Interactions between particulate air pollution and temperature in air pollution mortality time series studies
Introduction
There have been numerous community time series studies on the effects of particulate air pollution on mortality (Daniels et al., 2000; Dominici 2002a, Dominici 2002b; Katsouyanni 1997, Katsouyanni 2001; Kelsall et al., 1997; Michelozzi et al., 1998; Moolgavkar, 2000; Moolgavkar 1995a, Moolgavkar 1995b; Samet et al., 2000; Schwartz, 2000; Schwartz and Dockery, 1992; Schwartz and Marcus, 1990; Smith et al., 2000; Styer et al., 1995; Sunyer et al., 1996; Touloumi et al., 1996; Wong et al., 2001). These studies typically fit a generalized additive model (GAM) (Hastie and Tibshirani, 1990) or a generalized linear model (GLM) (McCullagh and Nelder, 1989) to concurrent time series of daily mortality, particulate air pollution, and meteorological covariates. The fitted models are then used to quantify the effect of particulate air pollution on mortality. A number of these studies investigated the interaction between particulate air pollution and season (Katsouyanni et al., 1997; Kelsall et al., 1997; Michelozzi et al., 1998; Moolgavkar 1995a, Moolgavkar 1995b; Schwartz, 2000; Smith et al., 2000; Styer et al., 1995; Sunyer et al., 1996; Touloumi et al., 1996; Wong et al., 2001). Few studies have looked at the interaction between particulate air pollution and temperature. Katsouyanni et al. (1993) investigated the interaction between air pollution and high temperature. They found that the interaction between high levels of particulate air pollution (measured by smoke) and high temperature was suggestive (p<0.20). Samet et al. (1998) investigated the sensitivity of the particulate air pollution (measured by total suspended particles) mortality effect estimate to alternative methods of controlling weather. They also looked at the interaction between particulate air pollution and weather but did not look explicitly at the interaction between particulate air pollution and temperature. Their analysis was carried out using four different weather models and Poisson log-linear regression. The first two weather models used parametric functions and nonparametric functions of weather, respectively. The third and fourth models used synoptic weather categories. They found little evidence of an interaction between particulate air pollution and weather.
The association between extremes of temperature and increased mortality is well known (Basu and Samet, 2002; Curriero et al., 2002; Mercer, 2003). It is also known that marked changes in ambient temperature can cause physiological stress and alter a person's physiological response to toxic agents (Gordon, 2003). This suggests that an interaction between particulate air pollution and temperature is plausible. We investigate this possibility by (1) using a two-dimensional smooth response surface to model the joint effect of particulate air pollution and temperature on mortality, in a continuous manner, and (2) stratifying the effects of particulate air pollution on mortality by temperature. These are similar to the methods used by Morris and Naumova (1998), who investigated the interaction of carbon monoxide and temperature on daily hospital admissions for congestive heart failure.
Section snippets
Data
The data used in this paper are concurrent daily time series of mortality, temperature, dew point temperature, and particulate air pollution from Cook County, Illinois and Allegheny County, Pennsylvania for the period 1987 to 1994, inclusive. These are the same Cook County and Allegheny County data used by Samet et al. (2000).
The mortality time series, aggregated at the level of county, are nonaccidental daily deaths of individuals aged 65 years and over. Deaths of nonresidents were excluded
Results
We start our investigation of the interaction between PM and temperature using response surfaces (Model 1) as an exploratory tool. The fitted response surfaces will then be used as a guide to help construct appropriate stratified models (Models 2B and 3).
In the results that follow parameter estimation uses the stricter convergence criteria suggested by Dominici 2002a, Dominici 2002b. The standard errors for the PM mortality effect estimates obtained from GAM estimation will be computed using
Simulation study
It can be important to allow for the possibility of an interaction between PM and temperature. For example, in Cook County the PM mortality effect estimates using no-interaction Model 2A along with GLM estimation and exposure measure PM1 were 0.36, 0.14, and 0.10, for α=0.5,1, and 2, respectively. The corresponding PM mortality effect estimates in the upper temperature stratum from interaction Model 2B were 0.62, 0.30, and 0.46. In this situation, allowing for an interaction between PM and
Discussion
In this study, we used two methods to investigate the interaction between PM and temperature. The results of fitting a joint PM–temperature response surface were suggestive of an interaction between PM and temperature. The response surfaces for both Cook County and Allegheny County indicate that the effects of PM on mortality are largest on days of extreme temperature. The response surfaces can also be used as an exploratory tool. Once fit, the shape of the surface can be used to aid the search
Acknowledgements
I thank Paul Switzer for his comments and Scott Zeger for the combined Cook County and Allegheny County mortality and meteorological data used in this paper. The author was partially supported by a Stanford Graduate Fellowship.
References (34)
Role of environmental stress in the physiological response to chemical toxicants
Environ. Res
(2003)Cold—an underrated risk factor for health
Environ. Res
(2003)- et al.
Does weather confound or modify the association of particulate air pollution with mortality? An analysis of the Philadelphia data, 1973–1980
Environ. Res
(1998) - et al.
Relation between elevated ambient temperature and mortalitya review of the epidemiologic evidence
Epidemiol. Rev
(2002) - et al.
Statistical Models in S
(1993) Model uncertainty and health effect studies for particulate matter
Environmetrics
(2000)Statistical issues in the study of air pollution involving airborne particulate matter
Environmetrics
(2000)- et al.
Temperature and mortality in 11 cities of the eastern United States
Am. J. Epidemiol
(2002) - et al.
Estimating particulate matter-mortality dose-response curves and threshold levelsan analysis of daily time-series for the 20 largest US cities
Am. J. Epidemiol
(2000) - et al.
Air pollution and mortalityestimating regional and national dose-response relationships
J. Am. Stat. Assoc
(2002)
On the use of generalized additive models in time-series studies of air pollution and health
Am. J. Epidemiol
Generalized Additive Models
Evidence for interaction between air pollution and high temperature in the causation of excess mortality
Arch. Environ. Health
Short-term effects of ambient sulphur dioxide and particulate matter on mortality in 12 European citiesresults from time series data from the APHEA project. Air Pollution and Health: a European Approach
Br. Med. J
Confounding and effect modification in the short-term effects of ambient particles on total mortalityresults from 29 European cities within the APHEA2 project
Epidemiology
Cited by (124)
On the linkage between urban heat island and urban pollution island: Three-decade literature review towards a conceptual framework
2021, Science of the Total Environment