14. Europe-wide hybrid land use
regression models (100x100 m)
Land use and road data, with
satellite observations and
dispersion model estimates as
additional predictors
PM2.5 Full Model
Inset 1
PM2.5 conc
µg/m3
Central exposure assessment
Local exposure models
Existing LUR and/or dispersion
models
23. Studies on air pollution and
Covid-19
• ‘ecological’: Cole, The Netherlands, investigating
tests, hospital admissions, deaths
• ‘semi-ecological’: Bowe, USA, hospital admissions
among veterans with positive Sars-CoV2 test
• ‘individual’: Elliott, UK, mortality among subjects in
UK Biobank who tested positive or negative for
Sars-CoV2
25. Features of Cole study
• Ecological analysis of 355 municipalities
• 46,000 positive tests, ~11,600 hospital
admissions, ~4,000 deaths until June 5,
2020
• Air pollution from 1x1 km maps 2015-2019
• Air pollution and covariates calculated as
means per km2 of municipality surface
area; average surface= ~120 km2
30. COVID-19 (March-June 2020) in The
Netherlands: Spatial correlations
School
vacation
February
2020
Carnival
Feb 23-26
North: Feb 15-21
Middle: Feb 22-28
South: Feb 22-28
33. Methods
• A national cohort of 169,102 US Veterans with
COVID-19 positive test aimed to examine the
association between chronic exposure to PM2.5 and
risk of hospitalization.
• Estimates of annual average PM2.5 for the US in 2018
were available from satellite-based PM2.5 estimates
at approximately 1 km2 resolution
• Association in prespecified subgroups to identify
sensitive populations.
34. Covariates
Individual level covariates
• Age
• Race (Black, Other, and White, self-report during clinical
encounters)
• Sex
• Smoking status
Contextual characteristics at county level
• Socioeconomic deprivation (ADI), constructed from: income,
education, employment, and housing quality.
• Population density
• Percentage with limited access to healthy food etc.
35. Non-linear exposure response curve of the association
between PM2.5 and risk of hospitalization among a national
cohort of US Veterans who tested positive for COVID-19.
36. Effect modification of the association
between PM2.5 and risk of hospitalization by
race, age, sex, and deprivation (ADI)
37. Eur J Epidemiol 2021
• Subjects from UK Biobank with extensive covariate
information
• 459 COVID-19 deaths until September 21, 2020
• Air pollution (PM2.5, black carbon, NOx)
estimated for the year 2010 at the home address
38. Eur J Epidemiol 2021
• Significant independent risk factors:
• Age, male sex, black vs. white
• Healthcare worker, current smoker, having
cardiovascular disease, hypertension, diabetes,
autoimmune disease, and oral steroid use at
enrolment
• Air pollution not significant
39.
40.
41.
42. Inequalities & Covid-19
• Higher exposure to SARS-CoV-2: crowded housing,
public transport, unable to work from home, work-
related infections etc.
• Greater vulnerability: poor diet, obesity, diabetes,
heart disease, lung disease
• Lower health literacy: less likely to understand and
follow precautionary measures such as social
distancing, use of face masks, unwillingness to get
vaccinated etc.
43. Air pollution & inequalities
• Exposure to air pollution may be higher in deprived
areas (but note: reverse examples exist!)
• Air pollution may be more damaging for subjects
who are in poor health already
• Air pollution may be more damaging for subjects
with poor diets, poor physical condition etc.
44. Air pollution and Covid-19
• Air pollution increases diabetes, heart disease, lung
disease
• Air pollution increases susceptibility to bacterial
and viral airway infections
• >> air pollution is likely to have some effect on risk
of SARS-CoV-2 infections; on Covid-19 disease
severity & hospitalisation; and on case-fatality rate
once disease has developed
• Field of study still in its infancy
45. 3.5.5. How do we make progress?
Where do we look for progress? Studies at the individual
level are urgently needed in which the incidence,
progression and remission of COVID-19 is investigated in
large, well characterised cohorts
• Define outcome measures
• Study well-defined patients
• Study well-defined population-based cohorts
• Develop suitable co-variates
• Test negative designs
• Virus sequencing (only small studies)
• Define relevant exposure time windows
• Time series studies with/without lockdown measures
46. WHO AQG were last
revised in 2005
• New update almost finished
• Systematic reviews published in Environment
International
• Update follows a detailed WHO handbook for
guideline development
• Publication expected summer 2021
46