Table 3.
Results of second-stage regression models, investigating the role of potential modifiersa of the estimated effectsb of CO on mortality.
| Total mortality
|
CVD mortality
|
|||||||
|---|---|---|---|---|---|---|---|---|
| 8 df/year
|
PACF
|
8 df/year
|
PACF
|
|||||
| Effect modifierc | 25thd | 75thd | 25thd | 75thd | 25thd | 75thd | 25thd | 75thd |
| No. of CO monitors | 0.71 (0.48 to 0.94) | 0.54 (0.34 to 0.74) | 1.18 (0.96 to 1.39) | 0.92 (0.73 to 1.11) | ||||
| Mean PM10 levels | 0.37 (0.08 to 0.66) | 0.49 (0.28 to 0.69) | 0.74 (0.46 to 1.02) | 1.07 (0.87 to 1.27) | ||||
| Mean O3 | 1.04 (0.67 to 1.41) | 0.82 (0.55 to 1.10) | 1.32 (0.96 to 1.68) | 1.09 (0.83 to 1.35) | ||||
| SMR | 0.79 (0.55 to 1.03) | 0.44 (0.22 to 0.66) | 1.29 (1.06 to 1.52) | 0.77 (0.56 to 0.98) | 1.06 (0.71 to 1.42) | 0.61 (0.30 to 0.93) | 1.40 (1.06 to 1.75) | 0.85 (0.55 to 1.14) |
| Population > 75 years of age (%) | 0.58 (0.25 to 0.92) | 0.94 (0.64 to 1.24) | 0.74 (0.41 to 1.06) | 1.25 (0.96 to 1.54) | ||||
| Geographic region | ||||||||
| Western cities | 0.75 (0.47 to 1.03) | 1.15 (0.90 to 1.40) | 1.06 (0.67 to 1.46) | 1.38 (1.00 to 1.76) | ||||
| Southern cities | 0.61 (0.32 to 0.91) | 1.08 (0.79 to 1.38) | 0.70 (0.26 to 1.14) | 0.90 (0.47 to 1.33) | ||||
| Eastern cities | 0.03 (−0.47 to 0.53) | 0.27 (−0.20 to 0.74) | 0.21 (−0.48 to 0.90) | 0.48 (−0.14 to 1.11) | ||||
SMR, standardized mortality rate.
Variables characterizing each city; only statistically significant effect modifiers under both methods for seasonality control, reducing the heterogeneity by > 10% are presented.
Effect estimates used from first-stage models are based on the PACF criterion and on 8 df/year for seasonality control.
The effect modifiers were included alternatively in the model.
Percent increase in mortality per 1-mg/m3 increase in the 2-day mean CO levels, estimated using fixed-effects model, for a city with levels of the corresponding effect modifier equal to the 25th and the 75th percentiles of its distribution.