Association of fine particulate matter and daily mortality : a case-crossover study in Hong Kong

Abstract

Background: Air pollution is a serious concern all over the world, and there have been numerous studies showing its adverse effects on health outcomes including disease-specific hospitalization and mortality. Most of these studies were conducted in Western settings and focused on NO2, SO2, PM10 and black smoke; limited results on PM2.5 in Asia have been published.

Objectives: This study attempted to identify association between PM2.5 concentration and daily mortality due to all-natural causes, cardiovascular diseases, and respiratory diseases in Hong Kong, a tropical city in Asia from 2008 to 2011.

Methods: The study used a case-crossover study design with time-stratified referent selection strategy. The referents were selected on the same day of the week of the same month and year as the event day. In this way, biases due to autocorrelation, time trend, and seasonal pattern were controlled for by study design. Covariates including temperature, humidity, and gaseous pollutants (NO2, SO2, O3 and CO) were adjusted for by statistical modelling. The statistical method applied was Conditional Logistic Regression. Sensitive analyses using matched by month strategy were also conducted to check the robustness of the main analyses. The health outcome variables included were mortality due to all natural causes, cardiovascular disease, and respiratory disease. Each model was examined for effects of PM2.5 at each lag 0 through lag 5 day and, for the current and lag 1 day moving average (lag 0-1). Effects were measured in Excess Risks (ER) associated with 10 μg/m^3 increase in PM2.5.

Results: Significant associations with PM2.5 were observed for mortality from all natural causes and cardiovascular disease at lag 0, lag 1, and lag 0-1; and from respiratory disease only at lag 1. At lag 0-1, PM2.5 was associated with mortality from all natural causes, cardiovascular disease, and respiratory disease with the ER (95% confidence interval) of 0.74% (95% CI: 0.29, 1.19), 1.4% (95% CI: 0.52, 2.27), and 0.67% (95% CI: -0.30, 1.64), respectively. After adjusting for each of the co-pollutants in two-pollutant model, the effect magnitude dropped except that one adjusting for CO. The ERs of mortality in co-pollutant models from all natural causes ranged from 0.23% to 1.72%, from cardiovascular disease ranged from 1.00% to 2.93%, and from respiratory disease ranged from -0.10% to 1.83%. The results were robust in sensitivity analyses.

Conclusion: My study provides some information to support formulation of air quality control strategies and policies, and for updating air quality standards. Such information includes the overall and seasonal patterns of air pollutants and mortality in Hong Kong, as well as the excess risks of mortality associated with increase in PM2.5. Studies with individual data stratified for subgroups can be conducted in the future to investigate effect modification of lifestyle factors for the individuals and population.