The relationship of airborne particulates below 10 micrometers in aerodynamic diameter (PM10) to mortality is critical for public health and regulatory decisions. To characterize this relationship we analyzed longitudinal data air pollution and mortality data for the 88 largest U.S. cities during 1987-1994, to determine how daily mortality varied with PM10.
At the first stage, we used log-linear models with non-parametric adjustments for weather variables as well as longer-term trends, reflecting seasonal variation and influenza epidemics, changes in behavior and medical practice, and other sources of auto-correlation. The logarithm of the expected value of daily mortality was modeled as a function of air pollution using natural cubic splines. At the second stage, spatial models were used to investigate heterogeneity of mortality, and of the shapes of PM10 mortality dose-response curves, across cities and regions. Using two-stage hierarchical linear regression, we examined dependence of mortality on mean pollution levels, demographic variables, reliability of pollution data, and specific constituents of particulate-matter mixtures.
Previous day PM10 concentrations were positively associated with total mortality in most locations, with a 0.4 percent rise in death rate per 10 unit increase in PM10. The Northeast region mortality increase was twice as high as the average for the other cities. For the nation overall, the pooled concentration-response relationship was linear, with little evidence for deviation from linearity down to the lowest levels. These results have significant regulatory implications. The modeling approaches can also be extended to other environmental pollution problems.
See biosun01.biostat.jhsph.edu/~fdominic/research.htmlLearning Objectives: At the conclusion of the session, the participant (learner) in this session will be able to: 1. Describe the use of loglinear modeling with natural cubic splines to relate relative mortality rates to particulate pollution levels using longitudinal data from the 88 largest U.S. cities, adjusting for weather and other covariates. 2. Describe the use of a spatial regression modeling to study geographical variation in relative mortality rates, shapes of dose-response curves, and the dependence of relative mortality on multiple measures of pollution and demography. 3. Discuss circumstances in which these statistical methods would, and would not, be appropriate for analyses of other environmental pollution problems.
Keywords: Biostatistics, Air Pollutants
Presenting author's disclosure statement:
Organization/institution whose products or services will be discussed: None
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.