Abstract.
Air pollution is both an environmental and a social problem, as it leads to a multitude of adverse on human health, ecosystems and the climate. Starting from the late 1980s a substantial literature has emerged on health effects from exposure to air pollution, which escalated at the end of 2013 with a study by the International Agency for Research on Cancer which found sufficient evidence to classify air pollution as a leading environmental cause of cancer deaths.
From the statistical point of view, studying the association between health and air pollution poses multiple challenges. First, data regarding pollutant concentration and health outcomes (i.e. mortality/morbidity data) are spatially misaligned, giving rise to the so-called change of support problem. Secondly, air pollution monitoring networks can be subject to selection bias if monitoring sites are not randomly distributed but are located in the more polluted areas, an issue referred to as preferential sampling. Finally, air pollution exposure estimate in ecological studies may suffer from potential confounding and measurement error. In this work we consider all these challenging issues while implementing a spatio-temporal model for air pollution health risk assessment, both through a simulation study and a real case application regarding drug prescriptions in England.