Estimates of the short term effects of air pollution in Italy using alternative modelling techniques

摘要

In the analysis of short term effect of air pollution on health, methods able to control for nonlinear confounding effect of temporal trend are required. We analyze the association between PM10 daily concentrations and Mortality/Hospital Admissions in the Italian Meta-analysis of Short-term effects of Air pollutants (MISA), using alternative modeling techniques: Generalized Additive Models with penalized regression spline fitted by the direct method in R software (GAM-R) and Generalized Linear Models with natural cubic spline (GLM+NS). We find that the two approaches provide similar results. If we are interested in overall estimates and a random effects meta-analysis model is specified, a certain robustness of results to change number of degrees of freedom for the spline is to be expected.