Carbon Monoxide and Risk of Hospitalization Due to COPD
Carbon Monoxide and Risk of Hospitalization Due to COPD
Table 1 provides descriptive statistics on pollution, weather, and hospitalization data. Carbon monoxide concentrations were low during the study period, with a daily average of 0.6 ppm and a 1-hour maximum of 0.8 ppm in background stations; the WHO 8-hour air quality guideline for carbon monoxide is 8.7 ppm. Carbon monoxide concentrations were moderately correlated with both nitrogen dioxide and PM2.5 levels; the correlation coefficients were 0.57 and 0.59, respectively. From January 1, 2001, to December 31, 2007, a total of 117,329 hospital admissions for COPD through emergency departments were recorded, with a daily average of 57 admissions. The mean age of the COPD patients was 80.7 years for female subjects and 74.2 years for male subjects.
Table 2 shows the percentage change in total emergency hospital admissions for COPD per each interquartile-range increment increase in pollutant concentration at lag 0, 1, 2 and 02 days in single-pollutant and 2-pollutant models. The former 3 were risk estimates for the single days of lag 0, 1, and 2, respectively, whereas lag 02 was an estimate of the association of cumulative exposure of from lag 0 to lag 2 days with hospitalization for COPD. Both nitrogen dioxide and PM2.5 were associated with higher risks of COPD hospitalizations. The risk estimates for carbon monoxide, however, were negative for all 3 single-day lags, and the lag 1 day carbon monoxide was associated with the largest COPD risk reduction. For the cumulative risk estimates of the distributed lag of 0–2 days, one interquartile-range increment increase in background carbon monoxide (0.4 ppm) was associated with −1.8% (95% confidence interval: −3.1, −0.4) change in COPD hospitalizations according to the single-pollutant model; the negative association became stronger when we adjusted for nitrogen dioxide or PM2.5 in the 2-pollutant models. Sensitivity analysis found that the risk estimates for carbon monoxide were largely robust to the degree of adjustment for seasonality and trend, model specifications for weather variables, the lag of carbon monoxide exposure, and the usage of other carbon monoxide measurement (1-hour maximum; Table 3). The results from these a priori models were also comparable to those from case-crossover analyses and those from generalized additive models chosen by minimizing the partial autocorrelation function of residuals.
Exposure to the 2 copollutants (PM2.5 and nitrogen dioxide) was associated with an increased risk of COPD hospitalization. One interquartile-range increment increase in PM2.5 (31.1 µg/m) corresponded to a 5.4% (95% confidence interval: 3.9, 6.9) increase in COPD hospitalizations according to the single-pollutant model; the risk estimate was robust to adjustment for carbon monoxide or nitrogen dioxide. One interquartile-range increment increase in nitrogen dioxide (24.2 µg/m) corresponded to a 4.2% (95% confidence interval: 2.7, 5.7) increase in COPD hospitalizations according to the single-pollutant model; the risk estimate was robust to adjustment for carbon monoxide but was attenuated to null after adjustment for PM2.5.
Figure 2 shows the percentage change in emergency hospital admissions for COPD per interquartile-range increase in pollutant concentrations at distributed lags of 0–2 days in single-pollutant and 2-pollutant models. The risk estimates were similar among female and male subjects, although the uncertainties of the risk estimates were larger for female subjects than for male subjects; one underlying factor could be the smaller number of COPD admissions among female subjects (11 cases per day on average) compared with male subjects (46 per day). Negative associations were found between carbon monoxide concentrations and COPD; the negative associations became stronger when we adjusted for PM2.5 or nitrogen dioxide. In both sexes, exposure to the 2 copollutants was associated with increased risks of COPD hospitalizations except that the risk estimates for nitrogen dioxide became statistically insignificant after adjustment for PM2.5 levels.
(Enlarge Image)
Figure 2.
Percentage change in emergency hospital admissions for chronic obstructive pulmonary disease per interquartile-range increment increase in pollutant concentrations at distributed lags of 0–2 days in single-pollutant and 2-pollutant models, Hong Kong, China, 2001–2007. The results for the 2 sexes were compared. The points indicate central estimates. CO, carbon monoxide; NO2, nitrogen dioxide; PM2.5, particulate matter with aerodynamic diameter less than 2.5 μm. Bars, 95% confidence intervals.
Figure 3 shows the second-degree polynomial exposure-response curves for daily average pollutant concentration at distributed lags of 0–2 days and risk of emergency hospital admissions for COPD in single-pollutant and 2-pollutant models. The exposure-response curves for carbon monoxide and COPD hospitalizations were approximately linear in all the single-pollutant and 2-pollutant models; the downward slope was steeper when we adjusted for PM2.5 or nitrogen dioxide. The exposure-response curves for PM2.5 and COPD hospitalizations were linear in all the single-pollutant and 2-pollutant models; the upward slope appeared steeper when we adjusted for carbon monoxide. The upward slope for the exposure-response curve of nitrogen dioxide and COPD hospitalizations was substantially reduced when we adjusted for PM2.5.
(Enlarge Image)
Figure 3.
Second-degree polynomial exposure-response curves for daily average pollutant concentrations at distributed lags of 0–2 days and risk of emergency hospital admissions for chronic obstructive pulmonary disease in single-pollutant and 2-pollutant models, Hong Kong, China, 2001–2007. The solid line represents central estimates; the dashed lines represent 95% confidence intervals. CO, carbon monoxide; NO2, nitrogen dioxide; PM2.5, particulate matter with aerodynamic diameter less than 2.5 μm.
Results
Table 1 provides descriptive statistics on pollution, weather, and hospitalization data. Carbon monoxide concentrations were low during the study period, with a daily average of 0.6 ppm and a 1-hour maximum of 0.8 ppm in background stations; the WHO 8-hour air quality guideline for carbon monoxide is 8.7 ppm. Carbon monoxide concentrations were moderately correlated with both nitrogen dioxide and PM2.5 levels; the correlation coefficients were 0.57 and 0.59, respectively. From January 1, 2001, to December 31, 2007, a total of 117,329 hospital admissions for COPD through emergency departments were recorded, with a daily average of 57 admissions. The mean age of the COPD patients was 80.7 years for female subjects and 74.2 years for male subjects.
Table 2 shows the percentage change in total emergency hospital admissions for COPD per each interquartile-range increment increase in pollutant concentration at lag 0, 1, 2 and 02 days in single-pollutant and 2-pollutant models. The former 3 were risk estimates for the single days of lag 0, 1, and 2, respectively, whereas lag 02 was an estimate of the association of cumulative exposure of from lag 0 to lag 2 days with hospitalization for COPD. Both nitrogen dioxide and PM2.5 were associated with higher risks of COPD hospitalizations. The risk estimates for carbon monoxide, however, were negative for all 3 single-day lags, and the lag 1 day carbon monoxide was associated with the largest COPD risk reduction. For the cumulative risk estimates of the distributed lag of 0–2 days, one interquartile-range increment increase in background carbon monoxide (0.4 ppm) was associated with −1.8% (95% confidence interval: −3.1, −0.4) change in COPD hospitalizations according to the single-pollutant model; the negative association became stronger when we adjusted for nitrogen dioxide or PM2.5 in the 2-pollutant models. Sensitivity analysis found that the risk estimates for carbon monoxide were largely robust to the degree of adjustment for seasonality and trend, model specifications for weather variables, the lag of carbon monoxide exposure, and the usage of other carbon monoxide measurement (1-hour maximum; Table 3). The results from these a priori models were also comparable to those from case-crossover analyses and those from generalized additive models chosen by minimizing the partial autocorrelation function of residuals.
Exposure to the 2 copollutants (PM2.5 and nitrogen dioxide) was associated with an increased risk of COPD hospitalization. One interquartile-range increment increase in PM2.5 (31.1 µg/m) corresponded to a 5.4% (95% confidence interval: 3.9, 6.9) increase in COPD hospitalizations according to the single-pollutant model; the risk estimate was robust to adjustment for carbon monoxide or nitrogen dioxide. One interquartile-range increment increase in nitrogen dioxide (24.2 µg/m) corresponded to a 4.2% (95% confidence interval: 2.7, 5.7) increase in COPD hospitalizations according to the single-pollutant model; the risk estimate was robust to adjustment for carbon monoxide but was attenuated to null after adjustment for PM2.5.
Figure 2 shows the percentage change in emergency hospital admissions for COPD per interquartile-range increase in pollutant concentrations at distributed lags of 0–2 days in single-pollutant and 2-pollutant models. The risk estimates were similar among female and male subjects, although the uncertainties of the risk estimates were larger for female subjects than for male subjects; one underlying factor could be the smaller number of COPD admissions among female subjects (11 cases per day on average) compared with male subjects (46 per day). Negative associations were found between carbon monoxide concentrations and COPD; the negative associations became stronger when we adjusted for PM2.5 or nitrogen dioxide. In both sexes, exposure to the 2 copollutants was associated with increased risks of COPD hospitalizations except that the risk estimates for nitrogen dioxide became statistically insignificant after adjustment for PM2.5 levels.
(Enlarge Image)
Figure 2.
Percentage change in emergency hospital admissions for chronic obstructive pulmonary disease per interquartile-range increment increase in pollutant concentrations at distributed lags of 0–2 days in single-pollutant and 2-pollutant models, Hong Kong, China, 2001–2007. The results for the 2 sexes were compared. The points indicate central estimates. CO, carbon monoxide; NO2, nitrogen dioxide; PM2.5, particulate matter with aerodynamic diameter less than 2.5 μm. Bars, 95% confidence intervals.
Figure 3 shows the second-degree polynomial exposure-response curves for daily average pollutant concentration at distributed lags of 0–2 days and risk of emergency hospital admissions for COPD in single-pollutant and 2-pollutant models. The exposure-response curves for carbon monoxide and COPD hospitalizations were approximately linear in all the single-pollutant and 2-pollutant models; the downward slope was steeper when we adjusted for PM2.5 or nitrogen dioxide. The exposure-response curves for PM2.5 and COPD hospitalizations were linear in all the single-pollutant and 2-pollutant models; the upward slope appeared steeper when we adjusted for carbon monoxide. The upward slope for the exposure-response curve of nitrogen dioxide and COPD hospitalizations was substantially reduced when we adjusted for PM2.5.
(Enlarge Image)
Figure 3.
Second-degree polynomial exposure-response curves for daily average pollutant concentrations at distributed lags of 0–2 days and risk of emergency hospital admissions for chronic obstructive pulmonary disease in single-pollutant and 2-pollutant models, Hong Kong, China, 2001–2007. The solid line represents central estimates; the dashed lines represent 95% confidence intervals. CO, carbon monoxide; NO2, nitrogen dioxide; PM2.5, particulate matter with aerodynamic diameter less than 2.5 μm.
