Associations between health and air pollution in time-series analyses.

The objective of this study was to determine the contribution of cyclic geophysical and sociocultural factors and of meterological and pollutant variations to explained variance in asthma admissions. A Poisson regression-type quasi-Fourier generalized linear model was built. We constructed an extended family of models to study the contribution of each variable. Daily observations of children's admissions for asthma were conducted during 2 y in hospitals in Melbourne, Australia. Pollutant data (i.e., fine particles, sulfur dioxide, nitrogen dioxide, and ozone) and meterological data (i.e., maximum and minimum temperatures, mean humidity, and mean barometric pressure) were collected. The result of the final model accounted for 57% of the variance in admissions. The contribution of pollution, with all periodic patterns estimated by Poisson ANOVA, was 14%, even though no pollutant alone made a significant contribution to explained variance. Autocorrelation analysis of residuals showed that the model accounted reasonably well for autocorrelation effects.