Assessing short-term impact of PM10 on mortality using a semiparametric generalized propensity score approach

摘要

In investigating short-term effects and impacts of air pollution on population mortality and morbidity, the exposure-response relationship is frequently assumed to be linear on a logarithmic scale, supporting the idea that an increase of the air pollution level yields the same percent increase in the occurrence of health events at any exposure level. This choice is in line with part of the epidemiological literature which does not report evidences of strong deviations from log-linearity within the ranges of exposure levels observed in urban areas [ – ]. However, recent works suggest a possible violation of the log-linearity assumption, especially in contexts where the range of exposure values was sufficiently wide to allow exploring the curve also for high air pollutant levels. For example, in a study aimed at investigating the relationship between the concentration of atmospheric particulate matter having diameter less than or equal to 2.5 m (PM ) and mortality in 272 Chinese cities, a leveling off in the exposure-response curves at high concentrations was reported in some areas of the Country [ ]. Another Chinese study found a non log-linear increase in the occurrence of respiratory diseases for daily PM levels above 50 g/m [ ]. Evidences of violation of the log-linearity assumption were found also in Li et al. [ ] for the relationship between PM and mortality and in Zu et al. [ ] for the relationship between ozone and hospitalizations from asthma. Finally, in the recent study conducted by Liu and colleagues on 652 cities around the world, the estimated pooled exposure-response functions between daily average levels of PM and PM and all-cause mortality showed no thresholds and seemed to flatten at high air pollutants concentrations [ ].