presents summary statistics for study variables. and present cross-sectional life expectancies plotted over air pollution data for the two time periods. At least five observations can be made based on these two figures: 1) PM2.5 concentrations generally declined during the 1980s and 1990s. 2) Life expectancies increased between the two periods. 3) For both periods there were cross-sectional associations between life expectancies and pollution. 4) Similar negative associations were observed when analyses were performed using county or metro level observations individually. 5) There was substantial variability or scatter around the regression line indicating that the association with air pollution explains only part of the cross-sectional variability and there are clearly other important factors that influence life expectancy.
Summary statistics of key study variables in the 217 analysis counties
Figure 2 Cross-sectional life expectancies for 1978–1982 plotted over 1979–1983 PM2.5 concentrations. Dots and number-labeled circles represent county level and metro-level population-weighted mean life expectancies, respectively. Metropolitan (more ...)
Figure 3 Cross-sectional life expectancies for 1997–2001 plotted over 1999–2000 PM2.5 concentrations. Dots and number-labeled circles represent county level and metro-level and population-weighted mean life expectancies, respectively. Metropolitan (more ...)
Estimates of the associations between PM2.5 and life expectancies using cross-sectional regression models were sensitive to inclusion of socio-economic, demographic, and proxy cigarette smoking variables, especially the proportion of high school graduates which was highly correlated with per capita income. For example, the association between PM2.5 and life expectancy was stronger in the less polluted time period without controlling for any covariates. Based on regression models without any covariates, 10μg/m3 higher PM2.5 concentrations were associated with 1.17 (SE = 0.27, p < 0.001) and 2.05 (SE = 0.48, p < 0.001) years lower life expectancy for the 1978–1982 and 1997–2001 periods, respectively. However, models that controlled for income, population, cross-county migration, proportion of the population who were black, Hispanic, or had urban residences, and that included proxy variables for smoking found smaller associations, especially in the second period. An increase of 10μg/m3 in PM2.5 concentrations was associated with a 0.62 (SE = 0.22, p < 0.001) and 0.53 (SE = 0.24, p < 0.05) years lower life expectancy for the 1978–1982 and 1997–2001 periods, respectively.
presents increases in life expectancies plotted over reductions in PM2.5 concentrations between approximately 1980 and 2000. Several additional important observations follow from this ) On average, life expectancy increased more in areas with larger reductions in air pollution. 2) Similar positive associations between life expectancy gains and reductions in PM2.5 concentrations were observed using both county-level and metro-level observations. 3) There was substantial variability or scatter around the regression line, indicating other unaccounted for factors influencing the changes in life expectancy.
Figure 4 Changes in life expectancies plotted over reductions in PM2.5 concentrations. Dots and number-labeled circles represent county level and metro-level and population-weighted mean life expectancies, respectively. Metropolitan area locations and number codes (more ...)
presents regression coefficients between changes in life expectancy and reductions in PM2.5 for models with various combinations of socio-economic, demographic and smoking proxy variables. Models restricted to only counties with a 1986 population ≥100,000, or to only the 51 largest counties in each metropolitan area also are presented. In all models, increased life expectancies were significantly associated with decreases in PM2.5. Based on Model 4, a decrease of 10μg/m3 PM2.5 was associated with an adjusted estimated increase in life expectancy equal to 0.77 (SE = 0.17) years. The estimated effect of reduced PM2.5 on life expectancy was not highly sensitive to controlling for changes in the socio-economic, demographic, or proxy smoking variables or to restricting observations to only large counties.
Table 2 Results of selected regression models including estimates of change in life expectancy (and standard errors) associated with a 10 μg/m3 reduction in PM2.5 controlling for different combinations of socio-economic, demographic and smoking variables (more ...)
In a variety of related sensitivity analyses, the effect estimate for change in PM2.5 was quite robust. In step-wise regressions, change in PM2.5 was generally the third variable to enter the model, following change in per capita income and change in COPD, and was stable to the inclusion of other variables. When models 4 and 7 in were re-estimated using weighted regression (weighting by the square root of the two-period average population), similar results were observed with a decrease of 10μg/m3 PM2.5 associated with an estimated increase in life expectancy equal to 0.66 (SE = 0.18) and 0.87 (SE = 0.24) years, respectively. Stratified estimates of Model 4 in were estimated using the 45 counties in the 15 least polluted metro areas for the early period (PM2.5 < 17μg/m3, see ) versus all other more polluted areas. A reduction of tenμg/m3 PM2.5 was associated with a 0.99 (SE = 0.42, p < 0.05) and 0.80 (SE = 0.22, p < 0.01) years increase in life expectancy for the least polluted versus other areas, respectively, finding no statistically significant differential pollution effects for the initially low polluted versus high polluted areas.
The effect estimate for change in PM2.5 was also not highly sensitive to the inclusion of survey-based estimates of metro-level changes in cigarette smoking. For example, when Model 4 in was re-estimated using data from the 140 counties in the 24 metro areas with matching smoking prevalence data, a reduction of ten μg/m3 PM2.5 was associated with an estimated increase in life expectancy equal to 0.80 (SE = 0.19, p < 0.05) and 0.83 (SE = 0.20, p < 0.05) years without and with the inclusion of the change in smoking prevalence variable, respectively. When model 7 in was re-estimated using only data from the 24 largest counties in the 24 metro areas with matching smoking prevalence data, a reduction of ten μg/m3 PM2.5 was associated with an estimated increase in life expectancy equal to 1.00 (SE = 0.29, p < 0.05) and 1.06 (SE = 0.30, p < 0.05) years without and with the inclusion of the change in smoking prevalence variable, respectively. When added to these models, change in smoking prevalence was not statistically significant (p > 0.15) and the estimated effect of a change in COPD death rate was largely unaffected. These results indicate that county-level changes in COPD were more robustly associated with county-level changes in life expectancy than metro-level estimates of changes in smoking based on limited survey data.