Use of novel statistical methods in assessing particulate air pollution and evaluating its association with mortality in China

Abstract

Particle matter (PM) has been associated with numerous adverse health effects including cardiovascular disease, chronic obstructive pulmonary disease and lung cancer in experimental studies and observation studies. The close and quantitative relationship between exposure to high concentrations of coarse particles (PM10) and fine particles (PM2.5) and increased mortality and morbidity in human has been confirmed in many epidemiology studies. The increases in urbanization and road motor vehicle use in China have raised concerns about the health effects of exposure to PM pollution from traffic emissions.

In Study I, we obtained hourly PM2.5 concentrations at 35 air quality monitoring (AQM) stations in Beijing between 2013 and 2014, and daily meteorological data and geographic information during the same time period. Based on the PM2.5 concentrations from different AQM station types, a two-stage method comprising a dispersion model and a generalized additive mixed model (GAMM) was developed to estimate the traffic and non-traffic contributions to daily PM2.5 concentrations separately. The method provides a new solution for ecological and epidemiological studies to estimate the road traffic contribution to PM2.5 concentrations when there is limited vehicle and emission profiles’ data.

In Study II, we used causes of death registry and daily AQM data from eight districts in Beijing between 2009 and 2010 to demonstrate an application of Bayesian model averaging (BMA) method and provide a novel modelling technique to assess the association between PM10 concentration and respiratory mortality. The BMA method within GAMM frame gave slightly but noticeable wider confidence intervals (CIs) for the single-pollutant model and the principal components based model, which indicates that BMA may provide a useful tool for modelling uncertainty in time-series studies when evaluating the effect of air pollution on fatal health outcomes.

In Study III, we evaluated the effects of PM2.5 concentrations on non-accidental mortality as well as their interactions with extreme weather conditions and weather types in Shanghai between 2012 and 2014. A fully Bayesian generalized additive model (GAM) was set up to link the mortality with PM2.5 and weather conditions. We found that the effects of PM2.5 on non-accidental mortality differed under specific weather conditions.

In Study IV, we compared the estimates from frequentist GAM and Bayesian GAM with simulated data. We also examined the sensitivity of Bayesian GAM to choices both of the priors and of the true parameter. The frequentist GAM and Bayesian GAM showed similar means and variances of the parameters of interest. However, the estimates from Bayesian GAM show relatively more fluctuation, which to some extent reflects the uncertainty inherent in Bayesian estimation.

In conclusion, PM pollution poses great threat to human health in China. Road traffic is one of the major sources of PM pollution, and our two-stage model is a useful tool to proportionate its contribution to PM pollution in large cities such as Beijing where daily meteorological and traffic data are available. Given the statistically significant interactions between PM2.5 and weather, and climate and pollution challenges, adequate policies and public health actions are needed, taking into account the interrelationship between the two hazardous exposures. Although computationally intensive, Bayesian approaches would be better solutions to avoid potentially over-confident inferences in traditional frequentist methods. With the increasing computing power of computers and statistical packages available, fully Bayesian methods for decision making may become more widely applied in the future.