Estimating the effect of annual PM2·5 exposure on mortality in India: a difference-in-differences approach

Abstract

Background

In 2019, the Global Burden of Diseases, Injuries, and Risk Factors Study attributed 0·98 million deaths to ambient air pollution in India based on potentially inappropriate exposure–response functions from countries with low air pollution levels. Instead, using data from India, we investigated long-term exposure to PM_2·5 and all-cause mortality with a causal inference method.

Methods

We collected national counts of annual mortality from 2009 to 2019 from the Civil Registration System at the district level to calculate annual district-level mortality rate as our main outcome and obtained annual PM_2·5 concentrations from a high-resolution spatiotemporal model. We applied an extended version of the difference-in-differences design by use of generalised additive models with quasi-Poisson distribution, including indicator variables and separate time trends for spatial administrative divisions. PM_2·5 concentrations obtained at 1 km × 1 km spatial resolution across the country were used to calculate annual district-level mean PM_2·5 concentrations. Similarly, we collected confounders at the district level, such as mean and SD of quarterly temperatures, gross domestic product per capita, population aged 60 years or older, clean cooking fuel usage, literacy in women, and median age. The spatial unit of analysis was administrative division.

Findings

The annual median population-weighted PM_2·5 was 38·9 μg/m³ (5–95th percentile 19·7–71·8 μg/m³). The full population lived in areas with PM_2·5 concentrations exceeding the 5 μg/m³ annual mean recommended in the WHO guidelines, and 1·1 billion of 1·4 billion (81·9% of the total population) lived in areas above the Indian National Ambient Air Quality Standards for annual mean PM_2·5 not exceeding 40 μg/m³. A 10 μg/m³ increase in annual PM_2·5 concentration was associated with an 8·6% (95% CI 6·4–10·8) higher annual mortality. Based on the Indian National Ambient Air Quality Standards, a total of 3·8 million (95% CI 2·9–4·9) deaths between 2009 and 2019 were attributable to PM_2·5, amounting to 5·0% (3·8–6·4) of total mortality. Based on the WHO guidelines, a total of 16·6 million (13·0–21·8) deaths were attributable to PM_2·5, amounting to 24·9% (19·5–32·5) of total mortality.

Interpretation

Our difference-in-differences approach allowed us to assess the full extent of registered deaths in the most populated country in the world, which has high levels of air pollution. We provide new evidence of increased mortality risk from long-term PM_2·5, which emphasises the need for tighter regulatory standards to potentially substantially reduce mortality across India.

Funding

Swedish Research Council for Sustainable Development.