A Bayesian hierarchical approach to calculate long-term exposure to air-pollution and COVID19 mortality: A nationwide study in England

• Activity Number: 388 - Advances in Disease Mapping
• Type: Invited
• Date/Time: Wednesday, August 5, 2020 : 1:00 PM to 2:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Garyfallos Konstantinoudis* and Tullia Padellini and Marta Blangiardo
• Companies: Imperial College London and Imperial College London and Imperial College London
• Keywords: CAR models; IGMRF; BYM; COVID19

Abstract

As of June 26 2020, COVID19 has caused more than 479,000 deaths globally. The UK is one of the countries most affected, with the deaths reaching 40,000 as of June 26 2020. Known factors that affect COVID19 mortality include age, sex, comorbidities and ethnicity. Long term exposure to air-pollution is also hypothesised to play a role. In this nationwide study in England we aim to quantify the effect of long term exposure to NO2 and PM2.5 on COVID19 mortality. We retrieve COVID19 cumulative deaths until June 14 from Public Health England and calculate exposure by calibrating the Pollution Climate Mapping model against a set of monitors, allowing for spatiotemporal random effects. We fit a Bayesian spatial model with the BYM prior while adjusting for confounders. We found a 3% (CrI: 2.9-3.1%) increased COVID19 mortality for every unit increase in the NO2 and 6.0% (5.1-7.0%) for every unit increase in the in the PM2.5 based on the models without confounding adjustment, whereas 0.0% (-0.6, 0.7) and 0.0% (-1.8, 1.8) in the fully adjusted models. In conclusion, we found lack of evidence that supports that long term exposure to NO2 and and PM2.5 affects COVID19 mortality in England.