Quantifying the effects of air pollution on respiratory ill health treated in primary care when the locations of the populations at risk are partially unknown

Most air pollution and health studies focus on severe outcomes such as hospitalisations and deaths, overlooking the impact that air pollution may have on non-hospitalised respiratory ill health treated in primary care. This paper presents a new study investigating the effects of NO2, PM10 and PM2.5 on the prescription rates of respiratory medications in Scotland between 2016 and 2020 at a monthly resolution. To enhance the spatial accuracy of the exposure estimates, air pollution predictions at a 1 km2 resolution are realigned to General Practioner (GP) surgery patient populations by accounting for where patients are likely to live rather than just where the GP surgery is. A Bayesian spatio-temporal conditional autoregressive model is utilised to account for spatial and temporal dependencies in the data, and this paper proposes two novel spatial neighbourhood matrices to better represent the spatial closeness among the patient populations registered at each GP surgery. These matrices improve model performance in capturing spatial correlation compared to standard distance-based approaches, such as using K-nearest neighbours approach. The results of the study suggest that particulate matter pollution has a significant impact on prescription rates for inhaled corticosteroids that are taken to prevent the symptoms of respiratory ill health, while NO2 demonstrates no such association.