to adjustment for PM2.5 with COVID-19 incidence and hospitalisation, while for COVID-19 mortality, results seemed most robust for PM2.5 (supplementary table S4). Associations remained unchanged after indirect adjustment for smoking and BMI (supplementary figure S7), in the population of those tested (supplementary figure S8), and for non-COVID-19 mortality (supplementary table S5). We found no associations in the first second pandemic wave, but significant positive associations in the second pandemic wave (supplementary figure S9).
We found stronger associations of PM2.5 and NO2 with COVID-19 incidence in those with lower SES and living in households with more than four inhabitants (figure 2), and in those with prior CVD, respiratory disease, ALRIs, lung cancer, dementia and diabetes (figure 3), with similar trends but weaker associations for COVID-19 hospitalisation, and no effect modification with COVID-19 mortality, notably limited by a small number of cases (supplementary figures S10–S13).
FIGURE 2.
Effect modification of the association between long-term exposure to air pollution and COVID-19 incidence among the 3 721 810 participants of the AIRCODEN cohort by sex, age and individual-level socioeconomic status characteristics at the study baseline on 1 March 2020: a) particulate matter with diameter ≤2.5 μm (PM2.5) and b) nitrogen dioxide (NO2). Results are presented per interquartile range increase: 0.53 μg·m−3 for PM2.5 and 3.59 μg·m−3 for NO2. The Wald test was used to calculate the global p-value. *: p<0.05.
FIGURE 3.
Effect modification of the association between long-term exposure to air pollution and COVID-19 incidence among the 3 721 810 participants of the AIRCODEN cohort by comorbidities at the study baseline on 1 March 2020: a) particulate matter with diameter ≤2.5 μm (PM2.5) and b) nitrogen dioxide (NO2). CVD: cardiovascular disease; ALRI: acute lower respiratory infection. Results are presented per interquartile range increase: 0.53 μg·m−3 for PM2.5 and 3.59 μg·m−3 for NO2. The Wald test was used to calculate the global p-value. *: p<0.05.
Discussion
In this large nationwide study, we detected strong associations between long-term exposure to air pollution and contracting SARS-CoV-2 infection and developing severe COVID-19 resulting in hospitalisation or death. People with chronic cardiometabolic and respiratory diseases, dementia and prior ALRIs, and those who are the most socioeconomically disadvantaged were most vulnerable.
Our findings generally agree with three other studies on long-term exposure to air pollution and COVID-19 incidence, hospitalisation or mortality, although with notable inconsistencies between studies [2, 13–15]. Our results corroborate those of Veronesi et al. [2] who in 62 848 residents of Varese (4408 positive in the first pandemic year) detected a 5.1% increase in COVID-19 rate for each 1 μg·m−3 increase in PM2.5. Sheridan et al. [14], in 424 721 subjects from UK Biobank (10 790 positive from 16 March 2020 to 31 December 2020), detected an OR of 1.06 and 1.05 for COVID-19 incidence for each 1.3 and 9.9 μg·m−3 increase in PM2.5 and NO2, respectively, but in contrast to our findings, found no associations with COVID-19 hospitalisations (n=1598) or deaths (n=568). Nobile et al. [13], in 1 594 308 subjects from Rome (79 976 positive and 2656 deaths from 1 January 2020 to 15 April 2021), found no associations with COVID-19 incidence, but reported 8% and 9% higher risk of dying from COVID-19 for each 0.92 and 9.22 μg·m−3 increase in PM2.5 and NO2, respectively. Furthermore, our results agree with Kogevinas et al. [15] who found associations with COVID-19 risk assessed in 9000 subjects by blood serum tests, self-reports, hospitalisations and death. Our finding of stronger associations with increasing severity of COVID-19 was also observed by Kogevinas et al. [15] and Nobile et al. [13], but not by Sheridan et al. [14]. We detect very strong associations with COVID-19, 10% increase in incidence and 23% in COVID-19 mortality per 0.53 μg·m−3 in PM2.5, substantially stronger than those in the aforementioned studies, but in line with early ecological studies [11]. Our association with COVID-19 mortality is 10 times that observed for PM2.5 and all-cause mortality (23% versus 2%) (table 2), in line with Nobile et al. [13] who reported eight times higher estimates for COVID-19 than for non-COVID-19 deaths (8% versus 1% per 0.92 μg·m−3). These strong associations may in part be explained by the residual confounding by smoking, BMI and other lifestyle factors, as observed in Sheridan et al. [14], or other unmeasured confounders. Strong associations may also be due to fact that the risk of all-cause mortality due to PM2.5 and NO2 is much higher in those ever tested than in the general population (supplementary figure S8), likely showing that those at higher risk of death are those being more tested. These strong associations may, however, be plausible, as we have previously reported considerably stronger associations of PM2.5 with all-cause mortality in Denmark than in other European countries [31], almost three times stronger than those in the Roman cohort utilised in the Nobile et al. [13] study. Furthermore, our finding of stronger associations at the lowest exposure supports stronger associations with COVID-19 in Denmark, with the low air pollution levels, as for all-cause mortality [20, 31, 32], than those observed in more polluted Italy, Spain and UK [2, 13–15]. However, some caution in interpreting our findings should be noted due to the Danish DEHM/UBM model estimates being less precise that the ELAPSE model estimates (supplementary figure S5) and that the large estimates could be more of a reflection of larger uncertainty (wider confidence intervals) rather than only strength of association. Notably, our findings were robust to adjustment for COVID-19 positivity and its geographical development over time, as the only study to date able to evaluate confounding by COVID-19 development over time. Furthermore, our study is unique in being able to utilise alternative air pollution exposures, which showed consistent results to the main analyses. Finally, our results were robust to limiting the population to tested individuals, together with other sensitivity analyses, supporting the plausibility of the observed associations (supplementary figures S3 and S5). Our results on air pollution leading to increased risk of contracting COVID-19 are also in line with the larger literature showing that long-term exposure to air pollution increases the risk of more severe COVID-19 outcome in those who contracted COVID-19 [16–24].
We found negative associations of O3 with all three COVID-19 outcomes, in agreement with two studies that had data on O3 [2, 15]. Veronesi et al. [2] also found significant negative associations of O3 with COVID-19 incidence, as did Kogevinas et al. [15] with all COVID-19 outcomes they considered. The negative associations we found might reflect the high negative correlation with NO2 and BC especially (supplementary figure S1), traffic pollutants which may be the most relevant for COVID-19, as suggested by two-pollutant models (supplementary figure S4). O3 and NO2 are negatively correlated because when O3 is close to combustion sources (e.g. major roads) it reacts with nitric oxide emitted from the combustion source to form oxygen and NO2. O3 therefore tends to be low near roadways, whereas BC emitted by traffic is high. NO2 is in part directly emitted from traffic and in part formed by atmospheric reactions, so it is also high near roadways.
Although the exact molecular mechanisms by which air pollution affects viral infection and the pathogenesis of COVID-19 remain unknown, there are several plausible pathways [6]. Exposure to air pollution may promote upregulation of the angiotensin converting enzyme 2 receptor relevant for viral entry, replication and assembly, and activate pro-inflammatory transcription factors, producing local inflammation. Furthermore, pollutant exposure reduces mucociliary clearance, promotes epithelial permeability, prevents macrophage uptake and disrupts natural killer cell function, all of which can increase viral spread and inflammation [6]. Subsequent enhanced inflammation can trigger neutrophil recruitment and further amplify inflammatory processes. Moreover, since pollution is believed to skew adaptive immune responses toward allergic/bacterial responses instead of antiviral immune responses, exposure to air pollution may result in enhanced virus-induced tissue damage and inflammation, promoting dysfunction of a number of organs, including the lungs, heart, kidney and brain, resulting in death [6]. Furthermore, air pollution likely additionally increases the risk of COVID-19 severity and death indirectly by increasing the risk of major respiratory and cardiometabolic diseases [1], which in turn increase COVID-19 severity/mortality [2–5].
We show, for the first time, the highest vulnerability to air pollution among those with the lowest SES, in part likely explained by indirect impacts of lifestyle, such as smoking, obesity, physical inactivity, etc., linked with COVID-19 incidence [5, 14]. Another explanation for the stronger association in the lowest SES is likely related to the lack of social support to lockdown, overcrowding and higher exposure. Furthermore, low SES is related to work in occupations that were exempt from lockdowns and working from home options, such as cleaning, security and service workers, bus drivers, etc., resulting in higher exposure to COVID-19, etc. Enhanced vulnerability to air pollution in large households likely reflects exposure to higher doses of virus and higher risk of infection. We also report vulnerability in chronic disease patients, in line with Veronesi et al. [2] who found stronger associations in coronary heart and obstructive lung disease patients, but with limited statistical power. We also note that effect modification results pointing to the higher vulnerability of chronic disease patients could be a reflection of more frequent testing among these individuals, considered to have a higher risk of infection or severe outcomes.
We find that associations of air pollution with COVID-19 were limited to the second pandemic wave, most likely explained by testing capacity with some effects from background non-pharmaceutical interventions [34] as well as behavioural modification. We should note that comparison of the results from the two waves demands some caution, as the two pandemic waves in Denmark were substantially different, presenting different stages of the pandemic and viral spread, prevention measures, testing capacity, as well as differences in personal behaviour and lifestyle. Only a very limited number of COVID-19 cases were detected in the first pandemic wave, notably those who travelled to Denmark from hotspot areas abroad (the first cases in Denmark were tourists from ski areas in Italy and Austria) early in the pandemic and close contacts (family, work colleagues and friends) of those infected. It is likely that these early infections of a new highly infectious virus in the completely closed society and controlled pandemic were likely spreading by mechanisms other than air pollution, mainly close contacts. The second pandemic wave was characterised by a large improvement in testing capacity (testing was offered only to those with severe symptoms or close contacts of cases in the first wave) free of charge to all, introduction of masks and reopening of the society (opening of work places, restaurants, theatres, cinemas, etc., conditional on a negative test). Thus in the second wave, when society was completely reopened and testing introduced for all, virus infection developed and took its natural course, resulting in many more people being infected, and in this more open and “normal” pandemic dynamic, a fraction of COVID-19 infections seemed to be driven by air pollution. Another explanation may be found in the new EU1 SARS-CoV-2 variant, which dominated the second wave in Denmark [35].
Our study is the largest to date and the first nationwide analyses of long-term exposure to air pollution and risk of COVID-19, facilitated by the internationally unique access to high-quality, centrally collected, individual-level Danish COVID-19 surveillance data for the entire population, in combination with national health, demographic and SES registers, and historical data on air pollution. As Denmark had one of the most intense testing strategies, with testing free of charge and easily accessible, we have arguably, especially in the second wave of the pandemic, been able to determine the most complete definition of incidence of SARS-CoV-2 infection, as >80% of the population was tested in the first 14 months of the pandemic, many multiple times. Acknowledging the limitations of this approach, the study benefited from detailed data on comorbidities, SES and spatiotemporal changes in COVID-19 pandemic positivity rates, and it is able to provide the first and the most comprehensive analyses of who is most susceptible to air pollution-related risk of COVID-19.
A major limitation of our study was the lack of data on smoking, physical activity, BMI, diet and nursing home residence. Notably, Denmark managed the COVID-19 pandemic exceptionally well with only minor excess mortality and without large COVID-19 mortality clusters in nursing homes as seen elsewhere [36]. Our definition of SARS-CoV-2 incidence is only partial, as the Danish COVID-19 surveillance system, especially in the early stages of the pandemic, could only identify a selected sample of all infected individuals, i.e. those with severe symptoms, close contacts of primary cases, hospitalised patients, etc. The testing policy was massively expanded in the second wave, where testing free of charge was offered to all, and a large portion of the population was tested daily, as a negative test was required to enter workplaces, universities, schools, restaurants, movie theatres, etc. We did not have data on vaccinations and could not address whether vaccination would have affected the observed associations. As vaccination was first introduced in Denmark at the end of the study, it would likely have had a minor impact on the observed associations. The first vaccination in Denmark was given on 27 December 2020 and only a fraction (9%) of the population (healthcare workers, the elderly and chronic disease patients) was vaccinated by 26 April 2021, the end of follow-up in this study. Another limitation is the definition of comorbidities based on hospital contact data only. This is a major limitation for some outcomes, such as ALRIs, as only a fraction of more severe ALRIs are captured, and not those treated by a general practitioner, for example, or those untreated. For chronic diseases such as major cardiometabolic diseases, lung cancer and dementia, this hospital contact definition better captures the true prevalence in the population than for ALRIs. Limitations in our exposure data include a larger uncertainty in the effect estimates (wider confidence intervals) with the Danish DEHM/UBM model than in those with the ELAPSE model (supplementary figure S5), and the possibility that large estimates could be more of a reflection of lower precision rather than only strength of association. Some inconsistencies, most pronounced for BC, between the two exposure models are also noted as a weakness. However, both exposure models, even where the size and precision of the observed associations differ, clearly show consistent associations with all three COVID-19 outcomes, for all three pollutants, supporting the plausibility of the air pollution link with COVID-19. Finally, we only had data for annual mean O3 in our main analyses with the Danish DEHM/UBM exposure model, whereas warm season average should be a better predictor of health-relevant exposures. However, our results are consistent with those for warm season O3 from the ELAPSE model (supplementary figure S5).
Conclusions
In a nationwide Danish study of the first 14 months of the COVID-19 pandemic, we found that long-term exposure to air pollution at low levels, well below current European Union limit values, is associated with increased risk of contracting SARS-CoV-2 and developing severe COVID-19 disease requiring hospitalisation or resulting in death. Chronic cardiometabolic, respiratory and neurodegenerative disease patients, individuals with prior ALRIs, and the lowest SES groups appear most susceptible and most likely to contract SARS-CoV-2 or develop COVID-19 due to air pollution. These findings contribute important new data to an increasing evidence base showing that air pollution is a risk factor for COVID-19.
Supplementary material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary material ERJ-00280-2023.SUPPLEMENT (810.9KB, pdf)
Shareable PDF
Footnotes
Author contributions: The study was conceptualised and designed by Z.J. Anderson and J. Zhang. Statistical analysis was conducted by J. Zhang, and Y-H. Lim helped with data management and statistical analysis. R. So helped with indirect adjustment for missing confounders analysis. J.T. Jørgensen helped with preparing the COVID-19 data for the analysis. M. Ketzel and J. Brandt developed the Danish DEHM/UBM model, and G. Hoek developed the ELAPSE model. Z.J. Andersen and J. Zhang drafted the manuscript. All authors have read and revised the manuscript and contributed to the interpretation of the results. All authors have approved the final draft of the manuscript.
Conflict of interest: T. Lange reports participation on data safety monitoring board for Novo Nordisk and Leo Pharma, outside the submitted work. S. Bhatt reports support for the present manuscript from Novo Nordisk Foundation. All other authors have no potential conflicts of interest to disclose.
Support statement: The research described in this article was conducted under contract to the Health Effects Institute (HEI) (4978/RFA20-1B/20-12), an organisation jointly funded by the US Environmental Protection Agency (EPA) (assistance award number CR-83590201) and certain motor vehicle and engine manufacturers. The contents of this article do not necessarily reflect the views of HEI, or its sponsors, nor do they necessarily reflect the views and policies of the EPA or motor vehicle and engine manufacturers. This study was also funded by the Novo Nordisk Foundation Challenge Programme (NNF17OC0027812). Funding information for this article has been deposited with the Crossref Funder Registry.
References
- 1.Health Effects Institute . State of Global Air 2020. 2020. www.stateofglobalair.org/sites/default/files/documents/2022-09/soga-2020-report.pdf Date last accessed: 13 May 2023.
- 2.Veronesi G, de Matteis S, Calori G, et al. Long-term exposure to air pollution and COVID-19 incidence: a prospective study of residents in the city of Varese, Northern Italy. Occup Environ Med 2022; 79: 192–199. doi: 10.1136/OEMED-2021-107833 [DOI] [PubMed] [Google Scholar]
- 3.Williamson EJ, Walker AJ, Bhaskaran K, et al. Factors associated with COVID-19-related death using OpenSAFELY. Nature 2020; 584: 430–436. doi: 10.1038/S41586-020-2521-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Atkins JL, Masoli JAH, Delgado J, et al. Preexisting comorbidities predicting COVID-19 and mortality in the UK Biobank community cohort. J Gerontol A Biol Sci Med Sci 2020; 75: 2224–2230. doi: 10.1093/gerona/glaa183 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Elliott J, Bodinier B, Whitaker M, et al. COVID-19 mortality in the UK Biobank cohort: revisiting and evaluating risk factors. Eur J Epidemiol 2021; 36: 299–309. doi: 10.1007/S10654-021-00722-Y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Woodby B, Arnold MM, Valacchi G. SARS-CoV-2 infection, COVID-19 pathogenesis, and exposure to air pollution: what is the connection? Ann NY Acad Sci 2021; 1486: 15–38. doi: 10.1111/NYAS.14512 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Jerrett M. Air pollution as a risk for death from infectious respiratory disease. Am J Respir Crit Care Med 2022; 205: 1374–1375. doi: 10.1164/RCCM.202202-0351ED [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Liu S, Lim YH, Chen J, et al. Long-term air pollution exposure and pneumonia-related mortality in a large pooled European cohort. Am J Respir Crit Care Med 2022; 205: 1429–1439. doi: 10.1164/RCCM.202106-1484OC/SUPPL_FILE/DISCLOSURES.PDF [DOI] [PubMed] [Google Scholar]
- 9.Bowe B, Xie Y, Yan Y, et al. Burden of cause-specific mortality associated with PM2.5 air pollution in the United States. JAMA Netw Open 2019; 2: e1915834. doi: 10.1001/jamanetworkopen.2019.15834 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Neupane B, Jerrett M, Burnett RT, et al. Long-term exposure to ambient air pollution and risk of hospitalization with community-acquired pneumonia in older adults. Am J Respir Crit Care Med 2010; 181: 47–53. doi: 10.1164/rccm.200901-0160OC [DOI] [PubMed] [Google Scholar]
- 11.Wu X, Nethery RC, Sabath MB, et al. Air pollution and COVID-19 mortality in the United States: strengths and limitations of an ecological regression analysis. Sci Adv 2020; 6: eabd4049. doi: 10.1126/SCIADV.ABD4049 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Villeneuve PJ, Goldberg MS. Methodological considerations for epidemiological studies of air pollution and the SARS and COVID-19 coronavirus outbreaks. Environ Health Perspect 2020; 128: 9. doi: 10.1289/EHP7411 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Nobile F, Michelozzi P, Ancona C, et al. Air pollution, SARS-CoV-2 incidence and COVID-19 mortality in Rome: a longitudinal study. Eur Respir J 2022; 60: 2200589. doi: 10.1183/13993003.00589-2022 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Sheridan C, Klompmaker J, Cummins S, et al. Associations of air pollution with COVID-19 positivity, hospitalisations, and mortality: observational evidence from UK Biobank. Environ Pollut 2022; 308: 119686. doi: 10.1016/J.ENVPOL.2022.119686 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Kogevinas M, Castaño-Vinyals G, Karachaliou M, et al. Ambient air pollution in relation to SARS-CoV-2 infection, antibody response, and COVID-19 disease: a cohort study in Catalonia, Spain (COVICAT Study). Environ Health Perspect 2021; 129: 117003. doi: 10.1289/EHP9726 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Bowe B, Xie Y, Gibson AK, et al. Ambient fine particulate matter air pollution and the risk of hospitalization among COVID-19 positive individuals: cohort study. Environ Int 2021; 154: 106564. doi: 10.1016/j.envint.2021.106564 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Chen C, Wang J, Kwong J, et al. Association between long-term exposure to ambient air pollution and COVID-19 severity: a prospective cohort study. CMAJ 2022; 194: E693–E700. doi: 10.1503/cmaj.220068 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Mendy A, Wu X, Keller JL, et al. Air pollution and the pandemic: long-term PM2.5 exposure and disease severity in COVID-19 patients. Respirology 2021; 26: 1181–1187. doi: 10.1111/RESP.14140 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Bozack A, Pierre S, DeFelice N, et al. Long-term air pollution exposure and COVID-19 mortality: a patient-level analysis from New York City. Am J Respir Crit Care Med 2022; 205: 651–662. doi: 10.1164/RCCM.202104-0845OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Chen Z, Sidell MA, Huang BZ, et al. Ambient air pollutant exposures and COVID-19 severity and mortality in a cohort of patients with COVID-19 in Southern California. Am J Respir Crit Care Med 2022; 206: 440–448. doi: 10.1164/rccm.202108-1909OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Hyman S, Zhang J, Andersen ZJ, et al. Long-term exposure to air pollution and COVID-19 severity: a cohort study in Greater Manchester, United Kingdom. Environ Pollut 2023; 327: 121594. doi: 10.1016/j.envpol.2023.121594 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Mendy A, Wu X, Keller JL, et al. Long-term exposure to fine particulate matter and hospitalization in COVID-19 patients. Respir Med 2021; 178: 106313. doi: 10.1016/j.rmed.2021.106313 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.López-Feldman A, Heres D, Marquez-Padilla F. Air pollution exposure and COVID-19: a look at mortality in Mexico City using individual-level data. Sci Total Environ 2021; 756: 143929. doi: 10.1016/j.scitotenv.2020.143929 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Jerrett M, Nau CL, Young DR, et al. Air pollution and meteorology as risk factors for COVID-19 death in a cohort from Southern California. Environ Int 2023; 171: 107675. doi: 10.1016/j.envint.2022 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.So R, Andersen ZJ, Chen J, et al. Long-term exposure to air pollution and mortality in a Danish nationwide administrative cohort study: beyond mortality from cardiopulmonary disease and lung cancer. Environ Int 2022; 164: 107241. doi: 10.1016/J.ENVINT.2022.107241 [DOI] [PubMed] [Google Scholar]
- 26.Brandt J, Silver JD, Frohn LM, et al. An integrated model study for Europe and North America using the Danish Eulerian Hemispheric Model with focus on intercontinental transport of air pollution. Atmos Environ 2012; 53: 156–176. doi: 10.1016/J.ATMOSENV.2012.01.011 [DOI] [Google Scholar]
- 27.Frohn LM, Geels C, Andersen C, et al. Evaluation of multidecadal high-resolution atmospheric chemistry-transport modelling for exposure assessments in the continental Nordic countries. Atmos Environ 2022; 290: 119334. doi: 10.1016/J.ATMOSENV.2022.119334 [DOI] [Google Scholar]
- 28.Hvidtfeldt UA, Ketzel M, Sørensen M, et al. Evaluation of the Danish AirGIS air pollution modeling system against measured concentrations of PM2.5, PM10, and black carbon. Environ Epidemiol 2018; 2: e014. doi: 10.1097/EE9.0000000000000014 [DOI] [Google Scholar]
- 29.Khan J, Kakosimos K, Raaschou-Nielsen O, et al. Development and performance evaluation of new AirGIS – a GIS based air pollution and human exposure modelling system. Atmos Environ 2018; 198: 102–121. doi: 10.1016/j.atmosenv.2018.10.036 [DOI] [Google Scholar]
- 30.de Hoogh K, Chen J, Gulliver J, et al. Spatial PM2.5, NO2, O3 and BC models for Western Europe – evaluation of spatiotemporal stability. Environ Int 2018; 120: 81–92. doi: 10.1016/J.ENVINT.2018.07.036 [DOI] [PubMed] [Google Scholar]
- 31.Stafoggia M, Oftedal B, Chen J, et al. Long-term exposure to low ambient air pollution concentrations and mortality among 28 million people: results from seven large European cohorts within the ELAPSE project. Lancet Planet Health 2022; 6: e9–e18. doi: 10.1016/S2542-5196(21)00277-1 [DOI] [PubMed] [Google Scholar]
- 32.Strak M, Weinmayr G, Rodopoulou S, et al. Long term exposure to low level air pollution and mortality in eight European cohorts within the ELAPSE project: pooled analysis. BMJ 2021; 374: n1904. doi: 10.1136/BMJ.N1904 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Shin HH, Cakmak S, Brion O, et al. Indirect adjustment for multiple missing variables applicable to environmental epidemiology. Environ Res 2014; 134: 482–487. doi: 10.1016/J.ENVRES.2014.05.016 [DOI] [PubMed] [Google Scholar]
- 34.Sharma M, Mindermann S, Rogers-Smith C, et al. Understanding the effectiveness of government interventions against the resurgence of COVID-19 in Europe. Nat Commun 2021; 12: 5820. doi: 10.1038/s41467-021-26013-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Hodcroft EB, Zuber M, Nadeau S, et al. Spread of a SARS-CoV-2 variant through Europe in the summer of 2020. Nature 2021; 595: 707–712. doi: 10.1038/s41586-021-03677-y [DOI] [PubMed] [Google Scholar]
- 36.COVID-19 Excess Mortality Collaborators . Estimating excess mortality due to the COVID-19 pandemic: a systematic analysis of COVID-19-related mortality, 2020–21. Lancet 2022; 399: 1513–1536. doi: 10.1016/S0140-6736(21)02796-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary material ERJ-00280-2023.SUPPLEMENT (810.9KB, pdf)
This one-page PDF can be shared freely online.
Shareable PDF ERJ-00280-2023.Shareable (339.8KB, pdf)