Air pollution and hospital admissions for respiratory disease

Several recent studies have reported associations between short-term changes in air pollution and respiratory hospital admissions. Most of those studies analyzed locations where there was a high correlation between airborne particles and sulfur dioxide (SO2), and between all air pollutants and temperature. Here, I seek to replicate the previous findings in a location where SO2 concentrations were trivial, and the correlation between both airborne particles and ozone with temperature was considerably lower than in previous studies. I constructed daily counts of admissions to all hospitals in Spokane, WA, for respiratory disease (International Classification of Diseases, 9th revision, codes 460-519) for persons age 65 years and older. I computed average daily concentrations of airborne particles whose diameter is 10 mu m or less (PM(10)) and ozone (O-3) from all monitors in each city, and I obtained daily average temperature and humidity from the U.S. weather service. SO2 concentrations in Spokane were so low that monitoring was discontinued. I regressed daily respiratory admission counts on temperature, humidity, day of the week indicators, and air pollution. I used a Poisson regression analysis and removed long wavelength patterns using a nonparametric smooth function of day of study. I dealt with a possible U-shaped dependence of admissions on temperature and/or humidity by using nonparametric smooth functions of weather variables as well. I then examined sensitivity analyses to control for weather. Both PM(10) and ozone were associated with increased risk of respiratory hospital admissions [relative risk (RR) = 1.085; 95% confidence interval (CI) = 1.036-1.136 for a 50-mu g per m(3) increase in PM(10), and RR = 1.244; 95% CI = 1.002-1.544 for a 50-mu g per m(3) increase in peak hour ozone]. The PM(10) association was insensitive to alternative methods of control for weather, including exclusion of extreme temperature days and control for temperature on multiple days. The ozone results were more sensitive to the approach for weather control. The magnitude of the PM(10) effect in this location, where SO2 was essentially not present, and where the correlation between PM(10) and temperature was close to zero, was similar to that reported in other locations in the eastern United States and Europe, where confounding by weather and SO2 is a more substantial concern.