This multicountry study will evaluate whether changes in mortality are associated with changes in ambient NO2 and PM2.5 levels before, during, and after the lockdown and disentangle the short-term effects of NO2 versus PM2.5 on mortality. The analysis will be conducted in four countries: China, Germany, Italy, and the United States.
To control the COVID-19 pandemic, unprecedented lockdown measures were implemented around the world, leading to large but temporary air pollution reductions. Both satellite and groundlevel measurements have shown substantial decreases in nitrogen dioxide (NO2) concentrations during the lockdowns at the city or county level, whereas changes in fine particulate matter (PM2.5) concentrations have been inconsistent, with reductions comparable to those of NO2, smaller reductions, or even increases. These disproportionate changes in ambient NO2 and PM2.5 pollution provide a rare opportunity for a natural experiment to address critical questions regarding the causal effects of NO2 and PM2.5 exposure on health.
We will first evaluate whether changes in mortality are associated with changes in ambient NO2 and PM2.5 levels, before, during, and after the lockdown (Aim 1). We propose to analyze this question by focusing on regions that implemented lockdown measures that led to air pollution reduction, but were not severely affected by the COVID-19 pandemic during the lockdown period: Jiangsu, China, Central and Southern Italy, Northern and Eastern Germany, and California, U.S. To estimate the effect of air pollution reductions on mortality during the lockdowns in these study regions, we will use a variant of the difference-in-difference approach. We will focus on total, nonaccidental, and cardiovascular mortality. These outcomes are unlikely to undercount cases outside of hospitals or to be greatly influenced by changes in diagnostic criteria for COVID-19.
To examine the whole mortality burden due to COVID-19, we will also examine excess total (i.e., all cause) mortality during the pandemic. This approach is designed to address a number of challenges that make it difficult to quantify air pollution effects in this setting, including varying interventions across space and time, disrupted health-care systems, and uncertain quality of health data. In addition we will leverage this unprecedented natural experiment of disproportionate reductions in NO2 and PM2.5 pollution during the COVID-19 lockdowns to perform stratified analyses to disentangle the short-term effects of NO2 versus PM2.5 on mortality (Aim 2), a challenge that has proven difficult in prior research due to the high correlation between these two pollutants.
We will first quantify the air pollution changes during the lockdown using a counterfactual modeling approach and then stratify the locations (cities, counties, or municipalities) in each country into a treatment group (large reduction in NO2, but little to no reduction in PM2.5) and a control group (comparable reductions in NO2 and PM2.5). We will apply the difference-in-difference approach developed in Aim 1 to estimate the health effects of NO2 in the treatment and control groups separately; and then calculate the difference between these two estimates to capture the effect of NO2 exposure, independent of PM2.5 exposure. This proposal specifically addresses the first objective of the RFA 20-1B (“What are the effects of the unprecedented interventions implemented to control the COVID-19 pandemic on emissions, air pollution exposures, and human health?”).