[Methods of statistical analysis to evaluate the short term effects of air pollution in the EpiAir Project]

The EpiAir Project, which is funded by the Italian Center for disease prevention and control, Ministry of Health, involves 10 Italian cities for the period 2001-2005, and has the following objectives: to monitor the environmental characteristics associated with relevant health effects; to systematically collect data on relevant health outcomes, such as mortality and hospitalizations, for which the impact of air pollution has been established; to report estimates of the health effects attributable to air pollution; to identify the subgroups of the population most vulnerable to the effects of air pollution. An important part of the project deals with estimating the association between several air pollutants (PM10, NO2, O3 ) and adverse health outcomes, such as mortality for a broad spectrum of causes and hospital admissions for single disease diagnoses or diseases groups. The analyses are implemented at the city level, and then pooled estimates are calculated with random-effects metaanalyses. The present paper provides details on the statistical methodology applied, focusing on: the definition of and control for the confounding factors in the city-specific analyses, estimating the association between health outcomes and air pollution at different temporal latencies (lags), identifying population subgroups most vulnerable to PM10, and analysing the heterogeneity of the city-specific results.