engleski

Air pollution and the emergency room admissions for chronic obstructive pulmonary disease in Zagreb, Croatia

Short-term effects of air pollution on aggravation of symptoms in adult persons with chronic obstructive pulmonary disease (COPD, ICD9 codes 490, 491, 492 and 496) were analysed for the period 1995-1997. Daily counts of emergency hospital visits due to aggravation of symptoms of chronic obstructive pulmonary disease were obtained from the emergency room registries of three hospitals in Zagreb. Average daily concentrations of black smoke and SO2 (measured at 4 sites) and NO2 (measured at 6 sites) were collected at the Institute for Medical Research and Occupational Health. Data on NO2 from 2 additional sites, average daily temperature, relative humidity and air pressure were obtained from the Meteorological and Hydrological Service of Croatia. Data on influenza epidemics were obtained from the official epidemiological reports (Epidemiološki vjesnik 1/95-12/97). Counts of emergency room visits were analysed using a generalised additive model (Hastie and Tibshirani, 1990) with logarithmic link and overdispersed Poisson variance. Additive components were modelled as cubic splines. The building of the model consisted of adding terms for holidays, day of the week, seasonallity, weather variables, influenza and finally the air pollution variables. The functional form of each term was assessed by comparing a modified AIC criterion for candidate models (the deviance was penalised by adding the effective degrees of freedom multiplied by the factor of 3 and the overdispersion parameter). The best term for each predictor was kept in the model disregarding the statistical significance. Meteorological, air pollution and influenza variables were lagged by up to 3 (for influenza 14) days. Serial autocorrelation was examined using autocorrelation and partial autocorrelation function for the deviance residuals of the final model. All analyses were performed with S-plus 2000 Professional running under Win98. Mean count of emergency visits during the analysed period was 3.5 (variance 6.5). Temperature ranged from -12.2 to 26.6°C (median 12.0°C). Relative humidity was between 35.1 and 100.7% (median 74.6%). Air pressure ranged from 979 to 1022mmHg (median 1002mmHg). Concentrations of SO2 and black smoke were well within the recommended annual mean and 98th percentile values. Mean NO2 concentration was below the recommended value, but its 98th percentile (63.3 ľgm-3) barely surpassed the recommended value of 60 ľgm-3. Relative humidity entered the model lagged by 1 day, air pressure, and SO2 were lagged by 2 days, and temperature, NO2, and black smoke by 3 days. Seasonallity was modelled using a cubic spline of the day since the beginning of the study period with 14 degrees of freedom. In addition to seasonallity, only air pressure and NO2 showed significant departures from linearity. The additive component for air pressure was U shaped, with optimal value (the least COPD risk) around 1005 mmHg. The additive component for NO2 showed a hockey-stick form with apparent threshold around 50 ľgm-3. Effects of NO2 and black smoke were significant (significance of the ?2 test for the change in deviance was p=0.0008 for NO2 and p=0.0402 for black smoke). Effect of SO2 was not significant (p=1.0000). Relative risk associated with the increase in NO2 from the 5th to the 99th percentile (i.e. 13.7 to 70.3 ľgm-3) was 1.146 (95% CI: 0.942-1.393). Relative risk associated with a 50 ľgm-3 increase in black smoke was 1.174 (95% CI: 1.011-1.364). The final model showed considerable overdispersion (equal 1.364). Maximum absolute autocorrelation and partial autocorrelation of residuals was 0.082.

nitrogen dioxide; sulphur dioxide; black smoke; COPD; emergency hospital visits; GAM

nije evidentirano