A Population-Dynamic Model of Daily Mortality and Air Pollution in Philadelphia Incorporating Precedent Frailty

C. Murray (1), F. W. Lipfert (2), R. E. Wyzga (3)

(1) U. of Houston, Houston, TX (2) Consultant, Northport, NY (3) Sr. Technical Executive, Electric Power Research Institute, Palo Alto, CA

     Abstract Number: 282
     Last modified: November 9, 2009

Abstract
Numerous time-series studies have found associations between ambient air pollution and daily mortality and various indicators of morbidity. While these studies have been able to identify very small effects by using large samples, they have been less successful in identifying the specific pollutants involved or in providing a rational basis for effective public health policies. This paper is directed to the public health context of air pollution effects on daily mortality, specifically to the question of life expectancy and the degree of prematurity of deaths associated with transient increases in air pollution, an issue referred to as mortality displacement or “harvesting.” State-space modeling is used to build on a previous analysis of elderly (ages 65+) daily mortality in Philadelphia from 1974-88, by estimating the effects of environmental factors associated with entry into the pool of frail individuals who are at high risk of imminent death. The size of this pool cannot be observed but is estimated using Kalman filtering. The environmental factors considered are ozone, various functions of ambient temperature, and total suspended particulate matter (TSP), which is considered here as an index of all types of airborne particles.
The new results show that this population at extreme risk fluctuates with season and other factors but averages less than 0.1% of the elderly in Philadelphia. TSP, ozone and temperature are among the factors associated with transition from a (presumably) healthy state to entry into this small high-risk pool. By considering successively longer moving averages of TSP, we show that cumulative short-term effects on entry to the at-risk pool tend to level off and decrease as periods of exposure longer than a few days are considered. This trend is consistent with the hypothesis that only acute mechanisms are relevant to entry to the at-risk pool. The magnitudes of environmental effects on elderly mortality are consistent with previous Philadelphia analyses based on conventional time-series methods. Relationships of the environment with entry to the at-risk pool have not previously been considered, although studies of daily hospital admissions might be considered relevant. This model suggests that the deaths of about half of the subjects who enter the at-risk pool due to environmental factors are associated with such factors. We also show that the daily life expectancy of subjects in the at-risk pool is relatively robust to alternative models and environmental fluctuations.
This analysis is believed to be the first to explicitly link environmental effects on morbidity and mortality in a common model; it also indicates that different agents may be involved in each process. The average life expectancy of persons in the at-risk pool is estimated to be 5-8 days, which may be reduced by less than one day by environmental effects. This model suggests that the processes leading up to severe frailty and high risk of imminent death may be more important from a public health perspective