The effect of air pollution & climate on hospital admissions for chronic obstructive airways disease: a non-parametric alternative

Medical researchers have discovered an alarming increase in rates of respiratory disease in many countries. Exposure to environmental factors such as air pollution, ozone and pollen seem to be possible causes. In recent years several studies have shown a short term association between exposure to air borne particles and ozone and an increase in respiratory hospital admissions. The aim of this paper is to examine methodological issues in epidemiological time series studies involving count data. As a vehicle of illustration, we consider the effect of air pollution and climate on daily counts of hospital admissions for chronic obstructive airways disease. We apply non-parametric (Generalized Additive) models and parametric (Generalized Linear) models, and examine the relationship between hospital admissions for chronic obstructive airways disease and the short term exposure to pollutants. We show that the Generalized Additive Model is superior to the Generalized Linear Model when capturing the non-linear pattern of the response variable with a number of the covariates. One problem in epidemiological time series studies is incorporating autocorrelation in the statistical models. We show that correct specification of the non-linear functional form, and allowing for seasonality, will in some cases remove the autocorrelation inherent in morbidity/mortality data. We also present an alternative residual analysis when the response is not normally distributed. Department of General Practice & Public Health, The University of Melbourne, VIC 3010, Australia. Department of Econometrics and Business Statistics, Monash University, VIC 3800, Australia.