Many analyses have been conducted worldwide to estimate the health benefits from air pollution abatement. A key issue in the estimation of these benefits is the unit risk, or the slope of the concentration-response function. Among all effects, premature mortality is with no doubt the most important. When local studies (i.e. studies conducted in the same region or city in which the benefits of abatement are being calculated) are available, they are usually used directly in benefit estimation. When no local studies are available, the current practice is to extrapolate results from a similar locality, or to take an average (sometimes weighted) of a given set of studies. However, these methods do no take into account the difference in conditions of the target place from the original study place. In this work we present a meta-analysis of studies of the short term impacts of particulate matter (PM10) impacts on premature mortality, considering explicitly the factors that may influence it. Environmental, demographic, geographical, and economic factors were considered as possible influencing the effect of particulate air pollution on premature mortality. Through a weighted regression model, a meta-analysis of the PM unit risk for a sample of 85 cities around the world was performed, considering as explanatory variables the average concentrations of air pollutants, the monitoring sites density, the mean temperature, the city's surface, the population density, the percentage of population over 65 years old, the average annual mortality rate, and the gross income per capita. Both fixed effects and random effects models were tested. The model selected was a fixed effects model which included population density, average concentration of PM10, and mean temperature. The effect of PM10 concentrations on unit risk was negative, with the slope (multiplied by 1000) decreasing 0.0035 (95%CI = 0.0015, 0.0055) for each μg/m3 of PM10. The effect of population density and temperature were both positive, with an increase of 1 degree Celsius in average temperature having an increase in the coefficient of 0.018 (95% CI = 0.0025, 0.033) The proposed model allows better predictions of the effect of particulate matter on mortality for any city, considering its environmental and demographic attributes.
(1) P. Universidad Catolica De Chile