A component-specific exposure-mortality model for ambient PM2.5 in China: findings from a nationwide epidemiology based on outputs from a chemical transport model

Long-term exposure to ambient fine particles (PM2.5) has been evidenced a leading contributor to premature mortality in China and many other countries. Previous studies assess the health risk using an exposure-response function, such as an exposure-mortality model (EMM) based on total concentration of PM2.5. However, the risk assessment method can be problematic by ignoring the unequal toxicity between different chemical components of PM2.5. To derive a components-specific EMM (CS-EMM), we conducted a whole-population-based epidemiology study in China. Based on Chinese Population Census data on 2000 and 2010 and the concentrations of ambient PM2.5 and its components in 2000 to 2010 assessed by satellite-based concentrations of PM2.5 and composition fractions simulated by a chemical transport model, we used a difference-in-difference approach to associate county-level changes of census-based total mortality with changes of PM2.5 and its components between 2010 and 2000. The chemical components of PM2.5 simulated by the model included sulfate (SO42-), nitrate (NO3-), ammonium (NH4+), organic carbon (OC), and black carbon (BC). We further compared CS-EMM with EMM based on a single pollutant of PM2.5 (PM2.5-EMM) or black carbon (BC-EMM), through evaluating their performance in risk assessment. Using census-based total mortality and cross validation we evaluated the performance of mortality prediction of an EMM, and found that the components-specific EMM (CS-EMM) outperformed PM2.5-EMM and BC-EMM. For instance, CS-EMM, PM2.5-EMM, and BC-EMM all overestimated the average number of city-level deaths by 117, 142, and 149, respectively, while CS-EMM overestimated the lowest. Moreover, CS-EMM had an advantage to interpret toxicity of PM2.5 mixture as an entirety. From 2000 to 2010, CS-EMM attributed 205,496 increase in PM2.5-associated mortality across China to the joint contribution of the growth of total concentration and the reduction of PM2.5 toxicity. Among the components, BC contributed 6.4% of PM2.5 concentration growth, but was corresponded to 46.7% of increment in PM2.5-associated deaths. This study developed a framework to establish and validate an exposure-response function based on PM2.5 components, and illustrated its advantages in terms of risk prediction and result interpretation in China. Our approach can be utilized to evaluate how the chemical composition modified the health impacts of PM2.5, and help the policymakers to target the toxic sources of air pollution.