Air pollution effects on life expectancy before and after restriction of fuel sulphur content in Hong Kong

Abstract

Background: In Hong Kong on July 1, 1990 the Government implemented a restriction of fuel sulphur content (the intervention), which led to reductions in ambient SO2, nickel and vanadium concentrations and to some health gains in children and adults. A reduction in mortality trend was observed over five years following the intervention. However the health benefits were not assessed directly from modeling of health outcomes and air pollutant concentrations. In this analysis we (a) assess trends in sixteen chemical sub-species of particulates five years before and after the intervention; and (b) extend the distributed lag models of time series methodology to estimate changes in life expectancy (LE ) from previous exposure as well as the LE gain attributable to the intervention. Methods: Daily mortality counts with date and causes of death for the period January 1985 to December 31, 2005 were retrieved from the deaths register of the Immigration Department with linkage to data in the Census and Statistics Department. Air pollutant concentrations data were obtained from the Environmental Protection Department. For (a), time trends were determined by general linear models with adjustment for meteorological conditions. Kolmogorov Zurbenko filter was applied with 90-day average of 16 chemical sub-species of particulates to determine the turning points for change in time trends. For (b), we modeled the ratio of the observed death rate (D(t)) and the reference (Dref(t)) as a linear function of past exposure in a window of up to four years and capture the corresponding coefficients as impact functions (f(i)) for the effects of air pollution at lag days i. The LE loss due to a permanent increase in air pollution was estimated as a cumulative sum of f(i) over i. Results: (a) There were changes in particulate sub-species at different time points which, for nickel and vanadium, were close to the day of implementation of the intervention. (b) Daily non-accidental deaths, regressed against PM10 and SO2 with observation windows up to 4 years, were statistically significant up to 3.5 years, with effect estimates smaller for exposure further in the past. The magnitude of the SO2 coefficients was comparable to those for PM10. The estimates of LE changes were consistent when different definitions for mortality rate and reference were used, and with results of other cohort studies. Conclusion: The changes in particulate sub-species may contribute to the health gains after the intervention. The proposed method in this study allows the determination of the mortality impacts of chronic exposure and the life expectancy change using distributed lags models of time series studies.