1664

Aim: Air pollutants may negatively affect a large spectrum of human health conditions. Among the statistical methods used to assess the impact of concentrations of ambient air pollutants on health are the time-series and case-crossover methods. Traditionally, both methods estimate risk in linear forms. Recent studies indicate that non-linear representations are more adequate and accurate. Methods: We investigate the association between daily emergency department (ED) visits for all respiratory conditions and daily concentration of ambient ozone. We retrieve ED data for Edmonton, Canada, for the period April 1992 - March, 2002, (3,652 days) using the ICD-9 codes 460-519. The case-crossover technique is realized as a conditional Poisson regression. We test lags from 0 to 8 days and determine 9 concentration-response (c-r) shapes. Results: 292,285 ED visits were identified and used in the study. Lagged exposures result in positive and statistically significant associations between ozone concentration and ED visits for respiratory conditions. The concentration-response shapes change according to the lags considered. Various forms of c-r shapes are observed, such as an almost linear shape, an S-profile, and others. The amalgamated by used lags c-r function provides a good tool to summarize the risk relating to the concentration of the considered air pollutant. In general, the results are non-linear parametric risk functions of the concentrations. Conclusion: The estimated functions allow the identification of concentrations with various intensities of impact on health. In some cases, these functions indicate a potential threshold concentration.