Partitioning the Effects of Weather and Air Pollution on Human Mortality in Santiago, Chile: 1988199

1
Poster U32A-0025
Partitioning the Effects of Weather and Air
Pollution on Human Mortality in Santiago, Chile
1988-1996
David S Grass and Mark A Cane contact
dgrass_at_ldeo.columbia.edu
LAMONT-DOHERTY EARTH OBSERVATORY THE EARTH
INSTITUTE AT COLUMBIA UNIVERSITY
INTRODUCTION One of the greatest challenges in
modern epidemiology has been quantifying the
impact on human health of prolonged exposure to
air pollution (Davis, 2002). Concurrently,
growing concern over the health consequences of
climate change has led to numerous studies of
weather and health. However most of these studies
only speculate as to the combined impacts of both
weather and pollution. These impacts are often
claimed to be synergistic or interactive, but
few studies have measured whether the combined
effects are antagonistic, additive, or
non-linear. There is also little agreement as to
whether pollution or weather is the dominant
environmental influence on human mortality.
Comparisons of the impacts of weather and
pollution in different locations are only
instructive when similar methodologies are used.
At present, there is no standard method for
addressing these questions, but one approach,
Kalksteins Temporal Synoptic Index (TSI), has
been widely applied and widely cited (Smoyer et
al., 2000). The present study suggests how an
elaborated method may provide a more reliable
means to evaluate the relative and absolute
impact of weather and pollution on human
mortality rates.
METHODS The first objective of this study was to
identify those weather patterns and air pollution
regimes most highly associated with increased
mortality. Weather classes are defined since
human physiology reacts to the complete set of
air mass conditions rather than to individual
elements in isolation. Variations in air
pollution concentrations are also largely
determined by the combined effects of multiple
weather elements. We applied a temporal synoptic
index which utilizes a semi-automated
classification scheme based on the use of
empirical orthogonal funtions (EOFs) and cluster
analysis to determine meteorologically homogenous
groups (Kalkstein, 1997). A parallel analysis
identifies oppressive pollution classes. The
differential contributions of weather and
pollution to mortality rates are calculated based
on the results of two-way contingency tables that
consider both effects simultaneously. Forward
stepwise multiple regression was used to
determine the relative importance of weather and
pollution variables in explaining the variation
in excess deaths.
STUDY AREA These issues are of particular concern
in Chile where pneumonia and respiratory illness
are the leading causes of death for children ages
one to four (INE, 1999). The metropolitan region
is home to 45 of the nations population (5.3 of
12 million) and 70 of its industry (Romero et
al. 1999). During most of the year Santiago is
subject to nocturnal inversions and subsidence
inversions associated with the South Pacific
Subtropical Anticyclone. Decreased convective
mixing traps contaminants in the closed basin
formed by the Andes, the Coastal Range, and the
Chacabuco hills to the north, leading to
increases in contaminant concentrations
(Sandoval, 1993). Critical air pollution episodes
are associated with synoptic meteorological
conditions that lead to the development of a
coastal low (Rutllant et al, 1994). During a
coastal low, skies are clear and humidity is low.
The descent of warm continental air and
penetration of sunlight leads to increased
surface temperatures. The initial phase of the
coastal low is associated with weak easterly
winds. The coastal low propagates southward,
eventually bringing cold humid maritime air into
the River Maipo Valley (Ulriksen, 1993).
RESULTS Pollution Classes 4 and 9 are associated
with the highest pollution levels and the highest
mortality rates of any class. The mean weather
conditions of pollution class 4 and 9 resemble
the coastal low described by Garreaud (2002).
Weather Classes 3 and 4 were associated with the
highest mortality rates. With the high mortality
pollution and weather classes, temperature was
the strongest predictor of mortality, followed by
NO2. There were 3 excess deaths/day for offensive
weather or offensive pollution alone, and 7
excess deaths for offensive weather and pollution
combined. Weighted averages of excess deaths, at
various lag times, were used to calculate the
relative risk of mortality due to disease on days
with both offensive weather and offensive
pollution (RR 1.10, 95 CI 1.05 1.16), days
with just offensive weather (RR 1.04, 95 CI
1.01 1.08), and days with just offensive
pollution concentrations (RR 1.04, 95 CI 1.02
1.07).
CONCLUSIONS This study proposes a new methodology
to partition the effects of weather and pollution
on mortality. While it was not possible to
attribute the deaths on offensive pollution days
solely to pollution levels, we can conclude that
NO2 had a significant effect on disease-related
mortality after controlling for the effect of
weather. These preliminary results suggest that
during the winter in Santiago, the effects of
weather and pollution on mortality are additive
and independent. Future applications of this
technique may succeed in providing quantitative
estimates of the individual and combined effects
of weather and air pollution.
ACKNOWLEDGEMENTS The authors would like to thank
Karen Tomic, Alexey Kaplan, Luis Cifuentes,
Ignacio Olaeta, Patricio Aceituno, Tracey
Holloway, José Rutllant, and Alla Fil for their
help with this study. This research was supported
in part by a NSF IGERT Joint Program in Applied
Mathematics and Earth Environmental Science.