Disentangling non-linear and time-varying effects in assessing the short-term impact of air pollution on mortality: evidence from a 12-year study in a high-risk Italian area

Introductioncurrent evidence on the short-term effects of air pollution on mortality often overlooks potential temporal variation and non-linear exposure-response relationships, which may bias effect estimates and limit the accuracy of health risk assessments. Methodsthis study addresses these gaps by examining temporal changes and non-linear associations between daily concentrations of PM10, PM2.5, NO2, and SO2 and mortality from natural, cardiovascular, and respiratory causes across eight municipalities in Tuscany, Italy, from 2008 to 2019. Environmental and mortality data were obtained from official sources; missing environmental data were handled through multiple imputation. Time-invariant and time-varying linear effects were estimated using Poisson regression models, and non-linear dose-response curves were assessed using splines. ResultsPM2.5 and SO2 were positively associated with natural and respiratory mortality, while PM10 and NO2 showed weaker or no associations. Stronger effects were observed during 2012-2015, despite lower pollutant concentrations. SO2 also exhibited a non-linear relationship with cardiovascular mortality, with greater effects at lower concentrations. Conclusionthese findings suggest that reductions in pollutant levels do not necessarily imply reduced health risks, potentially due to changes in pollutant composition or interactions with meteorological factors. This study underscores the importance of accounting for both temporal variation and potential non-linearity in air pollution health impact assessments. What is already known on this topicThere is strong evidence that short-term exposure to air pollutants such as PM10, PM2.5, NO2, and SO2 is associated with increased risks of natural, cardiovascular, and respiratory mortality. However, most existing studies rely on the assumption that these effects are constant over time and that the exposure-response relationship is linear. The possibility that health effects vary across time due to changing environmental or contextual factors - and that such variation may be non-linear - has received limited attention. What this study addsThis study investigates short-term mortality effects of air pollution over a 12-year period in eight municipalities in Tuscany, Italy, using both linear and non-linear models. It demonstrates that the effects of PM10, PM2.5, and SO2 on mortality are not temporally constant and that stronger associations are observed in periods with lower average pollutant levels. While SO2 shows a clear non-linear relationship, non-linearity alone does not fully explain the time variation observed for particulates, suggesting that changes in pollutant composition or environmental conditions may also play a role. How this study might affect research, practice or policyBy revealing that pollutant-related health risks can vary significantly over time and may not follow a simple linear pattern, this study underscores the importance of integrating temporal variability and non-linearity into air pollution epidemiology. These insights could improve the accuracy of health impact assessments, support more responsive air quality regulations, and inform future policies aimed at protecting public health under evolving environmental conditions.