Summary:
In recent years, much attention has been given to the human health effects of atmospheric pollutants, especially particulate matter. This has been the focus of particularly heated debate in the USA, as new regulations introduced in 1997 by the United States Environmental Protection Agency (USEPA) have considerably tightened the existing standard. Similar regulations are also being considered by several countries of the European Union. Much of the debate revolves around claims that particulate matter in the atmosphere directly influences mortality, hospital admissions with respiratory diseases, and so on. In this chapter, we take a critical look at one of these issues, the influence of PM10 (particulate matter of aerodynamic diameter no more than 10 µm) on deaths in the elderly population. Two data sets are considered, one from Birmingham, Alabama, and the other from Chicago. In both cases we find a significant PM10-mortality relationship in some of the models fitted, but not in others. Other issues considered include the existence of a threshold below which PM10 has no discernable influence, the interaction with other pollutants, and the mortality displacement or harvesting effect ( the theory that the direct effect of PM10 is limited to a very small subset of the population who are already critically ill and whose death is only advanced by a few hours or days as a result of air pollution). For the latter phenomenon, a compartment-type model is introduced and analyzed using a Markov chain Monte Carlo procedure. The results show that even when all these alternative effects are considered, there remains a considerable amount of unexplained association between particulates and mortality, but there appear to be too many uncertain issues to allow us to make definitive statements about a causal relationship.
Keywords:
Bayesian inference, Generalized additive model, Harvesting phenomenon, Influence of other atmospheric pollutants, Linear regression, Markov chain Monte Carlo
algorithms, Mortality displacement, Non-linear relationships, PM10, Poisson regression, Variable selection.
tr74.pdf