Seasonal adjustment methods for the analysis of respiratory disease in environmental epidemiology

Bircan Erbas, Rob J Hyndman


We study the relationship between daily hospital admissions for respiratory disease and various pollutant and climatic variables, looking particularly at the effect of seasonal adjustment on the estimated models. Often time series exhibit seasonal behaviour and adequate control for the presence of a seasonal component is essential before one attempts to model the complex pollution-health association. We show that if these factors are not adequately controlled for, spurious effects of pollutants and climate on morbidity/mortality can be induced. We present a method of seasonal adjustment called STL (Seasonal-Trend decomposition based on Loess smoothing), and apply it to pollution and climate data. We will use the seasonally adjusted series in a Generalized Linear Models and Generalized Additive Models analysis of the effects of pollution and climate on hospital admissions for Chronic Obstructive Pulmonary Disease in Melbourne, Australia for the period 1989–1992.