Here’s the 2017 study mentioned by name in the article: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(17)30505-6/fulltext https://www.thelance

Here’s the 2017 study mentioned by name in the article: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(17)30505-6/fulltext https://www.thelancet.com/journals/lancet/article/PIIS0140-6... I don’t pretend to understand fully, but these seem to be some sentences explaining their methodology: > Ambient PM2·5 was the fifth-ranking mortality risk factor in 2015. Exposure to PM2·5 caused 4·2 million (95% uncertainty interval [UI] 3·7 million to 4·8 million) deaths and 103·1 million (90·8 million 115·1 million) disability-adjusted life-years (DALYs) in 2015, representing 7·6% of total global deaths and 4·2% of global DALYs, 59% of these in east and south Asia. > Attributing deaths and disability-adjusted life-years (DALYs) to ambient air pollution requires spatially and temporally resolved estimates of population-weighted exposure, specification of a theoretical minimum risk exposure level (TMREL), estimation of relative risks across the exposure distribution, and estimates of the deaths and DALYs for diseases linked causally to air pollution. We combined estimates of exposure and relative risk to estimate the population-attributable fraction (PAF), the proportion of deaths and DALYs attributable to exposure above the TMREL. The numbers of deaths and DALYs for specific diseases were multiplied by the PAF to estimate the burden attributable to exposure. A more general description of the methods used to estimate the PAF and attributable burdens in GBD 2015 has been reported previously;3 here, we present details specific to air pollution. While I don’t know enough about this methodology to defend this study, I would just point out that no matter what you consider to be a cause of death, no matter how “direct” you think that cause is, is still “merely” a death of someone who would have died later anyway. This mode of argument gets used all the time to minimize deaths of e.g. COVID-19 (“most of the deaths are old people with multiple comorbidities”), but this is a weak mode of argument precisely because it could apply just as well to literally any situation. We generally wouldn’t apply this mode of argument to minimize, for example, a serial killer at a nursing home.