Publication

Background Short-term exposure to fine particulate matter (PM2.5) is associated with increased risk of hospital admissions and mortality, and health risks differ by the chemical composition of PM2.5. Policies to control PM2.5 could change its chemical composition and total mass concentration, leading to change in the subsequent health impact. However, there is little ence on whether associations between PM2.5 and health exhibit temporal variation. We investigated whether risks of hospitalisations from short-term exposure to PM2.5 varied over time in the USA. Methods We did a time-series analysis using a national dataset comprising daily circulatory and respiratory hospitalisation rates of Medicare beneficiaries (age >= 65 years) and PM2.5 in 173 US counties from 1999 to 2016. We fitted modified quasi-Poisson models to estimate temporal trends of associations within a county, and pooled county-level estimates using Bayesian hierarchical modelling to generate an overall estimate. Findings The study included 10 559 654 circulatory and 3 027 281 respiratory hospitalisations. We identified changes in the national average association between previous-day PM2.5 and respiratory hospitalisation over time, with a U-shape that is robust under stratification, linear, and non-linear models. The change in risk of respiratory hospitalisation per 10 mu g/m(3) increase in previous-day PM2.5 decreased from 0.75% (95% posterior credible interval 0.05 to 1.46) in 1999 to -0.28% (-0.79 to 0.23) in 2008, and then increased to 1.44% (0.00 to 2.91) in 2016. No statistically significant temporal change was observed for associations between same-day PM2.5 and circulatory hospitalisation. Interpretation Hospitalisation risk from PM2.5 changes over time and has increased over the past 7 years in study, especially in northeastern USA. The temporal trend differs by cause of hospitalisation. This study emphasises the necessity of evaluating temporal heterogeneity in health impacts of PM2.5 and suggests caution in applying association estimates to a different time period. Copyright (C) 2021 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY-NC-ND 4.0 license.