Daily mortality and air pollution in Atlanta: two years of data from ARIES.

Associations between daily mortality and air pollution were investigated in Fulton and DeKalb Counties, Georgia, for the 2-yr period beginning in August 1998, as part of the Aerosol Research and Inhalation Epidemiological Study (ARIES). Mortality data were obtained directly from county offices of vital records. Air quality data were obtained from a dedicated research site in central Atlanta; 15 separate air quality indicators (AQIs) were selected from the 70 particulate and gaseous air quality parameters archived in the ARIES ambient air quality database. Daily meteorological parameters, comprising 24-h average temperatures and dewpoints, were obtained from Atlanta's Hartsfield International Airport. Effects were estimated using Poisson regression with daily deaths as the response variable and time, meteorology, AQI, and days of the week as predictor variables. AQI variables entered the model in a linear fashion, while all other continuous predictor variables were smoothed via natural cubic splines using the generalized linear model (GLM) framework in S-PLUS. Knots were spaced either quarterly, monthly, or biweekly for temporal smoothing. A default model using monthly knots and AQIs averaged for lags 0 and 1 was postulated, with other models considered in sensitivity analyses. Lags up to 5 days were considered, and multipollutant models were evaluated, taking care to avoid overlapping (and thus collinear) AQIs. For this reason, PM(2.5) was partitioned into its three major constituents: SO(2-)(4), carbon (EC + 1.4 OC), and the remainder; sulfate was assumed to be (NH(4))(2)SO(4) for this purpose. Initial AQI screening was based on all-cause (ICD-9 codes <800) mortality for those aged 65 and over. For the (apparently) most important pollutants--PM(2.5) and its 3 major constituents, coarse PM mass [CM], 1-h maximum CO, 8-h maximum O(3)--we investigated 15 mortality categories in detail. (The 15 categories result from three age groups [all ages, <65, 65+] and five cause-of-death groups [all disease causes, cardiovascular, respiratory, cancer, and other "remainder" disease causes]). The GLM model outputs that were considered included mean AQI effects and their standard errors, and two indicators of relative model performance (deviance and deviance adjusted for the number of observations and model parameters). The latter indicator was considered to account for variations in the number of observations created by varying amounts of missing AQI data, which were not imputed. The single-AQI screening regressions on all-cause 65+ mortality show that CO, NO(2), PM(2.5), CM, SO(2), and O(3), followed by EC and OC, consistently have the best model fits, after adjusting for the number of observations. Their relative rankings, however, vary according to the smoothing knots used, and there is no correspondence between mean AQI effect and overall model fit.(Other regression runs often show that the best model fits are obtained with no AQI in the model.) There is no correspondence between mean AQI effect and statistical significance or between mean effect and serial correlation. There is a highly significant (.001 level) relationship between overall model fit and serial correlation; the best fitting models have the most frequent knot spacing and the most negative serial correlation. The regression analyses by cause of death find elderly circulatory deaths to be consistently associated with CO for all models.

Aerosol Research and Inhalation Epidemiological Study (ARIES): air quality and daily mortality statistical modeling--interim results.

The Aerosol Research and Inhalation Epidemiological Study (ARIES) is an EPRI-sponsored project to collect air quality and meteorological data at a single site in northwestern Atlanta, GA. Seventy high-resolution air quality indicators (AQIs) are used to examine statistical relationships between air quality and health outcome end points. Contemporaneous mortality data are collected for Fulton and DeKalb counties in Georgia. Currently, 12 months of air quality and weather data are available for analysis, from August 1998 through July 1999. The interim mortality analysis used Poisson regression in generalized additive models (GAMs). The estimated log-linear association of mortality with various AQIs was adjusted for smoothed functions of time and meteorological data. The analysis considered daily deaths due to all nonaccidental causes, deaths to persons 65 years or older, and deaths in each of the two constituent counties. The fine particle effect associated with the four mortality subgroups, using only today (lag 0), yesterday (lag 1), 2-day average (average of today and yesterday), and first difference (today minus yesterday) measurements of the air quality relative to today's number of deaths was positive for lag 0, lag 1, and 2-day average and positive only for decedents at least 65 years of age using first difference. The t values ranged from 0.81 to 1.15 for lag 0, 1.04 to 1.53 for lag 1, 1.10 to 1.66 for 2-day average, and -0.32 to 0.33 for first difference with 346 or 347 days of data. No statistically significant estimate of the linear coefficient was found for the other 14 air quality variables in our interim analysis for the four mortality subgroups. We discuss diagnostics to support these models. These interim analyses did not include an evaluation of sensitivity to a larger set of lag structures, nonlinear model specifications, multipollutant analyses, alternative weather model and smoothing model specifications, air pollution imputation schemes, or cause-specific mortality indicators, nor did they include a full reporting of model selection or goodness-of-fit indicators. No conclusion can be drawn at this time about whether the findings from subsequent studies have sufficiently greater power to detect effects comparable to those found in other U.S. cities including at least 2 or 3 years of data.

Is daily mortality associated specifically with fine particles? Data reconstruction and replication of analyses.

In 1996, Schwartz, Dockery, and Neas reported that daily mortality was more strongly associated with concentrations of PM2.5 than with concentrations of larger particles (coarse mass [CM]) in six U.S. cities ("original paper"/"original analyses"). Because of the public policy implications of the findings and the uniqueness of the concentration data, we undertook a reanalysis of these results. This paper presents results of the reconstruction of these data and replication of the original analyses using the reconstructed data. The original investigators provided particulate air pollution data for this paper. Daily weather and daily counts of total and cause-specific deaths were reconstructed from original public records. The reconstructed particulate air pollution and weather data were consistent with the summaries presented in the original paper. Daily counts of deaths in the reconstructed data set were lower than in the original paper because of restrictions on residence and place of death. The reconstruction process identified an administrative change in county codes that led to higher numbers of deaths in St. Louis. Despite these differences in daily counts of deaths, the estimated effects of particulate air pollution from the reconstructed dataset, using analytic methods as described in the original paper, produced combined effect estimates essentially equivalent to the originally published results. For example, the estimated association of a 10 micrograms/m3 increase in 2-day mean particulate air pollution on total mortality was 1.3% (95% confidence interval [CI] 0.9-1.7%, t = 6.53) for PM2.5 based on the reconstructed dataset, compared to the originally reported association of 1.5% (95% CI 1.1-1.9%, t = 7.41). For coarse particles, the estimated association from the reconstructed dataset was 0.4% (95% CI -0.2-0.9%, t = 1.43) compared to the originally reported association of 0.4% (95% CI -0.1-1.0%, t = 1.48). These results from the reconstructed data suggest that the original results reported by Schwartz, Dockery, and Neas were essentially replicated.