On the supra-linearity of the relationship between air pollution, mortality and hospital admission in 18 French cities

Purpose Understanding the relationship between an environmental determinant and a given health outcome is key to inform public health policies. The short-term mortality and morbidity responses to outdoor air pollutants are traditionally assessed as a log-linear relationship, but few studies suggest a possible deviation from linearity. This paper investigates the shape of the relationship between ozone, NO 2 and fine particulate matter (PM 10 and PM 2.5 ), mortality and hospital admissions in 18 French cities between 2000 and 2017. Method A multi-centric time series design, using quasi-Poisson generalized additive models, was used. Four approaches were compared to model concentration–response curves (log-linear, piecewise-linear with a priori defined breakpoints, piecewise-linear with no a priori breakpoint and cubic spline). Results All the models indicated evidence of supra-linearity between PM 10 , PM 2.5 , NO 2 , mortality and hospital admissions. For instance, with a log-linear model, a 10 µg/m 3 increase in PM 2.5 was associated with a 0.4% [95% CI 0.2; 0.7] increase in non-accidental mortality. When using a piecewise model with a priori set breakpoint at 10 µg/m 3 , the mortality increase was 3.8% [4.4; 6.3] below 10 µg/m 3 , and 0.3% [0; 0.6] above. Non-significant impacts of ozone were found for concentrations below 90 µg/m 3 to 120 µg/m 3 , with some variability in the identified threshold across the heath indicator studied. Conclusion The supra-linearity of the relationship between PM 10 , PM 2.5 , NO 2 , mortality and hospital admissions supports the need to further reduce air pollution concentrations well below regulatory values.