Reduction in air pollution and attributable mortality due to COVID-19 lockdown — Authors' reply

We thank Hicham Achebak and colleagues for their correspondence about our Comment.¹ We appreciate their comments on the potential biases in quantifying the changes in air pollution and attributable mortality during the COVID-19 lockdown, which provides us with the opportunity to discuss these issues in more detail.

First, we agree that meteorological changes could play a role in changes in air pollution during the COVID-19 lockdown in China. However, this potential bias was reduced to some extent in our difference-in-difference approach, which compared the air pollution changes during quarantine versus the before quarantine period with those in the previous 4 years. In our analysis, nitrogen dioxide (NO₂) and PM_2·5 dropped by 12·9 μg/m³ and 18·9 μg/m³ during quarantine across 367 Chinese cities, respectively. Similarly, in another analysis by He and colleagues,² PM_2·5 dropped by 21·1 μg/m³ because of the lockdown in China, which was robust to the adjustment of weather variables (22·3 μg/m³). This robustness might be due, in part, to the similarity in meteorological conditions (eg, air pressure, temperature, and wind field) during the COVID-19 lockdown in 2020 compared with previous years, as was found in the Yangtze River Delta region of China.³

We appreciate the authors' endeavor in proposing a meteorology normalisation technique based on machine learning and welcome different and well designed approaches to precisely quantify the air pollution changes from COVID-19 lockdown. However, we would like to caution that this method does not consider the annual trend of emission changes, especially in China, where dramatic declines in air pollution have occurred over the past few years because of aggressive clean air policies. The national population-weighted annual mean concentrations of PM_2·5 in China decreased from 61·8 μg/m³ to 42·0 μg/m³ from 2013 to 2017, with a 18·1 μg/m³ reduction from anthropogenic emission and a 1·7 μg/m³ reduction from meteorological variations.⁴ As explained in our Comment, our difference-in-difference approach also controlled for the long-term declining trend in air pollution in China.

Second, we used baseline mortality from the China Health and Family Planning Statistical Yearbook 2018 because of data availability. Baseline mortality for the year 2020 will not be available until mid or late 2021 and will be influenced by the disruption of health-care systems as a result of COVID-19, leading to challenges in disentangling the mortality changes due to an improvement in air pollution from mortality changes due to the effects of the pandemic. Moreover, using baseline cause-specific mortality from one year and exposure-response functions from the published literature is widely accepted in calculating air pollution mortality burden.5, 6 We applied nationwide exposure-response functions from the published literature,7, 8 rather than doing time-series analyses in our study. Therefore, our estimates should be interpreted as the air pollution-related premature deaths that would have occurred during the 34 days from Feb 10 to March 14, 2020, had there not been a lockdown.

Finally, we fully agree with the authors' comment about the potential short-term benefits from the air pollution improvements during the COVID-19 lockdown. Air pollution reduction and its associated health benefits during COVID-19 are likely to be temporary, as was seen with the slight reversal in improved health outcomes after the Beijing Olympics because of cessation of air pollution controls.⁹ Moving forward, a more sustainable and healthier society will require increased investment in clean and renewable energy, low-carbon infrastructure, active transportation, and climate-friendly lifestyles.¹⁰