spatial analysis of long-term exposure to air … · techniques at a sla level. spatial...
TRANSCRIPT
SPATIAL ANALYSIS OF LONG-TERM
EXPOSURE TO AIR POLLUTION AND
CARDIORESPIRATORY MORTALITY
IN BRISBANE, AUSTRALIA
BY
XIAO-YU WANG
Bachelor of Science, Postgraduate Diploma of Science
A thesis submitted for the Degree of Master of Applied Science in the School of Public Health and Institute of Health and Biomedical
Innovation, Queensland University of Technology
MARCH 2008
KEYWORDS
Air pollution, cardiorespiratory, cardiovascular, geographic information system,
interpolation, logistic regression model, mortality, nitrogen dioxide; ozone; particulate
matter; respiratory disease, spatial distribution, socio-ecological factors, sulphur
dioxide
I
SUMMARY
Air pollution is ranked by the World Health Organisation as one of the top ten
contributors to the global burden of disease and injury. Epidemiological studies have
shown that exposure to air pollution is associated with cardiorespiratory diseases.
However, most of the previous studies have looked at this issue using air pollution
data from a single monitoring site or average values from a few monitoring sites in a
city. There is increasing concern that the relationships between air pollution and
mortality may vary with geographical area, particularly for a big city. This thesis
consisted of three interlinked studies that aimed to examine the spatial variation in the
relationship between long-term exposure to air pollution and cardiorespiratory
mortality in Brisbane, Australia.
The first study evaluated the long-term air pollution trends in Brisbane, Australia. Air
pollution data used in this study were provided by the Queensland Environmental
Protection Agency (QEPA). The data comprised the daily average concentrations of
particulate matter less then 10 µm in aerodynamic diameter (PM10), nitrogen dioxide
(NO2), ozone (O3) and sulphur dioxide (SO2) between 1 January 1980 and 31
December 2004 in two monitoring sites (i.e. Eagle farm and Rocklea), and in other
available monitoring sites between 1 January 1996 and 31 December 2004.
Computerised data files of daily mortality between 1 January 1996 and 31 December
2004 in Brisbane city were provided by the Office of Economic and Statistical
Research of the Queensland Treasury. Population data and the Socio-Economic
II
Indexes for Areas (SEIFA) data in 2001 were obtained from the Australian Bureau of
Statistics (ABS) for each statistical local area (SLA) of the Brisbane city.
The long-term air pollution (the daily maximum 1-hour average or daily 24-hour
average concentrations of NO2, O3 and PM10) trends were evaluated using a
polynomial regression model in two monitoring sites (Eagle Farm and Rocklea) in
Brisbane, Australia, between 1980 and 2003. The study found that there were
significant up-and-down features for air pollution concentrations in both monitoring
sites in Brisbane. Rocklea recorded a substantially higher number of days with
concentrations above the relevant daily maximum 1-hour or 24-hour standards than
that in Eagle Farm. Additionally, there was a significant spatial variation in air
pollution concentrations between these areas. Therefore, the results indicated a need
to examine the spatial variation in the relationship between long-term exposure to air
pollution and cardiorespiratory mortality in Brisbane.
The second study examined the spatial variation of SO2 concentrations and
cardiorespiratory mortality in Brisbane between 1999 and 2001. Air pollutant
concentrations were estimated using geographical information systems (GIS)
techniques at a SLA level. Spatial distribution analysis and a multivariable logistic
regression model were employed to investigate the impact of gaseous air pollution on
cardiorespiratory mortality after adjusting for potential confounding effects of age,
sex, calendar year and SEIFA. The results of this study indicate that for every 1 ppb
increase in annual average SO2 concentration, there was an estimated increase of 4.4
% (95 % confidence interval (CI): 1.4 – 7.6 %) and 4.8 % (95 % CI: 2.0 – 7.7 %) in
cardiovascular and cardiorespiratory mortality, respectively. We estimated that the
III
excess number of cardiorespiratory deaths attributable to SO2 was 312 (3.4% of total
cardiorespiratory deaths) in Brisbane during the study period. Our results suggest that
long-term exposure to SO2, even at low levels, is a significant hazard to population
health.
The final study examined the association of long-term exposure to gaseous air
pollution (including NO2, O3 and SO2) with cardiorespiratory mortality in Brisbane,
Australia, 1996 - 2004. The pollutant concentrations were estimated using GIS
techniques at a SLA level. Logistic regression was used to investigate the impact of
NO2, O3 and SO2 on cardiorespiratory mortality after adjusting for potential
confounding effects of age, sex, calendar year and SEIFA. The study found that there
was an estimated 3.1% (95% CI: 0.4 – 5.8%) and 0.5% (95% CI: -0.03 – 1.3 %)
increase in cardiorespiratory mortality for 1 ppb increment in annual average
concentration of SO2 and O3, respectively. However there was no significant
relationship between NO2 and cardiorespiratory mortality observed in the multiple
gaseous pollutants model. The results also indicated that long-term exposure to
gaseous air pollutants in Brisbane, even at the levels lower than most cities in the
world (especially SO2), were associated with cardiorespiratory mortality.
Therefore, spatial patterns of gaseous air pollutants and their impact on health
outcomes need to be assessed for an evaluation of long-term effects of air pollution on
population health in metropolitan areas.
This study examined the relationship between air pollution and health outcomes. GIS
and relevant mapping technologies were used to display the spatial patterns of air
IV
pollution and cardiorespiratory mortality at a SLA level. The results of this study
show that long-term exposure to gaseous air pollution was associated with
cardiorespiratory mortality in Brisbane and this association appeared to vary with
geographic area. These findings may have important public health implications in the
control and prevention of air pollution-related health effects, since now many
countries and governments have paid more attention to control wide spread air
pollution and to protect our environment and human health.
V
CONTENTS
CHAPTER 1: INTRODUCTION ………………………………………………. …1
1.1 INTRODUCTION ………………………………………………………………...1
1.2 AIMS AND HYPOTHESIS ………………………………………………………3
1.3 SIGNIFICANCE OF THE STUDY ………………………………………………5
1.4 CONTENTS AND STRUCTURE OF THE THESIS …………………………….5
CHAPTER 2: LITERATURE REVIEW ……………………………………………7
2.1 INTRODUCTION ………………………………………………………………..7
2.2 EPIDEMIOLOGICAL EVIDENCE FROM INTERNATIONAL AND
AUSTRALIAN STUDIES ……………………………………………………...8
2.3 APPLICATIONS OF GIS AND SPATIAL ANALYSIS IN THE ASSESSMENT
OF THE HEALTH EFFECTS OF AIR POLLUTION ………………………..20
2.4 METHODOLOGICAL ISSUES ……………………….......................................25
2.5 GAPS IN CURRENT KNOWLEDGE BASE…………………………………...36
CHAPTER 3: STUDY DESIGN AND METHODS ……………………………….37
3.1 STUDY DESIGN …………………………………………………………………37
3.2 STUDY AREA ……………………………………………………………………37
3.3 DATA COLLECTION ……………………………………………………………38
VI
3.4 DATA LINKAGE …………………………………………………………………44
3.5 DATA ANALYSIS ………………………………………………………………..44
CHAPTER 4: AIR POLLUTION TRENDS IN BRISBANE, AUSTRALIA,
BETWEEN 1980 AND 2003 ……………………………………………………….49
ABSTRACT …………………………………………………………………………50
4.1 INTRODUCTION ……………………………………………………………….51
4.2 DATA COLLECTION AND ANALYSIS ………………………………….......52
4.3 RESULTS AND DISCUSSION ………………………………………………...54
4.4 CONCLUSIONS ………………………………………………………………...65
ACKNOWLEDGMENTS …………………………………………………………...67
REFERENCES ………………………………………………………………………68
CHAPTER 5: SPATIAL PATTERNS OF SO2 AND CARDIORESPIRATORY
MORTALITY IN BRISBANE, AUSTRALIA, 1999 – 2001 …………………….71
ABSTRACT …………………………………………………………………………73
5.1 INTRODUCTION ……………………………………………………………….74
5.2 METHODS ………………………………………………………………………75
5.3 RESULTS ………………………………………………………………………..81
5.4 DISCUSSION …………………………………………………………………...87
ACKNOWLEDGEMENTS …………………………………………………………91
REFERENCES ………………………………………………………………………92
VII
CHAPTER 6: SPATIAL ANALYSIS OF LONG-TERM EXPOSURE TO
GASEOUS AIR POLLUTANTS AND CARDIORESPIRATORY MORTALITY
IN BRISBANE, AUSTRALIA, 1996 – 2004 ……………………………………...97
ABSTRACT …………………………………………………………………………98
6.1 INTRODUCTION ……………………………………………………………….99
6.2 MATERIALS AND METHODS ………………………………………………100
6.3 RESULTS ………………………………………………………………………106
6.4 DISCUSSION ………………………………………………………………….112
6.5 CONCLUSION ………………………………………………………………...115
ACKNOWLEDGEMENTS ………………………………………………………..116
REFERENCES ……………………………………………………………………..117
CHAPTER 7: GENERAL DISCUSSION ……………………………………….122
7.1 AN OVERVIEW OF KEY FINDINGS FROM THIS STUDY ……………….122
7.2 ALTERNATIVE EXPLANATIONS …………………………………………..125
7.3 COMPARISON WITH OTHER STUDIES …………………………………...127
7.4 STRENGTHS AND LIMITATIONS ………………………………………….130
7.5 PUBLIC HEALTH IMPLICATIONS …………………………………………131
7.6 DIRECTIONS FOR FUTURE RESEARCH …………………………………..132
7.7 CONCLUSIONS AND RECOMMENDATIONS …………………………….133
APPENDIX I ………………………………………………………………………135 APPENDIX II ……………………………………………………………………..136 BIBLOGRAPHY ………………………………………………………………….137
VIII
LIST OF TABLES
Table 2.1 Air pollutants, the categories, sources, national standard and possible health
hazard ………………………………………………………………………..10
Table 3.1 Descriptions of monitoring sites in South-east Queensland region ………39
Table 3.2 Daily max 1-hour O3, NO2 SO2 and 24-hour average PM10 summary Jan.
1996 to Dec. 2004 …………………………………………………………...47
Table 4.1 The R-Squared values of four regression models for two monitoring sites
(Eag: Eagle Farm, Roc: Rocklea) ……………………………………………57
Table 4.2 Frequency of days with NO2, O3 and PM10 concentrations exceeding max 1-
h or 24-h standards at the two selected sites over observed periods ………...63
Table 4.3 Correlation (no. of observations) among daily air pollutant variables for two
monitoring sites ……………………………………………………………..64
Table 4.4 Correlation (no. of observations) between two monitoring sites for daily air
pollutants ……………………………………………………………………65
Table 5.1 Summary statistics of daily maximum 1-hour SO2 concentrations (ppb) and
land use type for six air monitoring stations in Brisbane metropolitan area
between 1999 and 2001 ……………………………………………………..82
Table 5.2 Summary statistics of daily number of cardiorespiratory mortalities in
Brisbane 1999-2001 …………………………………………………………83
Table 5.3 Odds ratio (95% Confidence Interval) for cardiorespiratory mortality
associated with annual average SO2 concentrations in Brisbane by SLA (1999-
2001) …………………………………………………………………………86
IX
Table 5.4 Estimated excess deaths of cardiorespiratory disease in higher SO2 SLAs
(1999 – 2001) ………………………………………………………………..87
Table 6.1 Maximum one hour concentration of gaseous air pollutants in all available
monitoring stations around and in Brisbane city, Australia
(1996 – 2004) ………………………………………………………………107
Table 6.2 Logistic Regression for gaseous air pollutants and cardiorespiratory
mortality disease ……………………………………………………………111
Table 7.1 Summary of mortality associated with air pollution in Brisbane,
Australia ……………………………………………………………………124
X
LIST OF FIGURES
Figure 2.1 An example of a PM2.5 pollution surface created by IDW and Kriging for
2/10/2003 …………………………………………………………………….35
Figure 3.1 Locations of monitoring sites in South-east Queensland region of
EPA ………………………………………………………………………….41
Figure 3.2 The distribution of deaths from RD, CVD, CRD and all causes excluding
external causes between 1996 and 2004 in Brisbane city …………………..42
Figure 3.3 The distributions of population in Brisbane city at the SLA level in
2001 …………………………………………………………………………43
Figure 4.1 The locations of Eagle Farm and Rocklea monitoring stations in
Brisbane ……………………………………………………………………...53
Figure 4.2 Annual means of daily maximum 1-h or 24-h concentrations of NO2, O3
between 1980 and 2003, PM10 data available for Eagle Farm between 1994
and 2003 and for Rocklea between 1986 and 2003. The line indicates trendline
of three air pollutant concentrations …………………………………………55
Figure 4.3 Seasonal (monthly) averages of daily maximum 1-h or 24-h mean levels
measured for NO2, O3 and PM10 during the study period, for Eagle Farm (eag)
and Rocklea (roc) ……………………………………………………………56
Figure 4.4 Daily maximum 1-h NO2, O3 and daily 24-h PM10 mean levels during the
study period. The broken line indicates the current air quality NEPM
Standard ……………………………………………………………………...62
XI
Figure 5.1 Location of ambient SO2 monitoring stations in and around the Brisbane
metropolitan area 1999 – 2001 (Brisbane CBD (QUT and SCI), Eagle Farm,
Flinders View, Springwood and Wynnum) ………………………………….76
Figure 5.2 Spatial patterns of annual average SO2 and number of people exposed
between 1999 and 2001 in the Brisbane at a SLA level (For each year, daily
maximum 1-hour SO2 concentrations data from all operational stations were
aggregated using IDW method to estimate the annual average SO2) ………..83
Figure 5.3 Spatial patterns of annual average standardised rates of respiratory,
cardiovascular and cardiorespiratory mortality; and annual average SO2
exposed in the Brisbane at a SLA level (1999 – 2001) ……………………...85
Figure 6.1 Locations of gaseous air pollution monitoring stations around urban
Brisbane city ………………………………………………………………..102
Figure 6.2 Spatial patterns of annual average concentrations of NO2, O3 and SO2;
annual average SMR of cardiorespiratory mortality and number of people
exposed in Brisbane at a SLA level (1996 - 2004) …………………..110 - 111
XII
LIST OF ABBREVIATION
ABS: Australian bureau of statistics
CI: Confidential interval
CRD: Cardiorespiratory disease
CVD: Cardiovascular disease
EPA: Environmental protection agency
GIS: Geographic information system
ICD: International Classification of Diseases
IDW: Inverse distance weighting
NO2: Nitrogen dioxide
O3: Ozone
OR: Odds ratio
PM: Particulate matters
PM2.5: Particulate matter with diameter less then 2.5 µm
PM10: Particulate matter with diameter less then 10 µm
RD: Respiratory disease
SEIFA: Socio-economic indexes for areas
SLA: Statistical local areas
SMR: Standardized mortality rate
SO2: Sulphur dioxide
WHO: World health organisation
XIII
STATEMENT OF AUTHORSHIP
To the best of my knowledge and belief, the work presented in this thesis is original
unless otherwise acknowledged in the text. The material, either in whole or in part,
has not been submitted for a degree or diploma at any other University.
Xiao-Yu Wang
March 2008
XIV
DEDICATION
To
my beloved husband
Hua
our son
Jimmy Zhenli
and our daughter
Cathy Zhenao
XV
ACKNOWLEDGEMENTS
I am greatly indebted to my supervisory team, A/Prof. Shilu Tong and Dr. Wenbiao
Hu, for their critical and thoughtful comments, support, continuing guidance,
encouragement, patience and constant help in all aspects throughout my study.
The author wishes to thank Dr. Ken Verall, Prof. Rod Gerber and Prof. Rodney Wolff
for their insightful comments on some of the manuscripts for this study.
The author would also like to thank Prof. Beth Newman and Genny Carter for their
general assistance during the study; the fellow research students and colleagues for
their interest and help, particular Dr. C. Ren, Ms. L. Chen, Ms. S. Naish, Dr. Y. Liu,
Ms. E. Winkler, Ms. L. McKinnon Ms. Y. Zhang, Ms. W. Yu, Mr. X. Qi; and all other
colleagues and folks in the school, faculty and IHBI for their advice and assistance
with this research and personal friendship.
Finally, the author would like to especially acknowledge the examination panel for
their contributions to improving this thesis.
XVI
PUBLICATIONS BY THE CANDIDATE
Journal Articles
Xiao-Yu Wang, Shilu Tong, Ken Verrall, Rod Gerber and Rodney Wolff. (2006) Air
pollution trends in Brisbane, Australia, between 1980 and 2003. Clean Air and
Environment Quality, 40 (1): 34-39.
Xiao-Yu Wang, Shilu Tong, Ken Verrall, Rod Gerber and Rodney Wolff. (2007)
Spatial Patterns of SO2 and Cardiorespiratory Mortality in Brisbane, Australia, 1999 –
2001. Environmental Health, 7 (3): 64-74.
Xiao-Yu Wang, Wenbiao Hu and Shilu Tong. (2007) Spatial analysis of long-term
exposure to gaseous air pollutants and cardiorespiratory mortality in Brisbane,
Australia, 1996 – 2004 (To be Submitted)
XVII
1
CHAPTER 1: INTRODUCTION
1.1 INTRODUCTION
Air pollution is a major environmental health problem scorecard from both indoor and
outdoor sources; it has been ranked as one of the top ten causes of global burden of
disease and injury by the World Health Organisation (Murray & Lopez, 1996). In some
Western countries, it is estimated that car emissions kill twice as many people as car
crashes (Kunzli et al., 2000; QEPA, 2001; 2007). There is also increasing global
awareness of the extreme levels of indoor and outdoor air pollution arising from the use
of coal and biomass (eg, wood, farm waste, and cow dung) for cooking and heating in
developing countries (World Health Organization, 1997). In urban Australia, the main
sources of air pollution include motor vehicle emissions, wood smoke from home
heating, and industry (Environmental Protection Agency, 2007; Kjellstrom et al., 2002;
QEPA, 2007). Bushfires are another important source of air pollution in some parts of
Australia (L Chen et al., 2006; Environmental Protection Agency, 2007; EPA, 2007;
QEPA, 2007; Sim, 2002).
Exposure to particulates and other pollutants in air has been associated with significant
public health impacts and is now tightly linked to a wide range of health outcomes
including cardiovascular disease, respiratory disease, lung cancer, emphysema, and
fibrosis (Brunekreef & Holgate, 2002). The rapid pace of urbanization globally means
that more people are living in large cities than ever before (Cifuentes et al., 2001). This
2
raises concerns regarding the public health impacts of exposure to traffic-related air
pollution in urban areas.
In order to investigate the relationship between air pollution and health outcomes (e.g.,
mortality and morbidity) many epidemiological studies have been conducted in North
America, Europe Asia, Australia and New Zealand over the past decade. These studies
have found the consistent associations between the particulate matter less than 10 μm
(PM10) and/or 2.5 μm (PM2.5) in aerodynamic diameter and cardiorespiratory morbidity
and mortality (Burnett et al., 2001; Chen et al., 2007; Chen et al., 2006; Fischer et al.,
2003; Hayes, 2003; Jerrett et al., 2005a; Jerrett and Finkelstein, 2005b; Kan & Chen,
2003; Kan et al., 2004; Krewski et al., 2005; Peng et al., 2005; Powe & Willis, 2004;
Ren & Tong, 2006; Ren et al., 2006; Venners et al., 2003; Wong et al., 2002).
Associations have also been reported for gaseous air pollutants, such as nitrogen dioxide
(NO2), ozone (O3) and sulphur dioxide (SO2) (Ballester et al., 2001; Krewski et al.,
2003; Venners et al., 2003).
However, most of the previous studies have examined the association between air
pollution and health outcomes using air pollution data from one monitoring station or
average values from a few monitoring stations (H Kan & Chen, 2003; Kunzli et al.,
2000; Ren et al., 2006; Rutherford et al., 2000; Sullivan & Beudeker, 1999). There is
increasing concern that the relationship between air pollution and mortality may vary by
geographical area (Burnett et al., 2001; L Chen et al., 2007; Krewski et al., 2003;
Maheswaran & Craglia, 2004; Peng et al., 2005a; Scoggins et al., 2004). Now it is
widely accepted that Geographical Information Systems (GIS) and related mapping
3
technologies provide important tools in the visualisation, exploration and modelling of
environment and health data.
Recently, GIS and spatial modelling have been used in studies of risk factors of vector-
borne diseases, water borne diseases, and sexual transmitted diseases, injury control and
prevention, and the analysis of disease control policy and planning (Botto et al., 2005;
Briet et al., 2005; Kitron, 1998; Odoi et al., 2004; Semaan et al., 2007; Yiannakoulias et
al., 2003). GIS and spatiotemporal modelling methods offer expanding opportunities for
epidemiological and public health research because they allow a user to display and
model the spatial relationships and patterns between causes and disease when
geographic distributions are part of the problem. Even when used minimally, they allow
a spatial perspective on disease. Used to their optimum level for spatiotemporal analysis
and decision making, they have rich potential for environmental planning, management,
health surveillance and assessment of risk factors (Briggs et al., 1997; Choi et al., 2006;
Christakos & Serre, 2000). Therefore, it is envisaged that GIS and spatial modelling
will play an increasingly important role in public health research.
1.2 AIMS AND HYPOTHESIS
Environmental health risk assessment and management is a great challenge to scientists,
decision-makers and general communities. There is a need to enhance our
understanding, predictability, and efficiency in the control and management of
environmental health hazards. However, it remains unclear how to best model the long-
term relationship between environmental exposures and community well-being and how
to utilise the available surveillance data, effectively and efficiently, for environmental
4
health decision-making. This study aims to focus on the development of a spatial and
temporal approach to environmental health modelling, with the display and modelling
of the long-term relationship between air pollution and cardiorespiratory disease using
GIS and spatial model in Brisbane, Australia.
1.2.1 The major objectives of this study
1. Visualise the long-term trends of air pollution in Brisbane, Australia;
2. Visualise the spatial distributions of gaseous air pollutants (NO2, O3 and SO2)
and cardiorespiratory mortality;
3. Quantify the relationship between long-term exposure to air pollution and
cardiorespiratory mortality.
1.2.2 Hypotheses
The central hypothesis to be tested is that the cardiorespiratory mortality is associated
with air pollution. This association varies by geographic area and can be assessed using
a GIS and spatial modelling approach. As a result of this study, the applications of GIS
and spatial modelling will be assessed in the process of environmental health modelling.
1.2.3 Specific hypotheses
1. Spatial and temporal distribution of gaseous air pollution (NO2, O3 and SO2) and
cardiorespiratory mortality can be assessed using GIS;
5
2. The relationship between exposure to air pollution and cardiorespiratory
mortality can be determined using GIS techniques, spatial analysis and
multivariable regression model.
1.3 SIGNIFICANCE OF THE STUDY
This is the first study to use relatively long-term data to examine the spatial variation in
the relationships between air pollution (NO2, O3 and SO2), socio-demographic factors
(SEIFA, population, sex and age etc) and the cardiorespiratory mortality in Brisbane,
Australia. Through the development of a proper modelling approach, the impact of
exposure to air pollution on community well-being and population health can be more
adequately and promptly assessed. Additionally, the methods developed through this
study may have a wider application to other epidemiological and public health
problems.
1.4 CONTENTS AND STRUCTURE OF THE THESIS
This thesis is presented using the publication model. As such, it contains three
manuscripts, each designed to stand on its own. Chapter 1 is an introduction. Chapter 2
critically reviews the literature relating to applications of GIS and spatial analysis in air
pollution and health outcome research. Chapter 3 provides the general study design and
methods. Chapters 4-6 are presented as three manuscripts. Each manuscript was written
in the conventional publication style for a particular journal. Because each manuscript
was designed to stand alone, there was an inevitable degree of repetitiveness in their
introduction, methods and discussion sections.
6
Chapter 7 provides an overview on the key findings of the study, and discusses these
findings in relation to the overall aims and specific objectives of the research project.
This chapter further discusses the similarities and differences between this study and
others, strengths and limitations, public health implications of the study findings, and
directions for future research.
Tables and figures are provided in the text to facilitate reading. The references for each
of the manuscripts are presented at the end of their corresponding chapters. A complete
list of biblography (including references cited in the Chapter 1, 2, 3 and 7) is provided
at the end of the thesis.
7
CHAPTER 2: LITERATURE REVIEW
2.1 INTRODUCTION
During early to middle twentieth century, air pollution episodes in Europe (e.g., London
air pollution in 1952) and the United States (e.g., in the Los Angeles basin in the late
1940s and 1950s) brought a wide attention to the study of air pollution related health
effects, including mortality and morbidity (Firket, 1936; Ministry of Health, 1954;
Schrenk et al., 1949). By the late 1970s, due mainly to changes of legislation,
concentrations of air pollutants in developed countries had been reduced greatly and
were no longer considered by many to be a public health concern. However, since the
early 1990s, more studies with sophisticated and sensitive monitoring techniques and
advanced statistical methods have demonstrated that air pollutant levels even below
ambient guidelines may still affect health (Brunekreef et al., 1995; Hoek et al., 2002;
Mage et al., 1999; Pope, 2000a; 2000b; Scoggins et al., 2004).
In recent years, adverse health effects, on mortality and morbidity, have been observed
at lower and lower air pollutant levels of (Bell et al., 2004; Vedal et al., 2003). As air
pollution and morbidity/mortality usually have strong spatial and temporal patterns,
increasing attention has been paid to the assessment of spatiotemporal aspects of the
relationship between air pollution and health outcomes. GIS and spatiotemporal
modelling methods are often used to display and model those aspects. These advanced
methods offer new and expanding opportunities for assessing the health impact of air
pollution.
8
The levels of air pollution often change over time and space. Earlier studies assumed
that the data for a single point was representative of a vast geographical area. Recently,
detailed understanding of complex urban air quality processes has been aided by the
application of urban airshed models. These models account for spatial and temporal
variations as well as differences in the reactivity of air pollutants and therefore can
provide a detailed spatial picture of pollutant levels (Carreras-Sospedra et al., 2006;
Scoggins et al., 2004). Coupled with GIS techniques, these models have great potential
to improve exposure measurements to link with health (Cicero-Fernandez et al., 2001;
English et al., 1999; Hoek et al., 2001)
There is a large amount of literature on “air pollution and health”. For example, about 500
publications on this topic per year were identified from Medline alone. Recently, GIS and
spatial analysis methods have been used to evaluate the health effect of exposure to low
levels of air pollution. The purpose of this review is to examine current literature on
major epidemiological studies of air pollution and health using GIS and spatial methods
published over the last two decades. Through this review, I will also evaluate the
methodology, strengths and limitations of GIS and spatial analysis tools, and to make
recommendations for future applications of GIS and spatial analysis in air pollution and
health research.
2.2 EPIDEMIOLOGICAL EVIDENCE FROM INTERNATIONAL AND
AUSTRALIAN STUDIES
9
Air pollution (i.e., airborne substances in quantities) could harm the comfort or health of
humans. These substances are called air pollutants and can be either particles, liquids or
gaseous in nature (QEPA, 2007).
2.2.1 Air pollutants of current interest in epidemiological studies
Air pollution studies of current interest include: particulate matter with diameter less than
10µm (PM10) or 2.5µm (PM2.5), ozone (O3), nitrogen dioxide (NO2), sulphur dioxide
(SO2), and carbon monoxide (CO). Table 2.1 shows the categories, sources, national
standard and possible health hazard of these pollutants (QEPA, 2007).
Particulate matters (PM)
Particulate air pollution is a mixture of solid, liquid, or solid and liquid particles
suspended in the air. The size of suspended particles varies from a few nm to tens of µm.
The largest particles (coarse fraction) are mechanically produced by attrition of larger
particles. Small particles (<1 µm) are largely formed from gases, and the smallest
(<0.1µm, ultrafine) are formed by nucleation resulting from condensation or chemical
reactions that form new particles. In practical terms, a distinction is made between PM10
(“thoracic” particles smaller than 10 µm in diameter that can penetrate into the lower
respiratory system), PM2.5 (“respirable” particles smaller than 2.5 µm that can penetrate
into the gas-exchange region of the lung), and ultrafine particles smaller than 100 nm
which contribute little to particle mass but which are most abundant in terms of
numbers and offer a very large surface area, with increasing degrees of lung penetration
(Brunekreef & Holgate, 2002). The main sources of ambient particles are fossil fuel
10
combustion, biomass burning and the processing of metals. Road transportation is the
major source of particles in urban areas (Pooley & Mille, 1999).
Table 2.1 Air pollutants, the categories, sources, national standard and possible health
hazard
Air
pollutant Caused Air NEPM* standards Hazard for human
PM10
Wood burning, diesel vehicles
and industry
50 µg/m3 (24-hour)
Can enter the human respiratory system and
penetrate deeply into the lungs causing
adverse effects.
O3
Secondary pollutant. Formed
from precursors such as auto
emissions 100 ppb (1-hour)
Affect the human cardiac and respiratory
systems, irritating the eyes, nose, throat, and
lungs.
NO2
Combustion (automobiles and
industry). Precursor for O3
120 ppb (1-hour)
Can increase a person's susceptibility to,
and the severity of, respiratory infections and
asthma. Long-term exposure to high levels of
nitrogen dioxide can cause chronic lung
disease.
SO2
Coal fired power plants,
smelters, food processing,
paper and pulp mills
200 ppb (1-hour)
Affect the respiratory system, the functions of
the lungs and irritate our eyes. When sulphur
dioxide irritates the respiratory tract it causes
coughing, mucus secretion, aggravates
conditions such as asthma and chronic
bronchitis and makes people more prone to
respiratory tract infections.
* NEPM: National Environment Protection Measure, Australia
Ozone (O3)
Ozone is an indicator of photochemical smog. Ozone is a colourless, highly reactive gas
with a distinctive odour. It is formed naturally by electrical discharge (lightening) in the
upper atmosphere at altitudes of between 15 and 35km. Stratospheric ozone protects
11
the Earth from harmful ultraviolet radiation from the sun (QEPA, 2007). Ozone is a
strong oxidising agent formed in the troposphere through a complex series of reactions
involving the action of sunlight on nitrogen dioxide and hydrocarbons (WHO, 2000).
Concentrations in city centres tend to be lower than those in suburbs and are mainly as a
result of the scavenging of ozone by nitric oxide originating from traffic (Brunekreef &
Holgate, 2002). It has been shown experimentally that exposure to ozone can affect the
human cardiac and respiratory systems, and cause irritation to the eyes, nose, throat and
lungs (Chan-Yeung, 2000; Schwela, 2000; Thurston & Ito, 2001).
Nitrogen oxides (NOx)
Nitrogen oxide (NO) and nitrogen dioxide (NO2) are produced from natural sources,
motor vehicle emissions and other fuel combustion processes (WHO, 2000). Nitric
oxide is colourless and odourless and is oxidised in the atmosphere to form nitrogen
dioxide. In ambient conditions, nitric oxide is rapidly transformed into nitrogen dioxide
by atmospheric oxidants such as ozone (Brunekreef & Holgate, 2002). Nitrogen dioxide
is an odorous, brown, acidic, highly-corrosive gas that can affect our health and
environment. Long-term exposure to nitrogen dioxide can result in chronic lung
diseases. It may also affect sensory perception by reducing a person’s ability to smell an
odour (Ackermann-Liebrich & Rapp, 1999).
Sulphur dioxide (SO2)
SO2 is a colourless gas with a sharp, irritating odour. It is produced from the burning of
fossil fuels (coal and oil) and the smelting of mineral ores that contain sulphur. When
12
SO2 combines with water, it forms sulphuric acid rain. When acid rain falls it can cause
deforestation, acidify waterways to the detriment of aquatic life and corrode building
materials and paints. It is a harmful gas pollutant formed when sulphur combines with
oxygen during the burning of fossil fuels. SO2 can affect the respiratory system, damage
the functions of the lungs and irritate the eyes. SO2 can cause coughing, mucus
secretion, and other conditions such as asthma and chronic bronchitis and can make
people more prone to respiratory tract infections. SO2 can also attach itself to particles
and, if these particles are inhaled, they can cause more serious effects (QEPA, 2001).
Some recent studies show a consistent effect of SO2 pollution on cardiorespiratory
mortality (Fischer et al., 2003; H D Kan et al., 2003; Powe & Willis, 2004; Venners et
al., 2003; T W Wong et al., 2002).
There is substantial evidence linking outdoor air pollution and health. Numerous studies
have examined the association between mortality and short-term fluctuations in air
pollution (Pope, 2000a). A smaller number of studies have reported an increase in
mortality with long-term exposure to outdoor air pollution (Hoek et al., 2002; Scoggins
et al., 2004). Several studies have paid special attention to spatial difference in the
health impact of air pollution across cities. I review those studies below.
2.2.2 Main findings from multi-city epidemiological studies of air pollution and
health outcomes
Epidemiological studies on the health effect of air pollution have been conducted in
various regions (e.g., America, Europe, Asia, Australia and New Zealand) (Burnett et al.,
2001; Chen et al., 2007; Chen et al., 2006; Fischer et al., 2003; Hayes, 2003; Jerrett et al.,
13
2005; Jerrett & Finkelstein, 2005; Kan & Chen, 2003; Kan et al., 2004; Krewski et al.,
2005; Peng et al., 2005; Powe & Willis, 2004; Ren & Tong, 2006; Ren et al., 2006;
Venners et al., 2003; Wong et al., 2002). These studies have used different health
indicators, different pollutants, different lag periods and different analytic approaches.
And these studies also included in both short-term and long-term health effects of air
pollution, with emphasis on mortality, morbidity and hospital admissions. In generally,
the short-term study is focused on a daily or weekly basis, and long-term study is focused
on seasonal or annual basis(Brunekreef & Holgate, 2002).
The Air Pollution and Health: a European Approach (APHEA) studies have provided
many new insights in Europe. The studies of APHEA provide quantitative estimates of
the short-term health effects of air pollution using an extensive database from 10 different
European countries with a total population exceeding 25 million through time series data
and meta-analysis. The project represents various social, environmental and air pollution
situations. Daily measurements of black smoke, SO2, NO2 and O3 were derived from
existing monitoring networks. The outcome data included daily counts of total and cause-
specific deaths and hospital emergency admissions. Generalised linear regression model
(GLM) with Poisson distribution link, dealt with autocorrelation and overdispersion, was
used in the analysis after controlling for all potential confounding factors. In APHEA-1,
daily mortality increased in six cities (Touloumi et al., 1997) by 2.9% per 50 µg/m3
increase in the 1-h maximum ozone concentration. Associations between NO2 and
mortality were also found, but these were sensitive to adjustment for black smoke,
suggesting that the nitrogen dioxide represented a mixture of traffic-related air pollution.
14
In the late 1990s, a new series of studies (APHEA-2) was performed using data on the
PM10 fraction. The APHEA-2 mortality study covered a population of more than 43
million people living in 29 European cities, which were all studied for more than 5
years in the early-mid 1990s (Katsouyanni et al., 2001). In this study, they considered
confounding from other pollutants especially meteorologic and chronologic variables.
They investigated several variables such as climate, population, and geography as
potential effect modifiers. For the individual city analysis, Generalized additive model
(GAM) extending GLMs, using a smoother to control for seasonal patterns, were
applied. The results confirm those previously reported on the effects of ambient
particles on mortality.
Atkinson (2001) estimated that all cause daily mortality increased by 0.6% (95% CI:
0.4–0.8) for each 10 µg/m3 increase in PM10 (Atkinson et al., 2001). The APHEA-2
hospital admission study covered a population of 38 million living in eight European
cities, which were studied for 3–9 years in the early to mid 1990s. Hospital admissions
for asthma and chronic obstructive pulmonary disease (COPD) among people older than
65 years were increased by 1.0% (0·4–1·5) per 10 µg/m3 PM10, and admissions for
cardiovascular disease (CVD) increased by about 0.5% (95% CI: 0·2–0·8) and 1.1%
(95% CI: 0.4–1.8) per 10 µg/m3 increment in PM10 and black smoke, respectively, (Le
Tertre et al., 2002).
In recent APHEA-2 study (Aga et al., 2003), the effects of ambient particles on
mortality among persons ≥ 65 yrs were investigated from 28 European cities. For
individual city analysis, GAM with a locally weighted regression (LOESS) smoother to
control for seasonal effects was applied. The results suggested that ambient particles
15
have the effects on mortality among the elderly, with relative risks comparable or
slightly higher than those observed for total mortality. The effects of air pollution on the
older persons are of particular importance, since the attributable number of events will
be much larger, compared to the number of deaths among the younger population.
In the United States, the National Morbidity, Mortality, and Air Pollution Study
(NMMAPS) was a large multi-city time series study of the short-term effects of ambient
air pollution on daily mortality and morbidity in the United States. The analyses of the
original 90 city, 8 year (1987-1994) database can be found in Samet et al. (2000a;
2000b; 2000c), Bell et al. (2004), Daniels et al. (2004), and Dominici et al. (2006).
All cause mortality increased by 0.5% (95% CI: 0.1%, 0.9%) for each increase of 10
µg/m3 in PM10. The estimated increase in the relative rate of death from cardiovascular
and respiratory disease was 0.7% (95% CI: 0.2%, 1.2%) (Samet et al., 2000c).
Bell (2004) analysed the 95 NMMAPS community data to examine the association
between ozone concentrations and mortality, showing that a 10 ppb increase in the
previous week’s ozone was associated with a 0.5% (posterior interval [PI], 0.3% - 0.8%)
increase in daily mortality and a 0.64% (95% PI, 0.31% - 0.98%) increase in
cardiovascular and respiratory mortality. And the estimates of the exposure over the
previous week were larger than those considering only a single day’s exposure (Bell et
al., 2004).
Daniels (2004) developed flexible modelling strategies for time-series data that include
spline and threshold concentration-response models for 20 largest US cities (1987 –
16
1994) using concentration of PM10 as the exposure measure. The model showed a linear
relation without indicating a threshold for the relative risks of death from all causes and
from cardio-respiratory causes. By contrast, for causes other than cardio-respiratory, the
relative risk did not increase until the PM10 concentration reached approximately 50
µg/m3. For total mortality, a linear model without threshold was preferred to either of
the threshold model and spline model. The findings were similar for combined
cardiovascular and respiratory deaths (Daniels et al., 2004).
Recently, Dominici (2006) in USA examined the short-term association between fine
particulate air pollution and hospital admissions and found that exposure to PM2.5 was
associated with different health outcomes(Dominici et al., 2006). The largest association
was observed for heart failure - viz., a 10 µg/m3 increase in PM2.5 was found to be
associated with a 1.3% (95% PI: 0.8%, 1.8%) increase in hospital admissions from heart
failure on the same day .
In Asia, there are also a number of studies of air pollution and health in China (include
Hong Kong and Taiwan), Japan, and other countries. Through the PubMed search of air
pollution and health outcomes in Asia, over half of the studies came from China. China
is a country undergoing rapid development. Its economy has rapidly developed in the
recent two decades. Economic development is usually linked with increase in energy
consumption and industry emissions, which in turn leads to worsening of air quality,
especially in metropolitan areas. In a study by Kan and Chen (2003), the relationship
between outdoor air pollution and daily mortality from June 2000 to December 2001 in
Shanghai was assessed using a semi-parametric GAM. The research showed that an
increase of 10 microg/m3 in PM10, SO2, and NO2 corresponded to a respective increase
17
in relative risk of mortality from all causes of 1.003 (95% CI, 1.001 - 1.005), 1.014
(95% CI, 1.008 - 1.020), and 1.015 (95% CI, 1.008 - 1.022). Risks for deaths from
cardiovascular and chronic obstructive pulmonary diseases (COPD) showed a stronger
association than those for total mortality for every pollutant considered. In the multi-
pollutants models, the association between SO2 and daily mortality was not affected by
the inclusion of other pollutants; however, for PM10 and NO2, the inclusion of other
pollutants may weaken their effects on mortality.
Since China joined the Global Environmental Monitoring System (GEMS) program in
the late eighties of the last century, regular systematic monitoring of air pollutants has
become routine practice. Yan (2002) found that the ambient air lead levels have been
declining since leaded gasoline was banned in 1997 in large cities. Unleaded gasoline
became available nationwide in 2000. Just 1 year after the introduction of unleaded
gasoline in Shanghai, blood lead in children gradually decreased from 83 to 80 μg/l, and
further down to 76μg/l after 2 years (Yan et al., 2002).
In recent years, more in-depth studies were conducted on the impact of air pollution on
human health and sustainable development, including inventory of air pollution sources,
spatiotemporal distribution of pollutants, and the exposure–response relationship
between air pollution level and mortality/morbidity (both acute and chronic effects).
These studies enabled quantitative assessment of the impact of air pollution upon health,
from which economic analysis could be made (H Kan & Chen, 2004; E Y Wong et al.,
2003). For instance, a study in Shanghai indicated that the health impact from exposure
to air pollution constituted a 1.6% loss in GDP in 2000. In this study, environment and
health impacts under various energy scenarios were also analyzed, thus providing
18
various options for decision makers. It also provides a practical new approach or an
alternative way of thinking in making similar health-based risk assessment in other
Chinese cities (H D Kan & Chen, 2002; Pan et al., 2007).
Wong et al (2002) investigated the association between air pollution and mortality in
Hong Kong. A Poisson regression was performed to examine the relationship between
concentrations of daily air pollutants and daily mortality from respiratory and
cardiovascular diseases during 1995 -1998 using an APHEA approach. They found
significant associations between the concentrations of all pollutants and mortalities from
all respiratory diseases and ischaemic heart diseases (IHD). The relative risks (RRs)
were calculated for all respiratory mortalities (for a 10 μg/m3 increase in the
concentration of one pollutant) ranged from 1.008 (for PM10) to 1.015 (for SO2) and
were higher for COPD with all pollutants except SO2, ranging from 1.017 (for PM10) to
1.034 (for O3). RRs for IHD ranged from 1.009 (for O3) to 1.028 (for SO2). In a
multipollutant model, O3 and SO2 were significantly associated with all respiratory
mortalities, whereas NO2 was associated with mortality from IHD. PM10 was not
associated with respiratory or cardiovascular mortalities in multipollutant analyses.
The meta-analysis of the Asian literature by the Health Effects Institute (HEI) found that
short-term exposure to air pollution is associated with increases in mortality and
morbidity in Asian populations. PM10 and NO2 were associated with all-cause mortality.
It was found that a 0.5% increase in mortality associated with a 10 ug/m3 increase in
PM10, a key measure of pollution, and a 1% increase in hospital respiratory admissions
associated with increased levels of NO2. Given the high levels of air pollution in many
Asian cites (>100 ug/m3), the public health impact could be substantial. While the small
19
number of cities available limited the comparisons that can be made at this stage of
analysis, the estimated effects are similar to those found in the extensive studies
conducted in western countries (Health Effects Institute, 2004). A key challenge for the
analysis is that the majority of the studies are concentrated in the more developed regions
of East Asia. Areas of South and Southeast Asia, many with high levels of air pollution
and high poverty are less well studied, limiting the understanding of whether air pollution
may have increased effects on the poor.
A recent study by Simpson (2005) examined the short-term effects of air pollution (O3,
NO2, bsp) on daily mortality (total all-cause, respiratory and cardiovascular by all ages,
and cardiovascular by ages ≥ 65) in four Australian cities (Brisbane, Melbourne, Perth
and Sydney), where more than 50% of Australians reside. 24-hour average bsp;
maximum 1 hour concentrations of NO2 and O3; and 4-hour maximum O3
concentrations between 1996 and 1999 were calculated from data provided through a
network of sites across each city(R Simpson et al., 2005). In this study, Poisson
regression models were conducted for each city utilising alternative modelling
approaches (e.g. GAM and GLM) within the APHEA2 framework, and the single-city
results were pooled in a meta-analysis. The results for the effect estimates in this study
for the single cities are similar to results for earlier single-city studies in Australia
(Morgan et al., 1998a; R Simpson et al., 2000; R W Simpson et al., 1997) in that the
same pollutants show significant impacts; however, the time periods are different. There
were significant effects on total mortality. No significant differences between cities
were found, the particle and NO2 impacts appear to be related in all analysis. Meta-
analyses carried out for three of the four cities yielded estimates for the increase in the
20
daily total number of deaths of 0.2% for a 10 µg/m3 increase in PM10 concentration, and
0.9% for a 10 µg/m3 increase in PM2.5 concentration.
2.3 APPLICATIONS OF GIS AND SPATIAL ANALYSIS IN THE
ASSESSMENT OF THE HEALTH EFFECTS OF AIR POLLUTION
The studies on the relationship between air pollution and mortality or morbidity have
primarily used time-series methods that examine variations in health outcomes through
time (usually daily variations) across whole cities and relate these to air pollution levels
(Morgan et al., 1998a; Morgan et al., 1998b). Because the health impact of air pollution
may involve a long and cumulative process, and air pollution levels differ with different
areas, it seems over-simplistic to use an average level of air pollution to represent a
whole city, particularly for large cities. Therefore, it is increasing realised that it’s
important to assess the spatial features of air pollution and its health effects within a city.
It is widely accepted that GIS and related mapping technologies provide important
potential benefits in the visualisation, exploration and modelling of the kind of spatial
data typical in many environmental health applications (Briggs & Elliott, 1995; Croner
et al., 1996; Vine et al., 1997). At the same time it is acknowledged that there are
challenges in using appropriate GIS techniques for different levels of analytical
sophistication appropriate in this field (Maheswaran & Craglia, 2004; Maynard &
Conway, 2006).
There are a growing number of spatial analyses in investigating the relationship between
air pollution and disease (Burnett et al., 2001; Pikhart et al., 2001; Scoggins et al., 2004).
21
Burnett (2001) presented a statistical model for linking spatial variation in ambient air
pollution to mortality. The methods are illustrated with an analysis of the American
Cancer Society cohort to determine whether all cause mortality is associated with
concentrations of sulphate particles. ARC/INFO was used to display the spatial
distribution of sulphate concentrations based on 1980-81 average values and specific
relative risks of mortality in 144 metropolitan statistical areas. The relative risk
associated with a 4.2 µg/m3 interquartile range of sulphate distribution for all causes of
death was 1.051 (95% CI 1.036-1.066) based on the Cox proportional hazards survival
model. They found that the estimates of the association between the individual risk
factors and mortality and their estimates of the uncertainty were nearly identical in the
Cox survival model and the random-effects Cox survival model, thus validating the use
of the Cox model to identify the set of individual risk factors for mortality (Burnett et
al., 2001).
Pikhart (2001) used a novel technique to estimate the outdoor concentrations of SO2 at
small-area level to study the long-term effects of SO2 on respiratory symptoms and
disease in children. This is as part of the international SAVIAH (Small-Area Variations
In Air quality and Health) study, based on 6,959 children aged 7-10 in Prague (Czech
Republic) and Poznan (Poland) whose parents completed a questionnaire covering
respiratory health, demographic and socio-economic factors and health behaviours.
Outdoor SO2 was measured by passive samplers at 80 sites in Poznan and 50 sites in
Prague during 2-week campaigns. GIS was used to estimate concentrations of SO2 at
each point (location). They found long-term outdoor winter concentrations of SO2 were
associated with wheezing/whistling and with asthma diagnosed after controlling for
confounding factors (Pikhart et al., 2001).
22
Scoggins (2004) used urban air shed modelling and GIS-based techniques to quantify
long-term exposure to relate air pollution levels to mortality in Auckland, New Zealand.
A logistic regression model was used to investigate how air pollution influences the
probability of death after adjustment for confounding factors. They found a positive
association between annual average NO2 and mortality at the census area unit levels.
This study provides additional evidence that long-term exposure to poor air quality,
even at levels below current standards, is a hazard to public health (Scoggins et al.,
2004).
Chen et al (2007) investigated the spatial variability of PM10 and its association with
respiratory emergency hospital admissions across six geographic areas in Brisbane,
Australia for the period of 1 January 1998 to 31 December 2001. A Poisson generalised
linear model was used to estimate the specific effects of PM10 on respiratory emergency
hospital admissions for each geographic area. And a pooled effect of PM10 was then
estimated using a meta-analysis approach for the whole city. The results of this study
indicate that the magnitude of the association between PM10 and respiratory emergency
hospital admissions varied across different geographic areas in Brisbane. This
relationship appeared to be stronger in areas with heavy traffic. They found an overall
increase of 1.04 in respiratory emergency hospital admissions associated with an
increase of 10 µg /m3 in PM10 in the single pollutant model. The association was also
statistically significant (an increase of 1.026) after adjusting for O3, but did not appear
to be affected by NO2. The effect estimates of PM10 were generally consistent for three
spatial methods used in this study, but appeared to be underestimated if the spatial
nature of the data was ignored (L Chen et al., 2007). This study only considered other
23
air pollutants as confounding factors but didn’t include other factors (e.g., socio-
economic factors). Additionally, it is subjective to divide the whole Brisbane city into
six geographic areas.
GIS has been increasingly used in air pollution and health research. However, a GIS
was limited to visualization of model results imported from conventional statistical
analysis packages, such as SAS or S-plus; thus a specialized programming designed to
address specific spatial statistical analysis was used from outside GIS (Chung et al.,
2004).
The spatial regression is designed to account for spatial autocorrelation. There is a
positive spatial autocorrelation if the differences between adjacent units are smaller than
the differences between nonadjacent units. On the other hand, a negative spatial
autocorrelation suggests that dissimilar values appear in adjacent (or connected) areas
(Cliff & Ord, 1981).
A spatial regression model was used to model the hospitalization rates for low back
problems by constructing weight matrix using SpaceStat software (Joines et al., 2003).
They found previous studies failed to detect significant sociodemographic spatial
autocorrelation because they did not account for spatial correlation. They also used a
weight matrix by averaging using neighbour values for each location calculated with
different contiguity. The impact of the resulting different number of counties used as
neighbour was examined by plotting autocorrelation statistics against the number of
counties involved in the analysis.
24
Similarly, to account for the possibility that missing or systematically mismeasured risk
factors may be spatially correlated with air pollution, Pope (2002) graphically examined
the correlations of residual mortality with distance between metropolitan areas and
incorporated a spatial random-effects component in the Cox proportional hazards model
constructed using S-Plus. The spatial random component included in the model
provides better estimates of the uncertainty of effect estimates for pollution-related
mortality.
Cressie and Read (1989) applied the fundamentals of time-series analysis to spatial
analysis, whereby “methods of data transformation, detrending, and autocorrelation
plotting modified both to mitigate and to exploit … spatial relationship”. Such spatial
techniques, combined with the current availability of demographic, health and
environmental data referenced to small area units, and GIS, has enabled spatial
approaches to the investigation of the association between air pollution and health
(Cressie & Read, 1989).
Geospatial analysis using GIS has been widely used in studies of health effects
associated with environmental influences (Briggs & Elliott, 1995; Croner et al., 1996;
Vine et al., 1997). However, caution regarding data availability and confidentiality,
boundary overlap issues, map display issues and the need for analysis external to the
GIS software has been advised. Further, daily variation in ambient air pollution is
unlikely to be biased by SES, and hence this does not presents a problem for time series
analysis. However, adjusting for confounding by social class is essential, for spatial
studies (D.J. Jolley et al., 1992a; D. J. Jolley et al., 1992b; 1996; Scoggins et al., 2004).
25
In summary GIS and spatial analysis methods have been used to evaluate the
assessment of the health effect of exposure to air pollution.
2.4 METHODOLOGICAL ISSUES
When performing a spatial analysis, mapping software is essential. As statistical
analysis often begins with a histogram to show a variable’s distribution, spatial analysis
begins with a variable’s spatial distribution and the final result of the analysis is also
often a map.
GIS is a computerized system for input, storage, management, display and analysis of
data that can be precisely linked to a geographic location (O'Sullivan & Unwin, 2003).
Typically, GIS datasets come as layers – there can be a layer for rivers, a layer for roads,
and a layer for post codes – all within a particular geographical boundary. A layer may
consist of one or more features, which include points, lines, or boundaries. Various
layers are superimposed to create a meaningful map. Each GIS layer has two views: a
map view and a data view (O'Sullivan & Unwin, 2003). The map acts as a visual
representation of data, and a particular attribute of the dataset can be displayed on the
map. The data view can be used to create a smaller dataset (or map) from a large dataset
using a query tool. A GIS layer dataset can also be combined with a user dataset to
create a new layer as long as there is a common attribute in the two datasets. For
example, Census 2000 non-GIS data can be joined with a zip code layer to create a new
layer where population can be displayed for each zip code in the USA (Rob, 2003).
Australia population (ABS, 2001) data also can display for each statistical local areas
(SLA).
26
There are four main sources for mapping software. First, there are commercially
available GIS such as MapInfo Professional (MapInfo Corporation, 2003), Maptitude
(Caliper Corporation 2000) and Environmental Systems Research Institute’s (ESRI)
ArcView GIS (ESRI 2000). This type of complete GIS software will be the most
expensive option, but a GIS software package will have the most advanced facilities for
displaying geographic data and will include functions based on years of geographical
research. ESRI offers a Spatial Analyst extension that includes geostatistical functions
such as Inverse Distance Weighting (IDW) and kriging. Second, there is commercially
available mapping software that is not as sophisticated as the GIS software so is less
expensive. For example, ESRI currently offers ArcExplorer as a free product for GIS
data viewing. Third, some statistical software vendors (e.g., SAS and S-PLUS) include
basic mapping functions as either a feature or extension of their software. S-PLUS also
offers S+ Spatial Stats and SPLUS for ArcView GIS as two add-on products to provide
an environment specialized to spatial analysis and integrated with ArcView GIS®. The
final option is to use one of the publicly available programs. Two GIS programs can be
freely downloaded, which include LandView from the U.S. Census Bureau and Epi
Map from the Centers for Disease Control and Prevention (CDC).
Therefore, GIS is an integrated set of computer hardware and software tools designed to
capture, store, retrieve and display spatially-referenced data (Bailey & Gatrell, 1995).
GIS can be used to generate maps and perform some spatial analyses. Some of these
technologies, like the Global position system (GPS) and remote sensing, are often used
to collect geographic data (ESRI, 2006).
27
Many studies have been conducted using daily average estimates of pollutant
concentrations and health outcomes obtained across a city or region (Dab et al., 1996;
Morgan et al., 1998b). Such studies have generally neglected spatial variation, and have
utilised regionally aggregated measures of pollution and health. However, there has
been increasing concern that the air pollution-health relationship may vary with
geographic area within a city. Thus, GIS and spatial modelling have been increasingly
applied in this field.
Spatial estimations of air pollution using interpolation techniques are accessible in a
number of statistical packages. Thus, previous GIS projects required export/import tools
to include statistical results into the map layers. Recent GIS packages include a number
of extensions that can solve the spatial interpolation, which result in better
interoperability and more efficient data management. Among the various methods of
geostatistical methods implemented in GIS (ESRI’s ArcGIS-Geostatistical Analyst),
there are two most common methods to carry out the spatial interpolation of air
pollution: IDW and Kriging. IDW interpolation estimate is a linear combination of the
observed values, inversely weighted by the distances of the observation locations from
the interpolation point (Isaaks & Srivastava, 1989; O'Sullivan & Unwin, 2003). Kriging
method uses variogram to express the spatial variation, and it minimizes the error of
predicted values, which are estimated by spatial distribution of the predicted values. It is
a powerful spatial analytical method (Isaaks & Srivastava, 1989; O'Sullivan & Unwin,
2003).
In epidemiological research, risk mapping often uses GIS in two steps: 1) getting data
into GIS and transforming data; and 2) spatial analysis and spatial statistical analysis.
28
The first step constitutes data pre-processing stage; and the second constitutes data
processing stage.
2.4.1 Data pre-processing stage
Getting data into GIS
Geocoding is increasingly popular as a tool for translating location information into
corresponding latitudes and longitudes in health databases for getting data into GIS.
Automated geocoding, coupled with automatic address standardization and functions,
has helped promote GIS-based analysis of data. Geocoding levels vary depending on the
reference data used (Chung et al., 2004).
Transforming data
Spatial data analysis often requires transformation of data. There are basically two types
of geographic data—vector and raster or grid. Vector data, based on coordinates, comes
as points, lines, or polygons, while raster data, based on grids or pictures, is more
commonly associated with spatial analysis. Points represent a single location and
generally have associated data for that specific event or location. Polygons represent
areas, and most often have aggregated data associated with them. Raster data do not
have direction or inside/outside features, but have a cell size and extent with
information attached to each cell. While the data are numeric, they can represent
measurements or categories. GIS tools are available for converting one format to
another before spatial analysis is taken.
29
Normally data are often on different units, formats, and at different scales. Location
information in health data (e.g., death certificates, hospital discharge, and health care
provider data) is most often available at the level of post code. GIS allows us to
combine all this information in a meaningful way, using overlay, buffering,
geoprocessing, zonal averages, and proportional allocation (Chung et al., 2004).
2.4.2 Data processing stage
Spatial analysis
Spatial analysis here is the study of spatial pattern using the basic GIS operations such
as spatial query and join, buffering, and overlaying. Generally speaking, things that are
closer together tend to be more alike than things that are farther apart. This “local”
spatial structure is a fundamental geographic principle for spatial analysis. Two
questions remained open: how many regions should we consider as local neighbourhood?
And should all of the regions be considered equally? Those questions are within users’
discretion - depending on study questions. Users define local neighbourhoods using
their preferred spatial proximity measures. Once the spatial structure is defined by the
chosen spatial proximity measure, a local spatial analysis can be done. That is, an
observed value for each region can be replaced with a “local” value calculated on the
basis of values for the neighbouring regions with a prespecified contributing weight
(Chung et al., 2004).
Spatial statistical analysis
30
Spatial statistical analysis is the application of statistical theory and techniques to the
description and modelling of spatially referenced data. Statistical methods such as
smoothing and approaches to identify clusters provide objective tools for measuring
data quality and for accounting for data uncertainties in mapping and assessment of
spatial patterns. Various smoothing procedures are used to eliminate variance instability
in the disease rates (or proportions). Observed rates are often extreme when population
at risk is too small (e.g., rural areas) or a disease in question is rare. Its ultimate goal is
to control such high variance, find areas with an excess rate, and interpret its categories.
There are three most popular methods for addressing this issue: (1) Kernel density
smoothing; (2) Empirical Bayes smoothing; and (3) Locally weighted regression.
Kernel Density Smoothing. If disease data have been geographically coded to their
latitude and longitude coordinates, or to quite small areas such as post codes, census
tracts, census block areas, or to other small postal code areas, they can be spatially
aggregated in very flexible ways according to the needs of the user. One popular
technique for creating a continuous map from such “point” data is kernel density
smoothing (Chung et al., 2004). A point dataset is mapped by creating a grid using an
inverse distance weighting function. GIS software (e.g., Spatial Analyst extension to
ArcView) allows users to select the distance or the number of points to be smoothed and
create a grid of user-selected fineness. Each point is averaged with the weighted value
of every other point within a specified distance (e.g., 1 mile) of that point based on
specified weight scheme such as the inverse square of the distance.
31
Talbot et al (2000) used kernel density smoothing. The number of low birth weight
(LBW) as well as the total number of births was aggregated for each zip code in New
York State. All births in a particular zip code were then assigned to the geographical
coordinates of the population-weighted zip code centroid. Then this zip code map was
overlaid with a layer with 1-km spacing of the grid points. The nearest zip code centroid
to the grid point is located. If the number of births is less than the minimum number,
then the next nearest zip code centroid is located and the number of births is added to
those from the previous zip codes. This process is continued until the total number of
births captured is greater than or equal to the minimum number of births. At this point,
the total number of births and LBWs captured in the selected zip codes are assigned to
the grid point (Talbot et al., 2000).
Empirical Bayes Smoothing. Disease mapping literature often uses empirical Bayes
smoothing. Yianakoulias (2003) smoothed fall injury rates in the elderly using empirical
Bayes estimates method. Observed rates are calculated as disease cases/population at
risk. The rates are assumed to follow as binomial (or Poisson) random variable.
Empirical Bayes methods shrink observed rates differentially toward the mean of the
distribution of rates in proportion to their expected variability based on the number of
observations in the small areas (Yiannakoulias et al., 2003). For example, when the
observed rates are based on small populations and the Bayes estimator is close to the
prior mean (or local rate using neighboring values). However, when the population is
large, the Bayes estimator approaches the observed rate.
Locally Weighted Regression (LOESS). Unlike the first two methods, a locally
weighted regression is a nonparametric smoothing method (Chung et al., 2004). In
32
locally weighted regression, points are weighted by proximity to the current value in
question using a kernel. A regression is then computed using the weighted points. It
performs a regression around a point of interest using only data that are “local” to that
point. The estimator variance is minimized when the kernel includes as many points as
can be accommodated by the model. Too large a kernel includes points that degrade the
fit; too small a kernel neglects points that increase confidence in the fit. There are a
number of ways one can set the smoothing parameter. As the parameter decreases, the
regression becomes more global. The variance-based method usually gives the best
performance (Chung et al., 2004).
Pope et al (2002) used nonparametric-smoothed estimates of air pollution-related
mortality using the robust locally weighted regression (LOESS) smoother. The
complexity of the location surface was controlled by a chosen 35% of a span parameter,
which is the proportion of the data used for the local regression. Increasing the span
increases the smoothness of the estimated surface.
2.4.3 Two popular interpolation methods
Deterministic method: IDW
IDW is a method of interpolation that estimates cell values by averaging the values of
sample data points in the neighbourhood of each processing cell (Isaaks & Srivastava,
1989; O'Sullivan & Unwin, 2003). The closer a point is to the centre of the cell being
estimated, the more influence, or weight, it has in the averaging process.
Statistical method: Kriging
33
Kriging is similar to IDW in that it weights the surrounding measured values to derive a
prediction for an unmeasured location (Isaaks & Srivastava, 1989; O'Sullivan & Unwin,
2003). The general formula for both interpolators is formed as a weighted sum of the
data:
where:
Z(si) = the measured value at the ith location.
λi = an unknown weight for the measured value at the ith location.
s0 = the prediction location.
N = the number of measured values.
In IDW, the weight, λi, depends solely on the distance to the prediction location.
However, with the Kriging method, the weights are based not only on the distance
between the measured points and the prediction location but also on the overall spatial
arrangement of the measured points. To use the spatial arrangement in the weights, the
spatial autocorrelation must be quantified. Thus, in ordinary kriging, the weight, λi,
depends on a fitted model to the measured points, the distance to the prediction location,
and the spatial relationships among the measured values around the prediction location
(ESRI, 2007).
Example of using two methods
Figure 1 (2.4) shows an example of a PM2.5 pollution surface created by IDW and
kriging for southern California on 2/10/2003 using ArcGIS 9.0 software, where PM2.5
34
data from 19 monitoring stations were used in the interpolation for southern California
(Wu et al., 2006). The results are similar using both methods.
Brigge (1997) suggest that Kriging estimates are better suited to national scales, while
IDW is more appropriate for regional mapping. There are some studies using one or
both methods on spatial modelling of air pollution in urban areas with GIS (Matejicek,
2005). Kriging has also been widely used in epidemiological studies. For example,
Goovaerts (2006) employed area-to-point Poisson Kriging approaches for spatial
support and population density in the mapping of cancer mortality risk in two areas of
USA (Goovaerts, 2006).
Many air pollution studies have employed these methods based on samples from
monitoring networks to make spatial prediction maps of air pollution concentrations
(Yuval, 2006). Matejicek (2006) used IDW method as a preliminary technique for a
quick look at interpolated data in the form of two-dimensional horizontal layers of
ozone distribution, without assessing prediction errors(Matejicek et al., 2006).
35
Figure 2.1 An example of a PM2.5 pollution surface created by IDW and kriging for
southern California on 2/10/2003.
36
2.5 GAPS IN CURRENT KNOWLEDGE BASE
Evidently from this review, many studies examining air pollution and health outcomes
have been carried out in Australia and overseas, and they have found that, to some extent,
air pollution can lead to an increase in cardiorespiratory morbidity/mortality in major
cities (Daniels et al., 2000; Daniels et al., 2004; Hajat et al., 2007; Lee et al., 2000; Peng
et al., 2005a; 2005b). However, only a small number of studies have considered the
spatial features of the air pollution - health relationship. There are still many knowledge
gaps to be filled as follows:
• How to determine population exposure distribution according to location?
• How to precisely estimate long-term exposure to air pollution at a SLA level
according to the available monitoring data?
• What is the best model to assess the spatiotemporal features of the relationship
between air pollution and health outcomes?
In summary, GIS and spatio-temporal modelling have great potential for environmental
and health outcomes research. The use of modelling techniques is still in its infancy.
Fully integrated and validated spatio-temporal approaches to environmental health
modelling need to be further developed.
37
CHAPTER 3: STUDY DESIGN AND METHOD
This chapter discusses the general study design and methods, as each manuscript
(Chapters 4-6) has its own separate detailed methods section.
The data sets required for this study have spatial attributes associated with either point
locations (monitoring sites) or small areas (SLAs). Sections 3.1 to 3.3 outline the study
design, the procedures to acquire the data, exploratory data analysis, and the protocol for
data analysis. Spatial linkage of the datasets and data analytic methods are described in
sections 3.4 and 3.5.
3.1 STUDY DESIGN
This study applied an ecological longitudinal research design in investigating the impact
of gaseous air pollution on cardiorespiratory mortality using GIS and logistic regression
models. The SLA is used in this study as a basic analytic scale to conduct a spatial
analysis of air pollution and cardiorespiratory mortality in Brisbane. The potential
confounding effects of age, sex and socio-economic factors on the investigation of the
relation between air pollution and cardiorespirator mortality were also considered and
assessed in the study.
3.2 STUDY AREA
38
The city of Brisbane was selected as the research site to assess the long-term air pollution
trends and the spatiotemporal relationship between air pollution and cardiorespiratory
mortality.
Brisbane was chosen because it has the relatively long-term monitoring data on air
pollution, and a close collaborative relation between the research team and the
Queensland Health, Environmental Protection Agency, Australian Bureau of
Meteorology and Australian Bureau of Statistics has been developed. Additionally,
Brisbane is the capital of Queensland with the highest population density in the state. It
is located in a sub-tropic climate area with the latitude 27o29’S and longitude 153o8’E. It
is Australia’s third largest city after Sydney and Melbourne, covering 1326.8 km2 urban
areas with the population size of 883,440 on 1 July 2001 (ABS, 2001).
Brisbane city contains 162 statistical local areas (SLAs). SLA is a medium level spatial
unit in the Australian Standard Geographical Classification (ASGC) system. The ASGC
is a set of hierarchically nested non-overlapping spatial units used by the ABS for
collecting and publishing the Australian Census since 1984. There are changes in
boundaries, names and codes over time, sometimes twice in one year (e.g., editions 7 and
8 were both released in 1989). The SLA is also the basic spatial unit used to collect and
disseminate statistics other than those collected from the Population Census. 162 SLAs
cover the whole of Brisbane city without gaps or overlaps (ABS, 2001).
3.3 DATA COLLECTION
3.3.1 Air pollution data
39
Air pollution data used in this study were provided by the Queensland Environmental
Protection Agency (QEPA). The data comprised the daily average concentrations of
particulate matter less then 10 µm in aerodynamic diameter (PM10), nitrogen dioxide
(NO2), ozone (O3) and sulphur dioxide (SO2) between 1 January 1980 and 31 December
2004 in two monitoring sites (i.e. Eagle farm and Rocklea), and in other available
monitoring sites between 1 January 1996 and 31 December 2004. During this study
period, there were many changes for monitoring sites in South-east Queensland region.
Some sites monitored the whole period and all pollutants, and some only monitored a
few years and a few pollutants (table 3.1).
Table 3.1 Descriptions of monitoring sites in South-east Queensland region
Sites Location Establishment
Area
classification
Pollutants
monitored
Brisbane CBD (qut) 2 George St (QUT) 1995 - commercial O3, NO2, SO2, PM10
Brisbane CBD (sci) 110 George St April 1998 - commercial O3, NO2, SO2
Darra (dar) Ashridge Rd, Darra 1979 - Feb. 2000 industrial PM10 (only 6 day
record)
Deception Bay
(dcb)
Foreshore area
Esplanade
June 1994 - residential O3, NO2
Eagle Farm (eag) Quarantine Centre
of DPI Curtin
Avenue
1978 - light industrial O3, NO2, SO2, PM10
Flinders View (rfv) Telstra Swanbank
Exchange Reif St
Jan. 1993 - residential O3, NO2, SO2, PM10
Fortitude Valley
(val)
Brunswick St,
Fortitude Valley
1978 - 1995 commercial O3, NO2
Helensvale (hel) Discovery Drive,
Helensvale
April 1998 to Oct.
2002
residential O3, NO2, PM10
40
Mount Warren Park
(mwp)
Bardyn Halliday
Drive, Mount
Warren Park
1994 -July 2002 residential O3, NO2
Mountain Creek
(mar)
Mountain Creek
Primary School
2001 - residential O3, NO2, PM10
Mutdapilly (mut) Gimpels Rd 1995 - rural O3, NO2
North Maclean
(nmc)
St Aldwyn Rd 1994 - rural O3, NO2
Pinkenba (pin) Pinkenba Primary
School
2001 - residential
(adjacent to
industry and
airport)
O3, NO2, SO2, PM10
Rocklea (roc) Oxley Common,
Sherwood Rd
1978 - residential / light
industrial
O3, NO2, PM10
South Brisbane
(sbr)
Road reserve,
South East
Freeway
2001 - commercial NO2, PM10
Springwood (spr) Springwood State
High School
March 1999 - residential O3, NO2, SO2, PM10
Woolloongabba
(woo)
PA Hospital,
Ipswich Rd
August 1976 - commercial PM10
Wynnum (wyn) Wynnum North Rd Jan. 1999 - May.
2002
residential
(adjacent to
industrial)
O3, NO2, SO2, PM10
Zillmere (zil) Pineapple St,
Zillmere
1995 – Oct. 1999 light industrial O3, NO2
The QEPA has a network of air quality monitoring stations in South-east Queensland,
which measure levels of pollutants in air. Daily air quality measurements have been
taken since 1978. The monitoring program provides information to identify long-term
trends in air quality across SEQ and helps assess the effectiveness of air quality
management strategies. Measurements are compared with national and overseas air
41
quality guidelines to determine current air quality (QEPA, 2007). 13 monitoring sites
were operating in South-east Queensland of EPA in 2004, including 8 stations of them
within Brisbane and 5 around Brisbane (Figure 3.1). South-east Queensland is a large
region with rapid development and population growth. Centred around Brisbane and its
suburbs, the region stretches from the Gold Coast north to the Sunshine Coast and west
to Toowoomba (QEPA, 2007).
Figure 3.1 Locations of monitoring sites in South-east Queensland region of EPA
3.3.2 Mortality data
Mortality data between 1 January 1996 and 31 December 2004 in Brisbane city were
provided by the Office of Economic and Statistical Research of the Queensland
Treasury. The data included day, month, year of death, sex, age, SLA of residence and
cause of death. The cause-specific deaths were categorised according to the
42
International Classification of Diseases Version 9 (ICD-9 code) (used until July 1999)
or Version 10 (ICD-10 code) and were defined as respiratory disease (RD) (ICD 9: 460-
519; ICD: 10 J00 – J99), cardiovascular diseases (CVD) (ICD 9: 390- 429; ICD: 10 I00
– I99) and cardiorespiratory diseases (CRD) (including both respiratory and
cardiovascular diseases). The annual number of deaths from different causes (including
RD, CVD, CRD and all causes excluding external causes) was listed in Figure 3.2.
200420032002200120001999199819971996
Year
6000
4000
2000
0
Dea
ths
num
ber
All causes (excluding external causes)
CRDCVDRD
Figure 3.2 The distribution of deaths from RD, CVD, CRD and all causes excluding
external causes between 1996 and 2004 in Brisbane city
3.3.3 Population and socio-economic data
Population data were obtained from the Australian Bureau of Statistics (ABS) for each
SLA of Brisbane city. The latest census data included sex and age (0-4, 5-9, …, 100+) by
SLA in 2001. There was substantial difference in the population size between SLAs. On
average, an SLA contained 5500 people, with a range from 250 to 15,631 people (ABS,
2001). Figure 3.3 shows the distribution of population in Brisbane city at the SLA level,
using the 2001 Census data, as well as digital statistical boundaries, base map data,
MapInfo (MapInfo Corporation 2003) and Microsoft Access softwares. During this study
43
period, the population of Brisbane city were 824,489 on 30 June 1996 and 958,504 on 30
June 2004, increasing by 16.25%.
Figure 3.3 The distributions of population in Brisbane city at the SLA level in 2001.
We also obtained the data on Socio-Economic Indexes for Areas (SEIFA) from the ABS
(ABS, 2001). It contained four summary indices (Advantage/Disadvantage Index,
Disadvantage Index, Economic Resources Index, and Education and Occupation Index)
to measure different aspects of socio-economic conditions by SLA. The indexes are all
constructed so that the higher values indicate higher advantage levels. High scores on
the Index of Relative Socio-Economic Disadvantage occur when the area has few
families of low income and few people with little training and in unskilled occupations.
Low scores on the index occur when the area has many low income families and people
with little training and in unskilled occupations. In this study, we selected the
Disadvantage Index to represent the socioeconomic status of each SLA, because of two
44
major reasons: firstly, it focuses on low-income earners, relatively lower educational
attainment, high unemployment and other variables reflecting disadvantage; secondly
this index was more strongly associated with cardiorespiratory mortality (rs = -0.19)
than other three indices of SEIFA (rs | ≤ | 0.18).
There were also some other confounding factors (e.g., smoking and occupational
exposure to air pollution; obesity rates and nutrition intake by area) should be
considered, but we do not have these information on SLA level.
3.4 DATA LINKAGES
For 162 SLAs within Brisbane, the digital base map data sets for constructing the GIS
database used in this study were obtained primarily from the ABS; and these data were
manipulated to facilitate the accurate identification of the spatial locations of SLA, and
their linkages with other data layers. The places of monitoring sites were geo-coded to
the digital base maps of localities utilising MapInfo and Microsoft Access software.
3.5 DATA ANALYSIS
3.5.1 Standardised mortality rates (SMR)
In order to describe and compare the spatial patterns of RD, CVD and CRD mortality
across SLAs, the direct method (i.e. using the Brisbane population as a reference) was
used to calculate the standardized mortality rate (SMR) for each SLA, adjusted for
45
differences in the age and sex distributions among SLAs (Selvin, 2001). The equation
for calculating a SMR is
∑∑= )(comparisoni
i
de
SMR
Where ∑ ie is the total number of expected cases generated using the reference
population rates for each SLA; ∑ )(comparisonid is the total population in the comparison
group. In this study, firstly, we calculated age-specific rates per 100,000 of RD, CVD
and CRD deaths (age group: 0-14, 15-64, 65-74 and 75+) for each SLA; then we
calculated the expected number of deaths in each age group by SLA; and finally we
summed the expected number of deaths and divided the Brisbane population to get
SMR per 100,000 for each SLA. A step by step example of how to calculate SMR in 5
SLAs of Brisbane is provided in Appendix II. In this study, we used same method, but
calculated SMR for whole Brisbane city including 162 SLAs.
3.5.2 Inverse distance weighted (IDW) interpolation
To assess the association between exposure to air pollution and cardiorespiratory
mortality, a spatial distribution model was developed using an inverse distance
weighted (IDW) interpolation for the annual average concentrations of air pollution at a
SLA level (Isaaks & Srivastava, 1989). This method is based on the assumption that the
interpolating surface should be influenced mostly by the nearby points and less by the
more distant points (Isaaks & Srivastava, 1989). It estimated the concentrations of air
pollution for a SLA without a monitoring site as a distance-weighted sum of the values
observed in some surrounding neighbourhoods. Consequently, an output surface of the
annual average of air pollution by SLA was produced for each year. Because the
46
mortality data were provided and summarized at the SLA level, the annual average of
air pollution was also produced for each SLA using GIS. It was calculated by adding air
pollution levels of all the grid cells (a grid cell size: 4 km2) in each SLA together and
then dividing the sum by the total number of cells in that specific SLA.
3.5.3 Visualisation of air pollution and mortality data
The original air pollution data (i.e. PM10, NO2, O3 and SO2) were point measurements
recorded at each monitoring site in the south-east Queensland region. Most of the SLAs
did not have a monitoring station.
MapInfo Professional software (MapInfo Corporation 2003) was used to display the
spatial distributions of cardiorespiratory deaths and air pollution data. The locations of
cardiorespiratory deaths and air pollution data were geo-coded to the digital base maps
of localities using MapInfo (MapInfo Corporation 2003) and Microsoft Access software
(Microsoft Office Access 2003).
3.5.4 Exploratory data analysis
Statistical analyses were conducted using daily data on air pollution concentrations in
monitoring sites and cardiorespiratory deaths for each SLA in Brisbane city. Then these
data were aggregated to the annual average levels of air pollution and cardiorespiratory
mortality at a SLA level. Missing values of air pollution concentrations were dealt with
using the listwise deletion approach (known as a complete case analysis).
47
On air pollution trends study for two monitoring sites (i.e. Eagle Farm and Rocklea), the
missing value were 9% with a range from 6.8% (Rocklea) to 11.2% (Eagle Farm) for
NO2, and 5.4% with a range from 5.2% (Rocklea) to 5.7% (Eagle Farm) for O3 during
the study period (1 Jan. 1980 to 31 Dec. 2003). The missing value for PM10 since the
sites have daily record was 3.2% with a range from 0.4 in Eagle Farm (1 Feb. 1998 to
31 Dec. 2003) to 6% in Rocklea (14 Mar. 1996 to 31 Dec. 2003), and for six day record
the missing value of PM10 was 5.5% with a range from 4.4% in Eagle Farm (1 Jan. 1994
to 31 Jan. 1998) to6.7% in Rocklea (26 Jan. 1986 to 13 Mar. 1996). The percentage of
air pollution data available in each monitoring site between Jan. 1996 and Dec. 2004
was summarised in Table 3.2. This percentage was calculated by available monitoring
period for each site. The annual average of air pollutants were used in this study.
Spearman rank-order correlations between air pollutants (i.e. PM10, NO2, O3, and SO2)
were calculated after outliers were removed (Kuzma, 1992). The outliers were defined
as those values outside the mean ± 3SD (standard deviation).
Table 3.2 Available daily max 1-hour O3, NO2 SO2 and 24-hour average PM10 summary (1996 – 2004)
O3 NO2 SO2 PM10
Sites % No of days %
No of days %
No of days %
No of days
Brisbane CBD (qut) 97.0 2141 93.5 2064 93.6 1587 95.8 3150 Brisbane CBD (sci) 98.8 1412 98.8 1412 94.4 2038 na na Deception Bay (dcb) 96.8 3184 94.3 3099 na na na na Eagle Farm (eag) 99.2 3261 96.8 3183 96.8 3182 79.2* 2605 Flinders View (rfv) 99.1 3259 98.1 3224 97 3188 97.3 2400 Helensvale (hel) 98.7 1606 98 1589 na na 97.3 1580 Mount Warren Park (mwp) 99.6 2366 92.8 2111 na na na na Mountain Creek (mar) 99.9 1065 99.2 1058 na na 95.7 1219 Mutdapilly (mut) 96.6 3176 94.4 3104 na na na na North Maclean (nmc) 99.2 3263 98.9 3251 na na na na Pinkenba (pin) 94.4 419 94.5 432 95 433 87.1 398 Rocklea (roc) 98.4 3235 97.2 3196 na na 92.3 3034 South Brisbane (sbr) na na 98.1 1114 na na 95.5 1092 Springwood (spr) 95.3 2032 95.4 2035 95.7 2041 98 2088 Woolloongabba (woo) na na na na na na 68.9* 2264 Wynnum (wyn) 100 505 100 505 100 505 95.9 1087
Zillmere (zil) 99.9 1400 94.1 1318 na na na na
48
* There were six day records before 31 Jan. 1998.
3.5.5 Multivariable logistic regression model
In this study, the relationships between annual average concentrations of air pollution
and cardiorespiratory mortality within 162 SLAs were assessed using logistic regression
model after adjustment for confounding effects of age, sex and social disadvantage in
both single pollutant and multiple pollutants models. The model is described as follows:
P (Deaths) = Exp(X) / [1+Exp(X)]
Where: P (Deaths) is the probability of dying or number of deaths divided by the
population for each SLA,
X = β0 + β1 (pollutant) + β2 (age) + β3 (sex) + β4 (SEIFA) for single pollutant
model or
X = β0 + β1 (PM10) + β2 (NO2) + β3 (O3) + β4 (SO2) + β5 (age) + β6 (sex) + β7
(SEIFA) for multiple pollutants model
All the analyses were conducted using SAS statistical software package version 9.1
(SAS Institute Inc, 2002).
In summary, a large scale spatial approach was used in this study. The aims were to
examine the spatial variation in the relationship between long-term exposure to gaseous
air pollutants (including nitrogen dioxide (NO2), ozone (O3) and sulphur dioxide (SO2)),
and cardiorespiratory mortality in Brisbane. The key findings were presented in each of
the following chapters.
49
CHAPTER 4: AIR POLLUTION TRENDS IN BRISBANE,
AUSTRALIA, BETWEEN 1980 AND 2003
Citation:
Xiao-Yu Wang, Shilu Tong, Ken Verrall, Rod Gerber and Rodney Wolff. Air pollution
trends in Brisbane, Australia, between 1980 and 2003. Clean Air and Environment
Quality, February 2006, 40 (1): 34-39.
Contribution of authors:
X. Y. Wang was the principal author of the manuscript, performed all data analysis and
wrote the manuscript. ST and KV contributed to the research design, collection of data,
development of analytical protocol, assisted with interpretation of the results and writing of
the manuscript. RG and RW contributed to the manuscript in terms of providing feedback
on the analyses and initial drafts.
50
ABSTRACT
This study aimed to evaluate long-term air pollution trends in Brisbane, Australia,
between 1980 and 2003. We obtained air pollution data recorded at the two monitoring
sites (Eagle Farm and Rocklea) from the Queensland Environmental Protection Agency.
Graphical and polynomial regression analyses were carried out on daily maximum 1-
hour average or daily 24-hour average concentrations of NO2, O3, and PM10. The results
show that there were significant up-and-down features for NO2 and O3 concentrations in
both sites during 1980 – 1993 whereas similar trend was observed with PM10
concentrations in Rocklea during 1985 – 1993. However, NO2 concentrations were
stable for both sites in recent years and a similar trend was observed with O3 and PM10
concentrations for both sites during 1994 – 2003. Annual average of daily 24-hour PM10,
NO2 and O3 concentrations fluctuated from 15.6 to 38.7µg m-3, 6.0 to 16.1ppb and 7.9
to 16.3ppb for Rocklea; from 17.5 to 29.7µg m-3, 8.9 to 14.9ppb and 11.8 to 18.2ppb for
Eagle Farm during the study period, respectively. Annual average daily maximum 1-
hour NO2 and O3 fluctuated from 15.5 to 31.8 ppb and 18.3 to 37.0 ppb for Rocklea;
from 20.0 to 31.8 ppb and 23.4 to 35.1 ppb for Eagle Farm during the study period,
respectively. For all the target compounds, Rocklea recorded a substantially higher
number of days with concentrations above the relevant daily maximum 1-hour or 24-
hour standards than that in Eagle Farm.
Keywords: Frequency; Nitrogen dioxide; Ozone; Particulate matter; Season
51
4.1 INTRODUCTION
Brisbane, the capital of Queensland, is the fastest growing city in Australia. With a
population of 1.6 million people, it is Australia’s third largest city after Sydney and
Melbourne (Wikipedia, 2005). With predominantly hot summers and mild winters, the
Brisbane climate suits outdoor activities with a sub-tropical lifestyle.
The air we breathe is a mixture of gases and small solid and liquid particles. Some of
the substances come from natural sources, while others are caused by human activities
such as the use of motor vehicles, industry, domestic activities, and business. Air
pollution occurs when the air contains substances in quantities that could harm the
comfort or health of humans and animals, or could damage plants or materials (QEPA,
2005). These air pollutants can be either particles, liquids or gas. Air pollution is one of
the major health hazards in metropolitan cities world wide.
Like many other developed countries in the world, the Australian government has paid a
great deal of attention to air pollution over the past decades. In 1976, the Queensland
Environmental Protection Agency (QEPA) established a network of air quality
monitoring stations in South East Queensland. Thirteen monitoring sites are located in
inner city, rural, industrial and residential areas. The current air quality monitoring
stations are monitoring a range of air pollutants including carbon monoxide (CO),
nitrogen dioxide (NO2), ozone (O3), sulphur dioxide (SO2), airborne particulate matter
with diameter ≤ 10 µm (PM10) and 2.5 µm (PM2.5), as well as visibility reduction.
52
Air pollution causes a variety of health problems. It has been ranked as one of the top
ten causes of the global burden of disease and injury by the World Health Organisation
(Murray & Lopez, 1996). Many studies have investigated the health impacts of air
pollution in Australia and overseas (Chan et al., 1997; Denison et al., 2001; Howie et al.,
2005; H Kan & Chen, 2003; Krewski et al., 2005; Morgan et al., 1998a; Morgan et al.,
1998b; Petroeschevsky et al., 2001; R W Simpson et al., 1997; Yang et al., 2004). There
are however, limited empirical data available on the spatial variation in the long-term
trends of air pollution in metropolitan areas (K S Chen et al., 2004; Jo et al., 2000; Wise
& Comrie, 2005).
4.2 DATA COLLECTION AND ANALYSIS
The data presented in this study were collected from two major air pollutant monitoring
sites in Brisbane (Figure 4.1). As these sites are among the earliest established in south-
east Queensland, they have the most complete air pollutants records (QEPA, 2001):
1) Eagle Farm is located 8 km east of the Brisbane Central Business District (CBD)
and is in a light industrial area (latitude -27.4381, longitude 153.0794). Pollutants
monitored included O3, NO2, SO2, PM10 and visibility-reducing particles.
2) Rocklea is located 14 km south of the Brisbane CBD and is in a light industrial and
residential area (latitude -27.5434, longitude 152.9986). Pollutants monitored
included O3, NO2, PM10, PM2.5 and visibility-reducing particles.
The Rocklea monitoring site moved from a roadside location to an open paddock in
1992.
53
Figure 4.1 The locations of Eagle Farm and Rocklea monitoring stations in Brisbane.
The data comprising daily maximum 1-hour and 24-hour average concentrations of
NO2, O3, and PM10 were recorded at both Eagle Farm and Rocklea monitoring sites. We
used the daily data on NO2 and O3 from 1980 to 2003 for both sites; PM10 data from
1994 to 2003 (once per six days recorded between 2 May 1994 and 29 January 1998,
daily recorded after 1 February 1998) in Eagle Farm; and from 1986 to 2003(once per
six days record between 26 January 1986 and 10 March 1996, daily record after 14
March 1996) in Rocklea. To examine the pattern of temporal variation, observation
values were averaged each year during the study period. The following months were
selected for each season: summer: December – February; autumn: March – May;
winter: June – August and spring: September – November.
Statistical analyses were carried out on daily maximum 1-hour or daily 24-hour average
concentrations for NO2, O3, and daily 24-hour average for PM10. Spearman rank-order
correlations between air pollutants (i.e. NO2, O3, and PM10) were calculated after
54
outliers were removed (Kuzma, 1992). The outliers were defined as those values outside
the mean ± 3SD (standard deviation).
A polynomial regression model was used to fit the annual average of air pollution
trends. In order to identify the most suitable model to reflect the trends of all air
pollutants over the study period, a variety of polynomial regression models were
examined.
4.3 RESULTS AND DISCUSSION
NO2
The annual and seasonal average of the daily maximum 1-hour and 24-hour average
NO2 concentrations in Eagle Farm and Rocklea were displayed in Figures 4.2A, 4.2B
and 4.3A, respectively.
It was observed that the levels of NO2 showed no significant change for recent years,
with more significant up-and-down features in other periods in both sites (Figures 4.2A
and 4.2B). The annual average levels of daily maximum 1-hour and 24-hour average
NO2 concentrations fluctuated from 15.5 to 31.8ppb and from 6.0 to 16.1ppb during
1980 – 2003 in Eagle Farm and Rocklea stations, respectively.
The overall seasonal pattern of daily maximum 1-hour NO2 concentrations was similar
to that of daily 24-hour average NO2 concentrations in both sites. They all increased
during the winter with the peak between July and August (30.1 – 33.3ppb for the daily
maximum 1-hour and 14.2 – 16.0ppb for 24-hour), and declined gradually with the
55
lowest concentrations recorded between January and February during the summer (14.7
– 16.4ppb for the daily maximum 1-hour and 6.9 – 7.1ppb for 24-houe) (Figure 4.3A).
Eagle Farm Rocklea
A
0
15
30
45
1980 1984 1988 1992 1996 2000
NO 2 (
ppb)
1-h24-h
B
0
15
30
45
1980 1984 1988 1992 1996 2000
NO 2 (
ppb)
1-h24-h
C
0
15
30
45
1980 1984 1988 1992 1996 2000
O3 (
ppb)
1-h24-h
D
0
15
30
45
1980 1984 1988 1992 1996 2000
O3 (
ppb)
1-h24-h
E
0
15
30
45
1994 1996 1998 2000 2002
PM10
(µg
m -3
)
24-h
F
0
15
30
45
1986 1989 1992 1995 1998 2001
PM10
(µg
m -3
)
24-h
Figure 4.2 Annual means of daily maximum 1-h or 24-h concentrations of NO2, O3
between 1980 and 2003, PM10 data available for Eagle Farm between 1994 and 2003
and for Rocklea between 1986 and 2003. The line indicates trendline of three air
pollutant concentrations.
56
A
0
15
30
45
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
NO
2 (p
pb)
1-h (eag) 1-h (roc) 24-h (eag) 24-h (roc)
B
0
15
30
45
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
O3
(ppb
)
1-h (eag) 1-h (roc) 24-h (eag) 24-h (roc)
C
0
15
30
45
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
PM10
(ug/
m3)
24-h (eag) 24-h (roc)
Figure 4.3 Seasonal (monthly) averages of daily maximum 1-h or 24-h mean levels
measured for NO2, O3 and PM10 during the study period, for Eagle Farm (eag) and
Rocklea (roc).
The trends of NO2 concentrations were fitted using the polynomial regression model
(Figures 4.2A and 4.2B). The R-squared values were 0.61 and 0.35 for daily
maximum1-hour and 24-hour average NO2 concentrations in Eagle Farm site; 0.79 and
0.86 in Rocklea site, respectively. This regression model was the most suitable for
describing the long-term trends of NO2 comparing with other models (e.g. logarithmic,
exponential or lower order in polynomial regression) for both sites (Table 4.1).
57
Nitrogen oxides (NOx) analysers report NO2 as the difference between total nitrogen
oxides and nitric oxide which requires splitting the sample stream in two and passing
one stream through a reduction catalyst. The more sophisticated instruments which was
introduced to the air monitoring network in 1994 are able to record NO2 levels with less
error and inter-annual fluctuations in observed NO2 concentrations has been reduced. A
downward trend in NO2 levels at both Eagle Farm and Rocklea occurred during the
1990s with a slight upward trend developing since 2000.
Table 4.1 The R-Squared values of four regression models for two monitoring sites
(Eag: Eagle Farm, Roc: Rocklea)
Polynomial Linear a Logarithmic a Exponential a
Order Eag Roc Eag Roc Eag Roc Eag Roc
NO2 1-h 2 0.38 0.39 3 0.39 0.41 4 0.59 0.64 5 0.61 0.78 6 0.61 0.79 0.37 0.06 0.35 0.01 0.39 0.05
NO2 24-h 2 0.05 0.38 3 0.1 0.39 4 0.32 0.72 5 0.33 0.85 6 0.35 0.86 0.05 0.03 0.07 0 0.05 0.02
O3 1-h 2 0.16 0.36 3 0.24 0.54 4 0.25 0.58 5 0.25 0.62 6 0.32 0.69 0.03 0.23 0.11 0.08 0.02 0.19
O3 24-h 2 0.31 0.4 3 0.43 0.47 4 0.44 0.52 5 0.44 0.53 6 0.45 0.63 0.2 0.24 0.35 0.09 0.18 0.2
PM10 24-h 2 0.55 0.73 3 0.89 0.74 4 0.9 0.83 5 0.92 0.92
6 0.92 0.92 0.12 0.67 0.34 0.79 0.1 0.73
58
a the order is 1.
O3
The levels of the annual average of daily maximum 1-hour and 24-hour average O3
concentrations in Eagle Farm and Rocklea were displayed in Figures 4.2C and 4.2D,
respectively. It was puzzling to note that the concentration of O3 slumped around 1990,
more markedly at Rocklea, but steadily increased since then. Rainfall was above
average in 1991 and consequent reductions in bushfire episodes may have reduced the
levels of precursors for O3 formation. There were frequent bushfires events in 1996 and
again in 2002-3 and this seems to be reflected in observed O3 concentrations,
particularly at the more inland site, Rocklea. Prior to 1992, O3 concentrations at
Rocklea were usually lower than those at Eagle Farm but in recent years higher
concentrations have been observed at Rocklea. This may be due to moving the Rocklea
monitoring station from a roadside location to an open paddock.
During the study period, the annual average values of daily maximum 1-hour and 24-
hour average O3 concentrations fluctuated from 18.3 to 37.0ppb and from 7.9 to
16.3ppb in Rocklea, respectively (Figure 4.2D). The trends of O3 concentrations were
also fitted using polynomial regression model. The R-squared values were 0.69 and 0.63
for daily maximum1-hour and 24-hour O3 concentrations in Rocklea, respectively
(Table 4.1).
When compared with Rocklea, no increasing level of annual average was observed for
both daily maximum 1-hour and 24-hour O3 concentrations in Eagle Farm, which
fluctuated between 1980 and 1995 (Figure 4.2C). The annual average values of daily
59
maximum 1-hour and 24-hour average O3 concentrations fluctuated from 23.4 to
35.1ppb and from 11.8 to 18.2ppb in Eagle Farm, respectively. The R-squared values of
annual average trends were 0.32 and 0.45 for daily maximum1-hour and 24-hour O3
concentrations, respectively (Table 4.1).
The seasonal (monthly) averages of the daily maximum 1-hour and 24-hour mean for
O3 concentrations were similar at two stations (Figure 4.3B). There were also
differences between the seasons. The daily maximum 1-hour and 24-hour mean O3
concentrations slowly increased during the late winter and spring, with their peak
attained between September and October (34.5 – 35.1ppb for the daily maximum1-hour
and 15.3 – 19.2ppb for 24-hour mean). The values then decreased gradually with the
lowest concentrations being recorded between May and June for both sites, (24.5 –
26.4ppb for maximum 1-hour and 9.8 – 11.0ppb for 24-hour mean).
Photochemical smog is formed by reactions involving NOx, volatile organic compounds
and sunlight. O3 is a major component of this smog and is used as an indicator. Most
(but not all) O3 events have occurred when there is bushfire smoke present. Other
events have occurred on meteorologically conducive days fuelled only by normal urban
precursors. The presence of reactive hydrocarbons in the atmosphere can speed up the
photochemical reactions. The government has recently sought to reduce evaporative
emissions of hydrocarbons by requiring lower vapour pressure of
petrol sold in Brisbane during the summer months.
PM10
60
PM10 particles are generated by a wide range of natural processes and human activities,
including windblown dust, industrial processes, motor vehicle emissions and bush fires
around Brisbane (QEPA, 2005).
The annual average of the daily 24-hour PM10 concentrations showed a well-defined
decreasing in Rocklea over the study period, except for some fluctuations in 1991, 1994
and 2002. The values of the daily 24-hour PM10 concentrations varied from 15.6 to
38.7µg m-3 in Rocklea (Figure 4.2F). The R-squared value of annual average trend was
0.92 for the daily 24-hour PM10 concentrations in Rocklea (Table 4.1).
In Eagle Farm, PM10 gradually decreased from 1994 to 1997 and then levelled off
although there were fluctuations. The daily 24-hour PM10 concentrations ranged from
17.5 to 29.7µg m-3 (Figure 4.2E). The R-squared value of annual average trend was
also 0.92 for the daily 24-hour PM10 concentrations in Eagle Farm. Therefore this model
was the most reliable for the daily 24-hour PM10 concentrations in both sites (Table
4.1).
The seasonal (monthly) averages of daily 24-hour mean PM10 concentrations show the
similar pattern in two sites (Figure 4.3C). They both gradually increased during the
winter and spring, with the peak between September and October (22.5 – 25.9µg m-3),
and then decreased gradually with the lowest concentrations being recorded in April
(16.5 – 16.7µg m-3).
NEPM standards
61
The days that the maximum 1 hour NO2 and O3, and 24 hour average PM10
concentrations exceeded the National Environment Protection Measure (NEPM) for
Ambient Air Quality standards were shown in Figure 4.4.
The standard of the NEPM is 120ppb for the daily maximum 1-hour NO2
concentrations. This standard was exceeded once at Eagle Farm (122ppb on 1/07/1983),
and four times at Rocklea (the highset value was 160ppb on 1/07/1986). The NO2 levels
measured in recent years have been well below the NEPM standard (Figures 4.4A and
4.4B).
For the daily maximum1-hour concentrations in Rocklea, the values of O3 exceeded the
1-h NEPM standard of 100ppb fifteen times (the highset values were 135ppb on
3/12/1996 and 5/03/1999) (Figure 4.4D). At Eagle Farm, only five daily maximum 1-
hour O3 records exceeded the 1-h standard of 100ppb (the highset value: 145ppb on
19/01/1989) (Figure 4.4C).
In addition, 31 records of the daily 24-hour PM10 exceeded the 50µg m-3 NEPM
standard (the highset value was 146.7µg m-3 on 23/10/2002) in Eagle Farm, and 43 in
Rocklea (the highset value was 177.2µg m-3 on 23/10/2002) over the observed period
(Figures 4.4E and 4.4F).
Excesses of the NEPM standard for O3 occurred occasionally, usually when suitable
weather conditions coincided with extra emissions from bushfires. However, with an
increase in urban growth and motor vehicle use, the number of days with high NO2
levels and photochemical smog could become more frequent. A dust storm passed over
62
the east coast of Australia during 23-24 October 2002. This has caused excesses of the
NEPM standard for PM10 during this period.
The frequency (i.e. percentage) was determined by calculating the total number of days
with concentrations exceeding the relevant standard over the total number of sample
days during the specified time period.
Eagle Farm Rocklea
A
0
50
100
150
200
1/01
/80
1/01
/83
1/01
/86
1/01
/89
1/01
/92
1/01
/95
1/01
/98
1/01
/01
NO 2 m
ax 1
-hr (
ppb)
B
0
50
100
150
200
1/01
/80
1/01
/83
1/01
/86
1/01
/89
1/01
/92
1/01
/95
1/01
/98
1/01
/01
NO 2 m
ax 1
-hr (
ppb)
C
0
50
100
150
200
1/01
/80
1/01
/83
1/01
/86
1/01
/89
1/01
/92
1/01
/95
1/01
/98
1/01
/01O
3 max
1-h
r (pp
b)
D
0
50
100
150
200
1/01
/80
1/01
/83
1/01
/86
1/01
/89
1/01
/92
1/01
/95
1/01
/98
1/01
/01
O3 m
ax 1
-hr (
ppb)
E
0
50
100
150
200
1/01
/94
1/01
/96
1/01
/98
1/01
/00
1/01
/02
PM 10
24-
hr (µ
g m
-3)
F
0
50
100
150
200
1/01
/86
1/01
/89
1/01
/92
1/01
/95
1/01
/98
1/01
/01
PM 10
24-
hr (µ
g m
-3)
Figure 4.4 Daily maximum 1-h NO2, O3 and daily 24-h PM10 mean levels during the
study period. The broken line indicates the current air quality NEPM standard.
63
Table 4.2 shows the frequencies of daily maximum 1-hour NO2 and O3 concentrations
were 0.01% and 0.06% in Eagle Farm; 0.05% and 0.18% in Rocklea. The frequencies of
daily 24-hour PM10 concentrations were 1.33% in Eagle Farm and 1.32% in Rocklea,
respectively.
Therefore, Rocklea recorded a substantially higher frequency of days with
concentrations above the relevant daily maximum1-hour standards compared to Eagle
Farm for NO2 and O3 target compounds, while the frequency for 24-hour PM10
concentrations was similar in two sites.
Table 4.2 Frequency of days with NO2, O3 and PM10 concentrations exceeding max 1-h
or 24-h standards at the two selected sites over observed periods
Frequency of days (%)
Site Compound 1-h standarda 24-h standardb Eagle Farm NO2 0.01 _ O3 0.06 _ PM10 _ 1.33 Rocklea NO2 0.05 _ O3 0.18 _ PM10 _ 1.32
a 1-h standard: 120 ppb, NO2; 100 ppb, O3; and not to be exceeded more than one day per year.
b 24-h standard 50 µg m-3, PM10; and not to be exceeded more than five days per year.
Relationship between pollutants and between sites
Table 4.3 shows a high degree of correlation between the daily maximum1-hour and 24-
hour concentrations of NO2 (rs= 0.88 – 0.89); and O3 (rs= 0.67 - 0.75) at both sites.
64
Table 4.3 Correlation (no. of observations) among daily air pollutant variables for two
monitoring sites
NO2 1-h NO2 24-h O3 1-h O3 24-h
Rocklea NO2 24-h 0.887**
(7719)
O3 1-h 0.054** 0.009
(7730) (7452)
O3 24-h -0.178** -0.277** 0.754**
(7562) (7476) (7891)
PM10 24-h 0.333** 0.408** 0.253** 0.02
(3103) (3051) (3088) (3079)
Eagle Farm NO2 24-h 0.883**
(7391)
O3 1-h 0.175** 0.116**
(7465) (7255)
O3 24-h -0.103** -0.254** 0.669**
(7366) (7312) (7910)
PM10 24-h 0.298** 0.309** 0.214** 0.034
(2267) (2253) (2267) (2272)
** Correlation is significant at p<0.01 level (2-tailed).
65
Daily 24-hour PM10 concentrations were weakly associated with the daily maximum1-
hour and 24-hour concentrations of NO2 (rs= 0.30 – 0.41) and with the daily maximum
1-hour concentrations of O3 (rs= 0.21 – 0.25) in two sites (Table 4.3).
The daily 24-hour and maximum1-hour concentrations of NO2 show a weaker
correlation with the daily maximum1-hour concentrations of O3 (rs= 0.12 and 0.18,
respectively) in Eagle Farm.
The daily maximum1-hour and 24-hour concentrations of NO2 show an inverse
correlation with the daily 24-hour concentrations of O3 (rs= -0.28 to -0.10) in two sites
(Table 4.3).
Table 4.4 shows the correlation between all compounds in both monitoring sites, from
high to low successively: the daily 24-hour PM10 (rs= 0.75); the daily maximum1-hour
and 24-hour concentrations of O3 (rs= 0.61 and 0.62, respectively); and the daily
maximum1-hour and 24-hour concentrations of NO2 (rs= 0.54 and 0.55, respectively).
Table 4.4 Correlation (no. of observations) between two monitoring sites for daily air
pollutants
NO2 1-h
NO2 24-h
O3 1-h
O3 24-h
PM10 24-h
Coefficient
0.55
0.54
0.62
0.61
0.75
(7128) (6695) (7654) (7372) (2215)
All correlations were significant at p<0.01 level (2-tailed).
4.4 CONCLUSIONS
66
Air quality has been measured in south-east Queensland since 1978. The present study
determined the long-term annual and seasonal trends of air pollution in two major
monitoring sites of Brisbane.
There was no significant change in NO2 concentrations over recent years but
more significant up-and-down features were observed in other periods at both
sites; O3 concentrations in Rocklea were usually lower than those at Eagle Farm
but higher in recent years; PM10 concentrations have slightly decreased between
1994 and 1997, and changed little for recent years in Eagle Farm but wavily
decreased in Rocklea. Between two monitoring sites, there appeared to be spatial
variation for most air pollutants examined in this study, and most of these
pollutants were only moderately correlated each other.
Daily 24-hour concentration of PM10 wavily decreased from 38.7 to 15.6µg m-3
during 1986~1999 in Rocklea and from 29.7 to 17.5 µg m-3 during 1994~1997 in
Eagle Farm, and then levelled off although there were fluctuations.
Daily 24-hour concentration of NO2 and O3 were overall stable between 1980
and 2003. The annual average daily maximum1-hour concentrations of NO2 and
O3 fluctuated from 15.5 to 31.8 ppb and from 18.3 to 37.0 ppb, respectively,
during 1980~2003. The variability for both NO2 and O3 was more obvious in
Rocklea.
For seasonal distributions, PM10 and O3 levels were higher in spring; and NO2
levels were higher in winter.
For all the target compounds, Rocklea recorded a substantially higher number of
days with concentrations above the relevant daily maximum 1-hour or 24-hour
standards compared to Eagle Farm.
67
ACKNOWLEDGMENTS
The data on air pollution trends were provided by the Queensland Environmental
Protection Agency. This study was partly funded by the Australian Research Council
(#DP0211869).
68
REFERENCES
Chan, Y.C., Simpson, R.W., Mctainsh, G.H., Vowles, P.D., 1997. Characterisation of
chemical species in PM2.5 and PM10 aerosols in Brisbane, Australia. Atmospheric
Environment 31, 3773-3785.
Chen, K. S., Ho, Y.T., Lai, C.H., Tsai, Y. A. and Chen, S. J., 2004. Trends in
Concentration of Ground-Level Ozone and Meteorological Conditions during High
Ozone Episodes in the Kao-Ping Airshed, Taiwan. Journal of the Air & Waste
Management Association 54 (1): 36-48.
Denison, L., Simpson, R., Petroeschevsky, A., Williams, G., 2001. Ambient air pollution
and daily hospital admissions in Melbourne 1994-1997. Melbourne EPA Victoria,
Report No.789
Howie, J., Tong, S, Verrall, K., Gerber, R., Wolff, R., 2005. Air Pollution and
Cardiopulmonary Diseases in Australia: A Review of Epidemiological Evidence.
Environmental Health 5 (1), 23-35.
Jo, W.K., Yoon, I.H., Nam, C.W., 2000. Analysis of air pollution in two major Korean
cities: trends, seasonal variations, daily 1-hour maximum versus other hour-based
concentrations, and standard exceedances.
Environmental Pollution 110 (1), 11-18.
Kan, H. Chen, B., 2003. Air pollution and daily mortality in Shanghai: a time series
study. Archives Environmental Health 58 (6), 360-367.
Krewski, D., Burnett, R., Jerrett, M., Pope, C. A , Rainham, D., Calle, E., Thurston, G.
and Thun, M., 2005. Mortality and long-term exposure to ambient air pollution:
69
ongoing analyses based on the American Cancer Society cohort. J Toxicol Environ
Health A. 68 (13-14):1093-109.
Kuebler, J. Bergh, H and Russell, A. G., 2001. Long-term trends of primary and
secondary pollutant concentrations in Switzerland and their response to emission
controls and economic changes. Atmospheric Environment 35, 1351-1363.
Kuzma, J. W., 1992. Basic Statistics for the Health Sciences, pp. 225-226. Mayfield
Publishing Company, California.
Morgan, G., Corbett, S., Woldarczyk, J., Lewis, P., 1998a. Air pollution and daily
mortality in Sydney, Australia, 1989 through 1993. American Journal of Public
Health 88 (5), 759-764.
Morgan, G., Corbett, S., Woldarczyk, J., 1998b, Air pollution and hospital admissions
in Sydney, Australia 1990 through 1994. American Journal of Public Health 88
(12), 1761-1766.
Murray, C.J. Lopez, A.D. (Eds), 1996. The global burden of disease: a comprehensive
assessment of mortality and disability from disease, injuries and risk factors in 1990
and projected to 2020, Harvard University Press Cambridge.
Petroeschevsky, A., Simpson, R.W., Thalib, L., Rutherford, S., 2001. Associations
between outdoor air pollution and hospital admissions in Brisbane, Australia.
Archives of Environmental Health 56 (1), 37-52.
QEPA. www.epa.qld.gov.au/environmental_management/air/air_quality_monitoring/
air_pollutants. Accessed 15/1/2005.
QEPA, annual report, 2003. Ambient air quality monitoring in Queensland 2003 annual
summary and trend report. http://www.epa.qld.gov.au/publications?id=1354.
70
Simpson, R.W., Williams, G., Petroeschevsky, A., Morgan, G., Rutherford, S., 1997.
Associations between outdoor air pollution and daily mortality in Brisbane, Australia.
Archives of Environmental Health 52 (6), 442-454.
Simpson, R., Denison, L., Petroeschevsky, A., Thalib, L., Willians, G., 2000. Effects of
ambient particle pollution on daily mortality in Melbourne,1991-1996. Journal of
Exposure Analysis and Environmental Epidemiology 10 (5), 488-496.
Yang, C.Y., Chen, Y.S., Yang, C.H., HO, S.C., 2004. Relationship between ambient air
pollution and hospital admissions for cardiovascular diseases in Kaohsiung, Taiwan.
Journal of Toxicology & Environmental Health: Part A 67 (6), 483-493.
Wikipedia. http://en.wikipedia.org/wiki/Brisbane Accessed 10/2/2005.
Wise, E. K. and Comrie, A. C., 2005. Meteorologically adjusted urban air quality trends
in the Southwestern United States. Atmospheric Environment 39, 2969-2980.
71
CHAPTER 5: SPATIAL PATTERNS OF SO2 AND
CARDIORESPIRATORY MORTALITY IN BRISBANE,
AUSTRALIA, 1999 – 2001
Citation:
Xiao-Yu Wang, Shilu Tong, Ken Verrall, Rod Gerber and Rodney Wolff (2007).
Spatial Patterns of SO2 and Cardiorespiratory Mortality in Brisbane, Australia, 1999 –
2001. Environmental Health, 7 (3): 64-74.
Contribution of authors:
X. Y. Wang was the principal author of the manuscript, performed all data analysis and
wrote the manuscript. ST contributed to the the research design, collection of data,
development of analytical protocol, assisted with interpretation of the results and writing of
the manuscript. Other authors contributed to the manuscript in terms of providing feedback
on the analyses and initial drafts.
72
Abbreviations
RD: Respiratory disease
CVD: Cardiovascular disease
CRD: Cardiorespiratory disease
ICD: International Classification of Diseases
PM10: Particulate matter with diameter less then 10 µm
NO2: Nitrogen dioxide
O3: Ozone
SO2: Sulphur dioxide
73
ABSTRACT
This study examined the spatial variation of sulphur dioxide (SO2) concentrations and
cardiorespiratory mortality in Brisbane, Australia, 1999 – 2001, using geographical
information systems techniques at a statistical local area level. For every 1 ppb increase
in annual average SO2 concentration, there was an estimated increase of 4.4 % (95 %
confidence interval (CI): 1.4 – 7.6 %) and 4.8 % (95 % CI: 2.0 – 7.7 %) in
cardiovascular and cardiorespiratory mortality, respectively. We estimated that the
excess number of cardiorespiratory deaths attributable to SO2 was 312 (3.4% of total
cardiorespiratory deaths) in Brisbane during the study period. Our results suggest that
long-term exposure to SO2, even at low levels, is a significant hazard to population
health. Therefore, spatial patterns of SO2 and other air pollutants and their impact on
health outcomes need to be assessed for an evaluation of long-term effects of air
pollution in metropolitan areas.
Key words: Air pollution, cardiorespiratory, mortality, spatial distribution, sulphur
dioxide
74
5.1 INTRODUCTION
Air pollution is a significant public health problem in metropolitan areas globally. It is
ranked by the World Health Organisation as one of the top ten contributors to the global
burden of disease and injury (Murray & Lopez 1996). In order to investigate the
relationship between air pollution and health outcomes, many epidemiological studies
have been conducted in North America, Europe Asia, Australia and New Zealand over
the past decade. These studies have found the consistent associations between the
particulate matter less than 10 μm (PM10) or and 2.5 μm (PM2.5) in aerodynamic
diameter and cardiorespiratory morbidity and mortality (Burnett et al. 2003; Chen et al.
2006 , 2007; Fischer et al. 2003; Hayes 2003; Jerrett et al. 2005; Kan et al. 2003 , 2004;
Krewski et al. 2005; Peng et al. 2005; Powe & Willis 2004; Ren & Tong 2006; Ren et al.
2006; Venners et al. 2003; Wong et al. 2002). Associations have also been reported for
gaseous air pollutants, such as nitrogen dioxide (NO2) (Fischer et al. 2003; Hoke et al.
2002; Kan et al. 2003, 2004; Scoggins et al. 2004; Simpson et al. 2005), ozone (O3)
(Bell et al. 2004; Fischer et al. 2003; Petroeschevsky et al. 2001; Simpson et al. 2005;
Wong et al. 2002) and sulphur dioxide (SO2) (Ballester et al. 2001; Krewski et al. 2003;
Venners et al. 2003).
The majority of the previous studies have examined the association between air
pollution and health outcomes using air pollution data from one monitoring station or
average values from a few monitoring stations. The spatial variation in the relationship
between air pollution and health outcomes is generally ignored. There is increasing
concern that the relationship between air pollution and mortality may differ with
geographical area within a city (Burnett et al. 2001; Chen et al. 2007; Krewski et al.
75
2003; Maheswaran & Craglia 2004; Peng et al. 2005; Scoggins et al. 2004). In fact, for a
metropolitan city, the ambient air pollution concentrations may vary dramatically with
different land use areas (Bell et al. 2004; Burnett et al. 2001; Krewski et al. 2003; Peng
et al. 2005). It is widely accepted that Geographical Information Systems (GIS) and
spatial modelling technologies provide important tools in the visualisation, exploration
and modelling of different kinds of spatial data, which may have many public and
environmental health applications (Maheswaran & Craglia 2004; Scoggins et al. 2004).
This study examined spatial variation in the relationship between sulphur dioxide (SO2)
concentrations and cardiorespiratory mortality using GIS techniques for three reasons:
firstly, ambient SO2 pollution is still a health concern in many metropolitan areas in
Australia (Australian Government Department of the Environment and Heritage, 2004);
secondly, few data are available on the spatial patters of ambient SO2 pollution and
mortality; and thirdly, the nature and magnitude of the impact of long-term exposure to
SO2 on mortality remain to be determined.
5.2 METHODS
Data collection
Brisbane is the capital city of Queensland. It is Australia’s third largest city after
Sydney and Melbourne, covering 1326.8 km2. The population of Brisbane city was
888,449 on 1 July 2001 (ABS, 2001).
In this study air pollution data included daily maximum 1-hour average concentrations
of SO2, NO2, O3 and daily 24-hour average concentrations of PM10 between 1 January
76
1999 and 31 December 2001, and the annual average and other measures were all based
on these data. Air pollution data were extracted from all available air monitoring
stations around and in Brisbane city by the Queensland Environmental Protection
Agency. There are two stations operating in the Brisbane central business district (CBD). The
first station was established in 1995 and located in an elevated position at the Queensland
University of Technology (QUT) campus. The second station was established in 1998 at the
Queensland Science Centre (SCI). Differential optical absorption spectroscopy technology is
used to measure O3, NO2 and SO2, as well as other pollutants (EPA, 2007). Figure 5.1 shows
six SO2 monitoring stations including four in Brisbane city and two around Brisbane
city. The shadow region is the metropolitan area of Brisbane city.
Figure 5.1 Location of ambient SO2 monitoring stations in and around the Brisbane
metropolitan area 1999 – 2001 (Brisbane CBD (QUT and SCI), Eagle Farm, Flinders
View, Springwood and Wynnum)
The Office of the Queensland Government Statistician provided data on daily mortality
in Brisbane city between 1 January 1999 and 31 December 2001. The data included day,
month, year of death, sex, age, statistical local area (SLA) of residence and cause of
77
death. An SLA is the basic spatial analytic unit used to collect and disseminate statistics
by the Australian Bureau of Statistics. 162 SLAs cover the whole of Brisbane city
without gaps or overlaps. In this study, cause-specific deaths were categorised
according to the 10th revision of the International Classification of Diseases (ICD10),
defined as respiratory diseases (RD) (ICD10 J00 – J99 excluding Influenza),
cardiovascular diseases (CVD) (ICD10 I00 - I99), and cardiorespiratory diseases (CRD)
(including both respiratory and cardiovascular diseases).
In order to calculate mortality rates for cardiovascular and respiratory diseases, we
obtained the population data for each SLA from ABS. The most recent census data was
from 2001 and included sex and age group (0-4, 5-9, …, 100+) by SLA. There were
substantial differences in the population size between SLAs. On average, an SLA
contained 5500 people, with a range from 250 to 15,631 people. The percentage
distributions of population by age group 0 – 14; 15 – 64; 65 – 74; 75 + in all SLAs were
between 4 - 31%; 48 - 89%; 2 - 13%; and 1 – 34%, respectively.
We also obtained the Socio-Economic Indexes for Areas (SEIFA) data for each SLA
from the ABS. It contained four summary indices (Advantage/Disadvantage Index,
Disadvantage Index, Economic Resources Index, and Education and Occupation Index)
to measure different aspects of socio-economic conditions for each SLA. These indices
have high degrees of correlation between each other (rs = 0.833 - 0.959). Therefore we
selected the Disadvantage Index to represent the socioeconomic status of a particular
SLA in this study, for two major reasons: firstly, it focuses on low-income earners,
relatively lower educational attainment, high unemployment and other variables
78
reflecting disadvantage; and secondly, this index was more strongly associated with
cardiorespiratory mortality (rs = -0.19) than other three indices of SEIFA (rs | ≤ | 0.18).
Data mapping and analysis
In this study, GIS was used to map the spatial patterns of annual average SO2
concentrations and mortality from respiratory, cardiovascular and cardiorespiratory
diseases. To assess the association between exposure to SO2 concentrations and
mortality at a SLA level, an Inverse Distance Weighted (IDW) method was used to
estimate the annual average SO2 concentrations for each SLA (Isaaks & Srivastava
1989; O’Sullivan & Unwin 2003). The IDW method is one of the commonly used GIS
techniques for interpolated data. The method assumes that each known value of the
measured variable in space has a local influence diminishing with distance. In the case
of point measurements of pollutant concentration, it weighs the points closer to the
prediction location greater than those farther away. It is widely used in the spatial
interpolation of air pollution (Wu et al., 2006). We used IDW method to estimate the SO2
concentrations for a SLA without a monitoring site as a distance-weighted sum of the
values observed at available surrounding monitoring stations. Consequently, an output
surface of the annual average of SO2 concentrations for all Brisbane area was produced
for each year. Such a surface was represented by values at each 1 km x 1km orthogonal
grid node over the total area. Then for a specific SLA, SO2 level was calculated by
adding SO2 levels of all grid cells together and then dividing the sum by the total
number of cells within the SLA. MapInfo Professional software (MapInfo Corporation
2003) was used to display the spatiotemporal distributions of air pollution and mortality
data. The digital base map datasets used for constructing the GIS were obtained
79
primarily from ABS. These data were manipulated by thematic mapping to facilitate the
accurate identification of the spatial locations of SLA and their linkages with the other
data layers. The SLAs of mortality and SO2 concentrations were geo-coded to the digital
base maps of localities using MapInfo and Microsoft Access software.
In order to compare the spatial patterns of mortality, standardized mortality rates (SMR)
for each SLA were calculated using the direct method, with the Brisbane city population
as a reference to adjust for differences in the age distributions among SLAs (Selvin
2001 The equation for calculating a SMR is
∑∑= )(comparisoni
i
de
SMR
Where ∑ ie is the total number of expected cases generated using the reference
population rates for each SLA; ∑ )(comparisonid is the total population in the comparison
group. In this study, firstly, we get the total number of deaths in the SLA by age group
(0-14, 15-64, 65-74 and 75+); secondly we calculated age-specific rates per 100,000 of
RD, CVD and CRD deaths (age group) for each SLA; then we calculated the expected
number of deaths in each age group by SLA; and finally we summed the expected
number of deaths and divided the Brisbane population to get SMR per 100,000 for each
SLA. A step by step example of how to calculate SMR in 5 SLAs of Brisbane is
provided in Appendix II. In this study, we used same method, but calculated SMR for
whole Brisbane city including 162 SLAs.
Statistical analyses were conducted on daily maximum 1-hour SO2 concentrations and
mortality of respiratory, cardiovascular and cardiorespiratory diseases in Brisbane, from
January 1999 to December 2001. Missing values were dealt with using the listwise
80
deletion approach (known as a complete case analysis) (Brown 1983). Listwise deletion
is an ad hoc method of dealing with missing data in that it deals with the missing data
before any substantive analyses are done. It is considered the easiest and simplest
method of dealing with missing data (Brown 1983). On average, the missing value was
4.9% in six monitoring stations during the study period, with a range from 0%
(Wynnum) to 9.9% (Brisbane CBD (SCI)). Logistic regression was used to investigate
how SO2 concentrations influenced the probability of deaths, while controlling for
confounding effects of age, sex, socioeconomic indices and other air pollutants (PM10,
NO2 and O3) for each SLA. The annual average concentrations of daily 24-hour average
PM10, daily maximum 1-hour average NO2 and O3 for all SLAs were also obtained by
IDW method. The logistic regression model is described as follows:
P (Deaths) = Exp(X) / [1+Exp(X)]
Where: P (Deaths) is the total number of deaths divided by the population for each
SLA;
X = β0 + β1 (SO2) + β2 (age) + β3 (Sex) + β4 (SEIFA) + β5 (PM10) + β6 (NO2) + β7
(O3)
Excess deaths in SLAs with higher levels of SO2
We compared SMR between SLAs with annual average SO2 concentrations greater than
4.20 ppb (85 SLAs) and less than or equal to 4.20 ppb (77 SLAs). A cut off point of
4.20 ppb was chosen as it represented the overall annual average of SO2 concentrations
in Brisbane. It was assumed that cardiorespiratory mortality was low in the areas with
81
low levels of SO2; and vice versa. The average annual number of excess deaths
attributable to air pollution occurring in areas with high SO2 pollution (i.e., > 4.20 ppb)
was calculated as follows:
Y = Ohigh – E; E = Olow/Plow*Phigh
Where: Y is excess number of deaths in high air pollution area (1999 – 2001); O is
observed number of deaths and E is expected number of deaths in the high air pollution
area after sex and age were adjusted for, while Plow and Phigh are the populations
residing in low or high air pollution area. All the analyses were conducted using the
SAS statistical software package version 9.1 (SAS Institute Inc, 2002).
5.3 RESULTS
Descriptive statistics for air pollution and mortality
Table 5.1 shows the SO2 concentrations in six monitoring stations across land use types.
During the study period, the mean of annual average SO2 concentrations was highest for
Flinders View (9.1 ppb) and lowest for Springwood (2.4 ppb). There was a large spatial
variation of SO2 between monitoring stations according to the correlation analysis.
Spearmen correlations were calculated for different air pollutants within same monitoring
station and for the same pollutant between monitoring stations. The correlations between daily
maximum one hour NO2, O3, SO2 and 24-hour average PM10 concentrations in each station
ranged from 0.01 (O3 concentrations with SO2 for Eagle Farm) to 0.52 (NO2 concentrations with
O3 for Wynnum). For the same air pollutant across different monitoring stations, correlations
82
ranged from 0.57 to 0.90 for PM10, 0.24 to 0.91 for NO2, 0.43 to 0.90 for O3 and -0.40 to 0.73
for SO2.
Table 5.1 Summary statistics of daily maximum 1-hour SO2 concentrations (ppb) and
land use type for six air monitoring stations in Brisbane metropolitan area between 1999
and 2001
Percentiles
Monitoring station Land use type Mean SD* Min 25 50 75 Max
Brisbane CBD (QUT) Commercial business 3.3 2 0 2 3 4 18
Brisbane CBD (SCI) Commercial business 2.9 2.3 0 2 2 4 27
Eagle Farm Light industrial 6.8 5.4 0 3 5 9 60
Flinders View Residential 9.1 9.7 0 2 6 13 81
Springwood Residential 2.4 3 0 1 1 3 46
Wynnum Residential adjacent to industrial 5.3 5.5 0 2 3 7 39
*SD: standard deviation
During 1999 - 2001, there were total 18,108 deaths in Brisbane and among them, 9,178
(51%) died from cardiorespiratory diseases. Table 5.2 shows the daily number of
cardiovascular deaths by cause and gender. The annual number of cardiovascular deaths
decreased from 2,688 (1999) to 2,473 (2001), while the annual number of respiratory
deaths varied slightly from 482 (1999) to 520 (2000).
Spatial distribution of SO2 and mortality
Figure 5.2 shows the annual average SO2 concentrations between 1999 and 2001 for
each SLA, and the number of people exposed to each level of SO2 concentration. The
highest SO2 levels were observed around Eagle farm stations, which are located in a
light to moderate industrial area.
83
Table 5.2 Summary statistics of daily number of cardiorespiratory mortalities in
Brisbane 1999-2001
Percentiles
Gender Disease Mean SD Min 25 50 75 Max
All Respiratory 1.4 1.2 0 0 1 2 8
Cardiovascular 7 3.1 0 5 7 9 31
Cardiorespiratory 8.4 3.4 1 6 8 10 33
Male Respiratory 0.7 0.9 0 0 0 1 5
Cardiovascular 3 1.8 0 2 3 4 13
Cardiorespiratory 3.7 2 0 2 3 5 13
Female Respiratory 0.7 0.8 0 0 0 1 5
Cardiovascular 4 2.2 0 2 4 5 21
Cardiorespiratory 4.7 2.5 0 3 4 6 23
Figure 5.2 Spatial patterns of annual average SO2 and number of people exposed
between 1999 and 2001 in the Brisbane at a SLA level (For each year, daily maximum
1-hour SO2 concentrations data from all operational stations were aggregated using
IDW method to estimate the annual average SO2)
84
Figure 5.3 shows the SMR of respiratory, cardiovascular and cardiorespiratory diseases
by SLA. The mean and the range of annual average SO2 concentrations in each category
of SMR were also included. Evidently, there was a strong spatial pattern of SMR across
SLAs.
Modelling the effect of SO2 pollution on mortality
We evaluated the independent associations of annual average SO2 concentrations with
different types of mortality using logistic regression models after adjustment for the
potential confounders. The results show that SO2 concentrations were statistically
significantly and positively associated with RD, CVD and CRD mortality (Table 5.3).
There were statistically significant crude associations between annual average SO2
concentrations and all three types of mortality. After controlling for the putative
confounders, there was an estimated increase of 4.4% and 4.8% in CVD and CRD
mortality, respectively, for every 1 ppb increase in annual average SO2 concentrations.
There was only a marginally significant relationship between SO2 and RD mortality
after controlling for potential confounding factors. For males, there was an estimated
increase of 5.8 % and 4.4 % in CVD and CRD mortality, respectively, for every 1 ppb
increase in annual average SO2 concentrations. However there was no significant
relationship between SO2 concentrations and RD mortality. For females, there was an
estimated increase of 16.3 % and 5.4 % in RD and CRD mortality, respectively, for
every 1 ppb increase in annual average SO2 concentrations. There was only a marginally
significant relationship between SO2 and CVD mortality after adjusting for confounding
factors (Table 5.3).
85
A. SMR of RD (SLAs)
B. SMR of CVD (SLAs)
C. SMR of CRD (SLAs)
Figure 5.3 Spatial patterns of annual average standardised rates of respiratory,
cardiovascular and cardiorespiratory mortality; and annual average SO2 exposed in the
Brisbane at a SLA level (1999 – 2001)
86
Excess deaths attributable to air pollution
Table 5.4 shows the excess number (ie, excess deaths due to exceedences of SO2 above
the annual average) of RD, CVD and CRD deaths based on the mean of annual average
SO2 concentrations (> 4.2 ppb) during the three year period. 6.3 % of respiratory deaths,
2.8 % of cardiovascular deaths and 3.4 % of cardiorespiratory deaths were attributable
to SO2 pollution. The excess number of cardiorespiratory deaths (312) was more than
twice that due to transport accidents (ICD 10 V01 – V99) (i.e. 144 deaths) in Brisbane
over the same period.
Table 5.3 Odds ratio (95% Confidence Interval) for cardiorespiratory mortality
associated with annual average SO2 concentrations in Brisbane by SLA (1999-2001)
Odds Ratio
Gender Disease Unadjusted (95% CI) Adjusted * (95% CI)
All Respiratory 1.154 (1.083-1.229) 1.069 (0.999-1.143)
Cardiovascular 1.115 (1.085-1.147) 1.044 (1.014-1.076)
Cardiorespiratory 1.122 (1.093-1.151) 1.048 (1.020-1.077)
Male Respiratory 1.063 (0.972-1.162) 0.987 (0.897-1.086)
Cardiovascular 1.120 (1.073-1.168) 1.058 (1.011-1.107)
Cardiorespiratory 1.109 (1.067-1.152) 1.044 (1.002-1.088)
Female Respiratory 1.257 (1.149-1.375) 1.163 (1.058-1.279)
Cardiovascular 1.110 (1.070-1.153) 1.037 (0.997-1.079)
Cardiorespiratory 1.131 (1.092-1.170) 1.054 (1.016-1.094)
* Adjusted for the confounding effects of age, disadvantage index and other air pollutants (i.e. PM10, NO2 and O3).
The regression coefficients are provided in the appendix I.
87
Table 5.4 Estimated excess deaths of cardiorespiratory disease in higher SO2 SLAs
(1999 – 2001)
Disease
Expected number
of deaths
Observed number
of deaths
Excess number of
deaths
Percentage of
increased deaths (%)
Respiratory 652 746 94 14.4
Cardiovascular 3377 3595 218 6.5
Cardiorespiratory 4029 4341 312 7.7
5.4 DISCUSSION
In this study, we used the IDW method to link annual average SO2 concentrations with
RD, CVD and CRD mortality in each SLA after adjustment for a range of confounding
factors. We found the statistically significant and positive association between long-
term exposure to SO2 and cardiorespiratory mortality.
SO2 is a colourless, irritating and reactive gas with a strong odour. In Australia,
emissions of sulphur dioxide are primarily from industrial operations that burn fuels
such as coal, oil, petroleum and gas and from wood pulping and paper manufacturing. It
is also emitted by vehicles. It irritates the eyes, nose and throat, and people with
impaired lungs or hearts and asthmatics are particularly at risk of exacerbating existing
health problems (EPA, 2007). Ambient SO2 concentrations are generally low. Levels of
SO2 vary between regions due to varied geographical distribution of major sources and
different topographical and meteorological conditions. SO2 levels in Australian cities
are low compared to the USA and Europe because of the limited number of major SO2
emitting industries and low sulphur fuels (ABS, 2006). Compared with other Australian
cities, SO2 level in Brisbane was higher than Sydney, but lower than Adelaide,
88
Melbourne and Perth (Australian Government Department of the Environment and
Heritage, State of the Air, 2004).
Our findings are consistent with previous studies that reported the adverse effects of
exposure to SO2 on health outcomes, even when the levels of SO2 are not considered
very high (Fischer et al. 2003; Kan et al. 2003; Powe & Willis 2004; Venners et al.
2003; Wong et al. 2002). For example, a study conducted in Shanghai, China (Kan &
Chen 2003) found the significant association of SO2 with death rates from all causes
and from RD and CVD. The concentration of SO2 was more strongly associated with
increased deaths than those of PM10 or NO2 in Shanghai. Several studies in Europe have
also reported a stronger association of SO2 with the excess mortality than that for other
air pollutants (Fischer et al. 2003; Powe & Willis 2004). In fact, no monitoring station
in Brisbane recorded a daily maximum 1-hour SO2 concentration that exceeded the
National Environment Protection Measure standard of 200 ppb (for 1-hour) at any time
during the study period. The highest SO2 levels were recorded at Flinders View, and the
rise in the 95th percentile concentration measured at this site since 1998 may reflect the
increase in the generating capacity of the Swanbank coal-fired power station during this
period (EPA Air quality reports, 2001). SO2 in Brisbane not like other air pollutants
emitted from vehicles (eg. NO2 and O3) is primarily from industrial sources, which are
largely concentrated near Eagle Farm.
The six SO2 monitoring stations were not well dispersed throughout the metropolitan
area of Brisbane (Figure 5.1). There was no monitoring station in the western area, but
the SO2 level of exposure estimated using the IDW method was relatively high in the
western suburbs. That is because there were higher levels of SO2 than other stations
89
recorded at the Flinders View monitoring station which is close to the western area
(Table 5.1).
In this study, the SO2 threshold of 4.2 ppb was chosen because it represented the annual
average of the daily maximum 1-hour SO2 concentrations across six monitoring stations
in Brisbane. Up to now, many epidemiological studies have been conducted to
investigate the relationship between air pollution (e.g. PM10, NO2, O3, CO and SO2) and
health outcomes, but only a few have used a spatial approach as we did in this study
(Burnett et al. 2001; Chen et al. 2007; Peng et al. 2005; Scoggins et al. 2004). Standards
for air pollution are concentrations over a given time period that are considered to be
acceptable in the light of what is known about the effects of each pollutant on health and
on the environment (EPA, 2007). They can also be used as a benchmark to see if air
pollution is getting better or worse. Different countries adopt different standards. For
example, annual mean of SO2 should not exceed 30 ppb in USA (US EPA, 2006) and 20
ppb in UK and Australia (UK Air Quality, 2006; DEH, 2006). However, our study
found that even lower level exposure to SO2 may still have the adverse effects on health
outcomes.
The understanding of the mechanisms behind the adverse health effects of SO2 is
improving. For example, some evidence shows that SO2 exposure can change the
expression of apoptosis-related genes, and it suggests that SO2 can induce apoptosis in
lung of rat and may cause some apoptosis-related diseases (Bai & Meng 2005).
Elucidating the expression patterns of those factors after SO2 inhalation may be critical
to the understanding of the mechanisms of SO2 toxicity. SO2 at higher concentrations
(28 and 56 mg/m3) has been shown to significantly decrease pulmonary CYP1A1 and
90
1A2 mRNA levels in animals (Qin & Meng 2005). Both CYP1A1 and 1A2 mRNA are
involved in the metabolism of many important drugs and environmental chemicals.
Furthermore, the decreases of activities and mRNA levels of these P450 enzymes
caused by SO2 at different concentrations in lungs and livers of rats follow linear dose-
response curves.
This study has four major strengths. Firstly, according to our knowledge, this is the first
study to examine spatial patterns of SO2 concentrations and mortality in a metropolitan
setting. Secondly, the datasets used in this study were quite comprehensive; there were
only 4.9% of SO2 missing data in six monitoring sites during the study period. Thirdly,
SO2 level in this study was much lower than the National Standard, but we still
observed its effects on mortality of CRD. Finally, we have controlled for most of the
known confounding factors including age, sex, SEIFA and other air pollutants at a SLA
level. Although particulate matter less than 2.5 µm (PM2.5) is another major concern in
air pollution-health studies, there is only one station (Rocklea) in Brisbane measured
PM2.5 for recent years. Therefore we did not select it as a confounding factor in this
study.
There are two major limitations in this study. Firstly, the dataset is not extensive enough
to detect small risks because only three-year data on air pollution and mortality were
included. Secondly, like other ecological studies, exposure misclassification could have
occurred to a certain extent as no information on personal exposure was available. In
future research, it would be desirable to undertake a prospective cohort study to
examine the spatial variation in the long-term relationship of SO2 and other air
pollutants (e.g. PM10, NO2 and O3) with cardiorespiratory mortality.
91
In conclusion, this study shows that long-term exposure to SO2 pollution was
statistically significantly associated with respiratory, cardiovascular and
cardiorespiratory mortality in Brisbane. Thus, spatial patterns of air pollution and
mortality in metropolitan areas need to be carefully examined in future research.
ACKNOWLEDGEMENTS
This study was partly funded by the Australian Research Council (DP0559655). Dr.
Shilu Tong is supported by a NHMRC research fellowship.
92
REFERENCES
ABS (Australian Bureau of Statistics), 2001.
http://www.abs.gov.au/ausstats/[email protected]/4079a1bbd2a04b80ca256b9d00
208f92/ab51cb9754fe337aca256bbf0000dbfc!OpenDocument. Accessed 7
February, 2005.
ABS (Australian Bureau of Statistics), 2006. Australia's Environment: Issues and
Trends, 2006.
http://www.abs.gov.au/Ausstats/[email protected]/Latestproducts/C0659BE5C7445912
CA2572340020FBBD?opendocument, Accessed 6 August, 2007
Australian Government Department of the Environment and Heritage, State of the Air:
Community Summary 1991-2001. 2004.
http://www.environment.gov.au/atmosphere/airquality/publications/status/index.
html,Accessed 8 January, 2007.
Bai, J., & Meng, Z. 2005, 'Effects of sulfur dioxide on apoptosis-related gene
expressions in lungs from rats'. Regulatory toxicology and pharmacology, vol.
43, no. 3, pp. 272-9.
Ballester, F., Tenias, J. M., & Perez-Hoyos, S. 2001, 'Air pollution and emergency
hospital admissions for cardiovascular diseases in Valencia, Spain'. Journal of
epidemiology and community health, vol. 55, no. 1, pp. 57-65.
Bell, M. L., McDermott, A., Zeger, S. L., Samet, J. M., & Dominici, F. 2004, 'Ozone
and short-term mortality in 95 US urban communities, 1987-2000'. The journal
of the American Medical Association, vol. 292, no. 19, pp. 2372-8.
Brown, C. H., 1983. Asymptotic comparison of missing data procedures for estimating
factor loadings. Psychometrika 48, 269-292.
93
Burnett, R., Ma, R., Jerrett, M., Goldberg, M. S., Cakmak, S., Pope, C. A., 3rd, et al.
2001, 'The spatial association between community air pollution and mortality: a
new method of analyzing correlated geographic cohort data'. Environmental
health perspectives, vol. 109, no. Suppl 3, pp. 375-80.
Chen, L., Mengersen, K., & Tong, S. 2007, 'Spatiotemporal relationship between
particle air pollution and respiratory emergency hospital admissions in Brisbane,
Australia'. The Science of the total environment, vol. 373, no. 1, pp. 57-67.
Chen, L., Verrall, K., & Tong, S. 2006, 'Air particulate pollution due to bushfires and
respiratory hospital admissions in Brisbane, Australia'. International journal of
environmental health research, vol. 16, no. 3, pp. 181-91.
DEH (Department of the Environment and Heritage), 2006. www.deh.gov.au, Accessed
9 November 2006
EPA (Environmental Protection Agency), 2007.
http://www.epa.qld.gov.au/environmental_management/air/air_quality_monitori
ng/southeast_queensland_monitoring_stations/#BrisbaneCBD. Accessed 20 July
2007.
EPA Air quality reports, 2001.
http://www.epa.qld.gov.au/environmental_management/air/air_quality_monitori
ng/air_quality_reports/. Accessed 10 June 2005.
Fischer, P., Hoek, G., Brunekreef, B., Verhoeff, A., & van Wijnen, J. 2003, 'Air
pollution and mortality in The Netherlands: are the elderly more at risk?' The
European respiratory journal. Supplement, vol. 40, pp. 34s-8s.
94
Hayes, M. V. 2003, '"Ecologic confounders" in the context of a spatial analysis of the
air pollution-mortality relationship'. Journal of toxicology and environmental
health. Part A, vol. 66, no. 16-19, pp. 1779-82.
Hoek, G., Brunekreef, B., Goldbohm, S., Fischer, P., & van den Brandt, P. A. 2002,
'Association between mortality and indicators of traffic-related air pollution in
the Netherlands: a cohort study'. Lancet, vol. 360, no. 9341, pp. 1203-9.
Isaaks, E. H., Srivastava, R. M. 1989. An Introduction to Applied Geostatistics. Oxford
University Press, New York, pp. 249-77.
Jerrett, M., Burnett, R. T., Ma, R., Pope, C. A., 3rd, Krewski, D., Newbold, K. B., et al.
2005, 'Spatial analysis of air pollution and mortality in Los Angeles'.
Epidemiology, vol. 16, no. 6, pp. 727-36.
Kan, H., & Chen, B. 2003, 'Air pollution and daily mortality in Shanghai: a time-series
study'. Archives of environmental health, vol. 58, no. 6, pp. 360-7.
Kan, H., Jia, J., & Chen, B. 2004, 'The association of daily diabetes mortality and
outdoor air pollution in Shanghai, China'. Journal of environmental health, vol.
67, no. 3, pp. 21-6.
Krewski, D., Burnett, R., Jerrett, M., Pope, C. A., Rainham, D., Calle, E., et al. 2005,
'Mortality and long-term exposure to ambient air pollution: ongoing analyses
based on the American Cancer Society cohort'. Journal of toxicology and
environmental health. Part A, vol. 68, no. 13-14, pp. 1093-109.
Krewski, D., Burnett, R. T., Goldberg, M. S., Hoover, B. K., Siemiatycki, J., Jerrett, M.,
et al. 2003, 'Overview of the reanalysis of the Harvard Six Cities Study and
American Cancer Society Study of Particulate Air Pollution and Mortality'.
Journal of toxicology and environmental health. Part A, vol. 66, no. 16-19, pp.
1507-51.
95
Maheswaran, R., Craglia, M. 2004. GIS in Public Health Practice. CRC Press, Boca
Raton, pp. 125-66.
Murray, C. J. L., Lopez, A. 1996. The global burden of disease: a comprehensive
assessment of mortality and disability from disease, injuries and risk factors in
1990 and projected to 2020. Distributed by Harvard University Press,
Cambridge, MA, 990 pp.
O’Sullivan, D., Unwin, D. 2003. Geographic Information Analysis. Wiley, Hoboken,
N.J. pp. 77-134.
Peng, R. D., Dominici, F., Pastor-Barriuso, R., Zeger, S. L., & Samet, J. M. 2005,
'Seasonal analyses of air pollution and mortality in 100 US cities'. American
journal of epidemiology, vol. 161, no. 6, pp. 585-94.
Petroeschevsky, A., Simpson, R. W., Thalib, L., & Rutherford, S. 2001, 'Associations
between outdoor air pollution and hospital admissions in Brisbane, Australia'.
Archives of environmental health, vol. 56, no. 1, pp. 37-52.
Powe, N. A., & Willis, K. G. 2004, 'Mortality and morbidity benefits of air pollution
(SO2 and PM10) absorption attributable to woodland in Britain'. Journal of
environmental management, vol. 70, no. 2, pp. 119-28.
Qin, G., & Meng, Z. 2005, 'Effect of sulfur dioxide inhalation on CYP1A1 and
CYP1A2 in rat liver and lung'. Toxicology letters, vol. 160, no. 1, pp. 34-42.
Ren, C., & Tong, S. 2006, 'Temperature modifies the health effects of particulate matter
in Brisbane, Australia'. International journal of biometeorology, vol. 51, no. 2,
pp. 87-96.
Ren, C., Williams, G. M., & Tong, S. 2006, 'Does particulate matter modify the
association between temperature and cardiorespiratory diseases?' Environmental
health perspectives, vol. 114, no. 11, pp. 1690-6.
96
Scoggins, A., Kjellstrom, T., Fisher, G., Connor, J., & Gimson, N. 2004, 'Spatial
analysis of annual air pollution exposure and mortality'. The Science of the total
environment, vol. 321, no. 1-3, pp. 71-85.
Selvin, S., 2001. Epidemiologic analysis: a case-oriented approach. Oxford University
Press, New York, pp. 161-85.
Simpson, R., Williams, G., Petroeschevsky, A., Best, T., Morgan, G., Denison, L., et al.
2005, 'The short-term effects of air pollution on hospital admissions in four
Australian cities'. Australian and New Zealand journal of public health, vol. 29,
no. 3, pp. 213-21.
UK Air Quality, 2006. www.airquality.co.uk, Accessed 9 November 2006
US EPA (U.S. Environmental Protection Agency), 2006. http://epa.gov/air/criteria.html,
Accessed 9 November 2006
Venners, S. A., Wang, B., Xu, Z., Schlatter, Y., Wang, L., & Xu, X. 2003, 'Particulate
matter, sulfur dioxide, and daily mortality in Chongqing, China'. Environmental
health perspectives, vol. 111, no. 4, pp. 562-7.
Wong, T. W., Tam, W. S., Yu, T. S., & Wong, A. H. 2002, 'Associations between daily
mortalities from respiratory and cardiovascular diseases and air pollution in
Hong Kong, China'. Occupational and environmental medicine, vol. 59, no. 1,
pp. 30-5.
Wu, J., Winer, A. M., & Delfino R, J. 2006, 'Exposure assessment of particulate matter
air pollution before, during, and after the 2003 Southern California wildfires'.
Atmospheric Environment, vol. 40, pp. 3333-48.
97
CHAPTER 6: SPATIAL ANALYSIS OF LONG-TERM
EXPOSURE TO GASEOUS AIR POLLUTANTS AND
CARDIORESPIRATORY MORTALITY IN BRISBANE,
AUSTRALIA, 1996 – 2004
Citation:
Xiao-Yu Wang, Wenbiao Hu and Shilu Tong (2007). Spatial analysis of long-term
exposure to gaseous air pollutants and cardiorespiratory mortality in Brisbane,
Australia, 1996 – 2004
To be submitted
Contribution of authors:
X. Y. Wang was the principal author of the manuscript, performed all data analysis and
wrote the manuscript. ST and WH contributed to the the research design, collection of
data, development of analytical protocol, assisted with interpretation of the results and
writing of the manuscript.
98
ABSTRACT
This study examined the association of long-term exposure to gaseous air pollution
(including nitrogen dioxide (NO2), ozone (O3) and sulphur dioxide (SO2)) with
cardiorespiratory mortality in Brisbane, Australia, 1996 - 2004. The pollutant
concentrations were estimated using geographical information systems (GIS) techniques
at a statistical local area (SLA) level. Logistic regression was used to investigate the
impact of NO2, O3 and SO2 on cardiorespiratory mortality after adjusting for potential
confounding effects of age, sex, calendar year and socio-economic indexes for areas
(SEIFA). The results show that the mean of annual average NO2, O3 and SO2
concentrations was19.8, 30.1 and 5.4 ppb, respectively. There was estimated 3.1% (95%
CI: 0.4 – 5.8%) and 0.5% (95% CI: -0.03 – 1.3 %) increase in cardiorespiratory
mortality for 1 ppb increase in annual average concentration of SO2 and O3, respectively
in the multiple gaseous pollutants model. Our results indicate that long-term exposure to
gaseous air pollutants in Brisbane, even at the levels lower than most of cities
(especially SO2), was associated with cardiorespiratory mortality. Therefore, spatial
patterns of gaseous air pollutants and their impact on health outcomes need to be
assessed for an evaluation of long-term effects of air pollution in metropolitan areas.
Key words: Cardiorespiratory mortality; Nitrogen dioxide; Ozone; Sulphur dioxide;
Spatial distribution
99
6.1 INTRODUCTION
Air pollution is ranked by the World Health Organisation as one of the top ten
contributors to the global burden of disease and injury (Murray and Lopez, 1996).
Exposure to gaseous air pollutants, even at a low level, has been associated with
cardiorespiratory diseases (Vedal et al., 2003). Most recent epidemiological studies of
air pollution have used time-series analyses to explore the relationship between daily
mortality or morbidity and daily ambient air pollution concentrations based on the same
day or previous days (Chock et al., 2000; Cifuentes et al., 2000; Hajat et al., 2007; Lee
et al., 2000; Moolgavkar, 2003; Roberts, 2004; Simpson et al., 2005; Yang et al., 2004).
Existing studies typically fit a generalized additive model (GAM) (Hastie and
Tibshirani, 1990) or generalized linear model (GLM) (McCullagh and Nelder, 1989) to
concurrent time series of daily mortality, ambient air pollution, and meteorological
covariates. However, most of the previous studies have examined the association
between air pollution and health outcomes using air pollution data from a single
monitoring site or average values from a few monitoring sites to represent the whole
population of the study area. In fact, for a metropolitan city, ambient air pollution levels
may differ significantly among the different areas.
There is increasing concern that the relationships between air pollution and mortality
may vary with geographical area (Burnett et al., 2001; Chen et al., 2007; Krewski et al.,
2003; Maheswaran and Craglia, 2004; Peng et al., 2005; Scoggins et al., 2004). A
number of studies have shown that groups with low socio-economic status tend to live
in areas with higher levels of air pollutants than groups with high socio-economic status
(Finkelstein et al., 2005; Mitchell and Dorling, 2003; Neidell, 2004). Additionally, some
100
studies have indicated that socio-economic status can act as a confounder when
investigating the relation between geographical location and health (Glover et al., 2004;
Hodgson and Mathur, 2000; Scoggins et al., 2004; Taylor et al., 2004).
This study aims to examine the spatial variation in the relationship between long-term
exposure to gaseous air pollutants (including nitrogen dioxide (NO2), ozone (O3) and
sulphur dioxide (SO2)), and cardiorespiratory mortality in Brisbane, Australia, during
the period 1996 - 2004. We did not include particulate matter in this study because the
geophysical features of particles are very different to gaseous air pollutants.
6.2 MATERIALS AND METHODS
Brisbane is the capital of Queensland. It is the third largest city after Sydney and
Melbourne in Australia, covering 1326.8 km2 for urban areas (ABS, 2001). The
population of Brisbane city were 824,489 on 30 June 1996 and 958,504 on 30 June
2004, increasing by 16.25% during this period. The main sources of air pollution in
Brisbane include motor vehicle traffic, controlled and uncontrolled burning in the
vicinity of the city, and a number of industrial sources including a power station, oil
refineries, oil terminal and a brewery.
The data analysis was undertaken at a statistical local area (SLA) level. SLA is the basic
spatial unit used to collect and analyse data, and 162 SLAs cover the whole of Brisbane
city without gaps or overlaps (ABS, 2001). A SLA typically contains 5500 people, with
a range from 250 to 15,631 people. The average size of a SLA was approximately 8 km2
(ranged from 0.4 to 184.8 km2). Health data are routinely collected at the SLA level.
101
Air pollution data
Daily data on maximum one hour O3, NO2 and SO2 concentrations in parts per billion
(ppb) for the period of 1 Jan. 1996 to 31 Dec. 2004 were obtained from thirteen
monitoring stations by the Queensland Environmental Protection Agency (QEPA),
including 8 stations within Brisbane city and 5 around the city (Figure 6.1). These data
were used to simulate the urban air quality of Brisbane. There are two stations operating
in the Brisbane central business district (CBD). The first station was established in 1995
and located in an elevated position at the Queensland University of Technology (qut)
campus. The second station was established in 1998 at the Queensland Science Centre
(sci). Differential optical absorption spectroscopy technology is used to measure O3,
NO2 and SO2, as well as other pollutants (QEPA, 2007).
102
Figure 6.1 Locations of gaseous air pollution monitoring stations around urban
Brisbane city
The daily gaseous air pollutants of O3, NO2 and SO2 concentrations were aggregated to
annual means which were used as a long-term air pollution exposure indicator. Missing
values of air pollution concentrations were dealt with using the listwise deletion
approach, which is considered as the easiest and simplest method of dealing with
missing data (Brown, 1983). Geographical information systems (GIS) techniques were
used in mapping the spatial patterns of annual average O3, NO2 and SO2 concentrations
at a SLA level. The original air pollution data from QEPA were point measurements
recorded at each monitoring station and most of the SLAs do not have a monitoring
station. In order to assess the association between exposure gaseous pollutants and
cardiopulmonary mortality at the SLA level, we estimated the annual average potential
concentrations of gaseous pollutants by SLA using an Inverse Distance Weighted
(IDW) method (Isaaks and Srivastava, 1989; O’Sullivan and Unwin, 2003). That is to
103
estimate the gaseous pollutants of O3, NO2 and SO2 concentrations for an SLA without a
monitoring station as a distance-weighted sum of the values observed at available
surrounding monitoring stations. Consequently, an output surface of the annual average
of O3, (NO2 or SO2) concentrations for all the Brisbane area was produced for each year.
Such a surface was represented by values at each 1 km x 1km orthogonal grid over the
total area. Then for a specific SLA, O3, NO2 and SO2 levels were calculated by adding
O3, NO2 and SO2 levels of all the grid cells together and then dividing the sum by the
total number of cells within the SLA. MapInfo Professional software (MapInfo
Corporation, 2003) was used to display the spatial distributions of air pollution data.
The digital base map datasets used for constructing the GIS were obtained primarily
from the Australian Bureau of Statistics (ABS). These data were manipulated by
thematic mapping to facilitate the accurate identification of the spatial locations of SLA
and their linkages with the other data layers. O3, NO2 and SO2 concentrations in each
SLA were geo-coded to the digital base maps of localities using MapInfo and Microsoft
Access software. The concentrations of gaseous air pollutants which were calculated at
a SLA level, ranged from 26.6 to 34 (ppb) for O3, 18.6 to 23.7 (ppb) for NO2, and 3.9 to
6.1 (ppb) for SO2.
In the study we assumed that annual average concentrations of O3, NO2 and SO2 are
appropriate indicators of long-term exposure to air pollution. As gaseous pollutants
often share the same source or are subject to the same environmental influences, levels
tend to be highly correlated over time. A number of other studies used a similar
approach to estimating long-term exposure to ambient air pollution levels (Chen et al.,
2006; Scoggins et al., 2004; Vedal et al., 2003; Zhang et al., 2006).
104
Mortality data
Mortality data for the same study period were provided by the Office of Economic and
Statistical Research of the Queensland Treasury. The data included date of death; sex,
age, SLA of residence and cause of death. The cause-specific deaths were categorised
according to the International Classification of Diseases Version 9 (ICD-9 code) (used
until July 1999) or Version 10 (ICD-10 code) and were defined as cardiorespiratory
diseases (CRD) (ICD 9: 390- 519; ICD: 10 I00 – I99 and J00 – J99) (including both
respiratory and cardiovascular diseases). During the study period, 51,233 deaths were
recorded, including 27,480 deaths for cardiorespiratory diseases.
To compare the spatial patterns of cardiorespiratory mortality across SLAs, the direct
method (i.e. using the Brisbane population as a reference) was used to calculate the
standardized mortality rate (SMR) for each SLA, adjusted for differences in the age and
sex distributions among SLAs (Selvin, 2001). The equation for calculating a SMR is
∑∑= )(comparisoni
i
de
SMR
Where ∑ ie is the total number of expected cases generated using the reference
population rates for each SLA; ∑ )(comparisonid is the total population in the comparison
group. In this study, firstly, we get the total number of deaths in the SLA by age group
(0-14, 15-64, 65-74 and 75+) and sex; secondly we calculated age-specific rates per
100,000 of CRD deaths (age group) for each SLA; then we calculated the expected
number of deaths in each age group by SLA; and finally we summed the expected
number of deaths and divided the Brisbane population to get SMR per 100,000 for each
105
SLA. A step by step example of how to calculate SMR in 5 SLAs of Brisbane is
provided in Appendix II. In this study, we used same method, but calculated SMR for
whole Brisbane city including 162 SLAs. SMR was used to display geographical
variation of mortality across different SLAs.
Population and socio-economic indexes for areas (SEIFA) data
The 2001 Census provided information on resident population, sex, and age groups by
SLA for the Brisbane city. We also obtained the SEIFA data for each SLA from the
Australian Bureau of Statistics in 2001. SEIFA contained four summary indices
(Advantage/Disadvantage Index, Disadvantage Index, Economic Resources Index, and
Education and Occupation Index) to measure different aspects of socio-economic
conditions by SLA. These indices have high degrees of correlation between each other
(r = 0.87 - 0.97).
In this study, we selected the Disadvantage Index to represent the socio-economic status
of each SLA because of two major reasons: firstly, it focuses on low-income earners,
relatively lower educational attainment, high unemployment and other variables
reflecting disadvantage; and secondly this index was more strongly associated with
cardiorespiratory mortality (r = -0.19) than other three indices of SEIFA (r |≤| 0.16).
Modelling effect of gaseous air pollutants on cardiorespiratory mortality
Pearson correlation was conducted to examine associations of the same pollutant across
different stations or different pollutants in the same station. Logistic regression was
106
used to investigate how gaseous air pollutants (NO2, O3 and SO2) influenced the
probability of deaths, while controlling for confounding effects of age, sex, calendar
year and SEIFA. The dependent variable was the annual mortality rates for each SLA.
The independent variables were the annual means of gaseous air pollutant. Both single
pollutant and multiple pollutants models were performed in relation to cardiorespiratory
mortality. The model was used as follows:
P (Deaths) = Exp(X) / [1+Exp(X)]
Where:
P (Deaths) is the number of deaths divided by the population for each SLA X = β0 + β1
(age) + β2 (sex) + β3 (year) + β4 (SEIFA) + β5 (annual average of gaseous pollutant) for
single pollutant model, and
X = β0 + β1 (age) + β2 (sex) + β3 (year) + β4 (SEIFA) + β5 (NO2) + β6 (O3) + β7 (SO2) for
multiple pollutants model. Separate logistic regression models were run using SAS
statistical software package version 9.1 with LOGISTIC procedure (SAS Institute Inc.,
2002).
6.3 RESULTS
Descriptive statistics for gaseous air pollutants and mortality
The daily maximum one hour NO2, O3 and SO2 concentrations recorded in Brisbane are
presented in Table 6.1. For NO2, the overall average was 19.8 ppb, with the greatest
difference in the average concentrations (17 ppb) observed between the stations of
South Brisbane and North Maclean. For O3, the average was 30.1 ppb, with the greatest
difference in the average concentrations (11.9 ppb) observed between the stations of
107
Flinders View and Brisbane CBD (sci). For SO2, the measures were more homogeneous
across stations (between 2.5 to 7.8 ppb) and average was 5.4 ppb.
Table 6.1 Maximum one hour concentration of gaseous air pollutants in all available
monitoring stations around and in Brisbane city, Australia (1996 – 2004)
NO2 O3 SO2
Monitoring site Area classification Mean SDa Mean SDa Mean SDa
Deception Bay (dcb) residential 16.8 8.6 32.1 9.9 _ _
Eagle Farm (eag) light industrial 21.6 9.7 29.9 10.8 6.4 5.4
Mount Warren Park
(mwp)
residential 15.1 8.1 32.1 12.4 _ _
North Maclean (nmc) rural 11 5.8 35 13.9 _ _
Pinkenba (pin) Residential (adjacent to
industry & airport)
18.9 6.6 29.9 8.7 7.8 7.8
Brisbane CBD (qut) commercial 21.7 8.8 28.2 13.5 4.9 3.9
Flinders View (rfv) residential 19.8 8.1 35.6 13.8 7.7 8.7
Rocklea (roc) residential / light
industrial
20.1 9.1 34.3 13.5 _ _
South Brisbane (sbr) commercial 28 9 _ _ _ _
Brisbane CBD (sci) commercial 25.4 10.4 23.7 7.9 3.2 2.4
Springwood (spr) residential 19.4 7.7 27 8.7 2.5 2.8
Wynnum (wyn) residential (adjacent to
industrial zone)
20.8 10.1 25.8 7.6 5.1 5.4
Zillmere (zil) light industrial area 18.8 8.3 27.1 11.3 _ _
aSD: Standard deviation
Pearson correlations between gaseous pollutants were calculated in three different types
of monitoring stations: Eagle Farm (representing the light industrial area), Brisbane
CBD (qut) (representing the commercial area) and Springwood (representing the
residential area) as they all have long-term monitoring data. The correlations between
daily maximum one hour NO2, O3 and SO2 concentrations in each station ranged from
108
0.10 (O3 concentrations with SO2 for Eagle Farm and O3 concentrations with NO2 for
Brisbane CBD (qut)) to 0.48 (SO2 concentrations with NO2 for Brisbane CBD (qut)).
There were similar results for other stations. For the same gaseous air pollutant across
different monitoring stations, there were moderate to high correlations for NO2
concentrations (ranged from 0.35 to 0.89), and O3 concentrations (ranged from 0.33 to
0.89), but only moderate correlations for SO2 concentrations (ranged from -0.22 to
0.53).
Spatial distribution of gaseous air pollutants and mortality
Figure 6.2 shows the spatial patterns of annual average concentrations of gaseous
pollutants (NO2, O3 and SO2) and annual SMR of cardiorespiratory mortality. The
number of people exposed was shown for each of the gaseous pollutants, and the
average of gaseous pollutants was also shown in each category of cardiorespiratory
mortality at a SLA level.
The highest values of NO2 were recorded around the centre of Brisbane (i.e., Kangaroo
Point and Woolloongabba) (Figure 6.2A). The lowest levels were recorded around
southeast Brisbane (i.e., Parkinson-Drewvale and Stretton-Karawatha).
The spatial pattern of O3 appears opposite to the NO2 plot, although with a different
dynamical range (Figure 6.2B). The negative correlation was found (r = -0.69) between
O3 and NO2. The highest O3 levels were recorded around Rocklea station that located in
an open area of the former Department of Primary Industries Animal Husbandry
Research Farm and is surrounded by light industry and residential areas.
109
The spatial features of SO2 appeared to differ from those of NO2 and O3 (Figure 6.2C).
Negative correlation was also found (r = -0.29) between NO2 and SO2. The highest SO2
levels were observed around Pinkenba and Eagle farm stations, which are located in a
light to moderate industrial area.
There was a strong spatial pattern of cardiorespiratory mortality across SLAs (Figure
6.2D). The average SMR of CRD was 321.2 per 100,000 person annual and the highest
SMR of CRD (i.e., 1256 per 100,000 person–years) was found in Sandgate during the
study period. Sandgate, with 11% of the population aged over 75, is one of the oldest
suburbs in Brisbane and situated on the coastline, along Moreton Bay. There were three
SLAs no CRD records (i.e., Gumdale, Riverhills and Tingapla), with less than 4% of the
population aged over 75 in these areas during the study period.
Modelling the effect of gaseous air pollution on cardiorespiratory mortality
The results of single gaseous pollutant model indicated that there was a positive
relationship between cardiorespiratory mortality and annual average O3 and SO2. After
controlling for potential confounding factors (age, sex, calendar year and SEIFA), there
was an estimated 0.8% (95% CI: 0.0 – 1.5%) increase in cardiorespiratory mortality per
1 ppb increase in annual average concentrations of O3; and 3.5% (95% CI: 0.9 – 6.2%)
increase in cardiorespiratory mortality per 1 ppb increase in annual average
concentrations of SO2. However, there was no significant relationship between annual
average concentrations of NO2 and cardiorespiratory mortality (Table 6.2).
110
A. Max one hour NO2 (ppb) & number of people exposed
B. Max one hour O3 (ppb) & number of people exposed
C. Max one hour SO2 (ppb) & number of people exposed
Colour NO2 (ppb)
Number
of SLAs
Number of
people
exposed
22.23 ~ 23.74 32 172,760
21.42 ~ 22.23 32 202,268
20.84 ~ 21.42 32 178,440
20.38 ~ 20.84 33 150,720
18.63 ~ 20.38 33 183,767
Colour O3 (ppb)
Number
of SLAs
Number of
people
exposed
31.92 ~ 34.01 32 137,443
29.81~ 31.92 32 209,534
29.28 ~ 29.81 32 160,159
28.60 ~ 29.28 33 215,965
26.55 ~ 28.60 33 164,854
Colour SO2 (ppb)
Number of
SLAs
Number of
people
exposed
5.46 ~ 6.14 32 148,263
5.10 ~ 5.46 32 177,814
4.78 ~ 5.10 32 175,144
4.49 ~ 4.78 33 194,915
3.93 ~ 4.49 33 191,819
111
D. SMR of CRD deaths /105 at a SLA level
Figure 6.2 Spatial patterns of annual average concentrations of NO2, O3 and SO2;
annual average SMR of cardiorespiratory mortality and number of people exposed in
Brisbane at a SLA level (1996 - 2004)
In the multiple gaseous pollutants model, there were similar associations between
annual average O3 or SO2 and cardiorespiratory mortality. After controlling for potential
confounding effects of age, sex, calendar year and SEIFA, there was an increase of
0.5% (95% CI: -0.03 – 1.3 %) and 3.1% (95% CI: 0.4 – 5.8%) in cardiorespiratory
mortality for 1 ppb increase in annual average concentrations of O3 and SO2,
respectively (Table 6.2).
Table 6.2 Logistic Regression for gaseous air pollutants and cardiorespiratory mortality disease
Odds Ratio
Pollutant Single pollutant (95% CI)a Multiple pollutants (95% CI)b
NO2 0.990 [0.977‐1.003] 0.993 [0.980‐1.007]
O3 1.008 [1.000‐1.015] 1.005 [0.997‐1.013]
SO2 1.035 [1.009‐1.062] 1.031 [1.004‐1.058] aAdjusted for the confounding effects of age, sex, calendar year and SEIFA
bAdjusted for the confounding effects of age, sex, calendar year, SEIFA and other gaseous air pollutants
SMR of
CRD
Number
of NO2 (ppb) O3 (ppb) SO2 (ppb)
Colour (per 105) SLAs Average Average Average
415~1268 32 21.16 30.06 5.05
308~415 32 21.43 29.82 4.86
251~308 32 21.37 29.84 4.96
200~251 33 21.25 29.93 5.08
0~200 33 21.06 29.93 4.94
112
The regression coefficients are provided in the appendix I.
6.4 DISCUSSION
This study found positive associations between long-term exposure to SO2 or O3 and
cardiorespiratory mortality at a SLA level. This significant association remained even
after simultaneously adjusting for confounding effects of age, sex, calendar year and
SEIFA. However, there was no significant relationship between the annual means of
NO2 and cardiorespiratory mortality in Brisbane.
Spatial distribution of gaseous air pollutants
In order to simulate the urban air quality of Brisbane at a SLA level, the IDW method
was used to estimate the gaseous air pollutant concentrations of O3, NO2 and SO2 for
each SLA (Isaaks and Srivastava, 1989; O’Sullivan and Unwin, 2003). This method can
provide a more complete spatial picture of air quality than monitoring data which is
only available from a few stations. This is necessary for examining the effects of long-
term exposure to air pollution because no single site can represent air quality of a whole
city. Another powerful geo-coding method is kriging, but the results of using IDW and
kriging methods are similar for small number of monitoring stations. In general, the
IDW method is relatively easer and its algorithm is simpler than that produced by
kriging (Wu et al., 2006).
Relationship between long-term exposure to air pollution and mortality
113
Both short- and long-term exposures to air pollution have an effect on mortality and
hospital admissions. However, relatively few studies have addressed the long-term
effects of air pollution on mortality, and most of them have shown an association with
particulate matter (PM) (Fisher et al., 2002; Pope et al., 2002, 2004). Long-term
exposure to PM results in a substantial reduction in life expectancy. Evidently, the long-
term effects clearly have greater significance to public health than the short-term effects
(WHO, 2005). Two studies by Pope et al. (2002, 2004) have shown an association
between long-term exposure to particulate air pollution and mortality from
cardiovascular disease and lung cancer. The strongest association was found for PM2.5
(Pope et al., 2002). There was an estimated increase of 12% and 14% in the deaths
from cardiovascular diseases and lung cancer, respectively, for 10 μg/m3 increase in
PM2.5 (Pope et al., 2002, 2004). A recent study also reported significant impacts of
exposure to PM10 on the survival of acute myocardial infarction patients (Zanobetti and
Schwartz, 2007).
In a German cohort study (Gehring et al., 2006), the long-term exposure to air pollution
was found to be associated with women’s cardiopulmonary (age 50-59 years) mortality.
Scoggins et al. (2004) used urban airshed modelling and GIS-based techniques to
quantify long-term exposure to ambient air pollution levels and its impact on mortality.
After controlling for confounding factors, they found 1.8% increase in circulatory and
respiratory mortality associated with every 1 µg/m3 increase in annual average NO2.
The air pollution levels in Brisbane were lower than most cities in the world. The results
of this spatial analysis still show the statistically significant association between the
annual mean of daily maximum one hour SO2 and cardiorespiratory mortality even after
114
controlling for confounding factors at a SLA level. We also found significant effects of
O3 on cardiorespiratory mortality with a single pollutant model, but only marginally
significant association was observed with a multiple pollutants model.
Socio-economic status and cardiorespiratory mortality
Mortality rates (overall, premature, and from specific causes) and years of life lost have
been shown to increase with decreasing socio-economic status (Turrell and Mathers,
2000). Measures of morbidity (adverse birth outcomes and behavioural disturbances),
health behaviours (smoking prevalence and physical inactivity), and risk factors
(obesity and high cholesterol levels) are also significantly higher in people in the lowest
socio-economic groups compared to the highest ones. There were similar gradients in
inequalities among socio-economic groups in rates of mortality, morbidity and health
behaviours and risk factors in NSW (Moore and Jorm, 2001).
In this study, a composite indicator was used to measure socio-economic status. To
some extent, this would have reflected of other possible confounding factors such as
smoking and occupational exposure to air pollution. Smoking is a risk factor that has a
clear social gradient, with disadvantaged people more likely to be smokers. People in
lower occupational classes are likely to have high smoking prevalence rates.
Furthermore, workers from deprived areas are more likely to occupy positions in the
less skilled occupational categories, which might have higher exposure to air pollution
(Ministry of Health, 2002).
Strengths and limitations in this study
115
Systematically, this study has two major strengths. Firstly, few studies systematically
examine spatial distributions of cardiorespiratory mortality associated with long-term
exposure to gaseous air pollution in a metropolitan setting. Secondly, we have
controlled for most of the known confounding factors including age, sex, calendar year
and SEIFA at a SLA level.
The limitations of this study should also be acknowledged. Firstly, seasonal difference
in air pollution and mortality was not examined as we only used annual average of data.
Secondly, as an ecological study, exposure misclassification bias is inevitable to some
extent. For example, exposure at people’s homes and work places may differ. However,
most of the deaths occurred in the elderly who were likely to stay at home most of the
time. Hence, the extent of misclassification bias may be limited. However, the findings
of this study should be interpreted cautiously, and need to be confirmed by further
research.
6.5 CONCLUSION
This study used GIS and a logistic regression model to analyse the relationship between
long-term exposure to gaseous air pollutants and cardiorespiratory mortality. The results
of this study indicate that long-term exposure to gaseous air pollutants in Brisbane, even
at the levels lower than most of cities (especially SO2), was associated with
cardiorespiratory mortality. Therefore, the spatial features of air pollution and health
outcomes should be considered when modelling air pollution and health relationships,
particularly for large cities.
116
ACKNOWLEDGEMENTS
This study was partly funded by the Australian Research Council (DP0559655). Assoc.
Prof. Shilu Tong is supported by a NHMRC research fellowship. We thank the
Queensland Environmental Protection Agency, the Office of Economic and Statistical
Research of the Queensland Treasury and the Australian Bureau of Statistics for
providing the data for this study.
117
REFERENCES
ABS (Australian Bureau of Statistics), 2001. http://www.abs.gov.au
Brown, C. H., 1983. Asymptotic comparison of missing data procedures for estimating
factor loadings. Psychometrika 48, 269-292.
Burnett, R., Ma, R., Jerrett, M., Goldberg, M., Cakmak, S., Krewski, D. et al., 2001.
The spatial association between community air pollution and mortality: a new
method of analyzing correlated geographic cohort data. Environ. Health Perspect.
109, 375-380.
Chen, L., Verrall, K., Tong, S., 2006. Air particulate pollution due to bushfires and
respiratory hospital admissions in Brisbane, Australia. Int. J. Environ. Health Res.
16, 181-191.
Chen, L., Mengersen, K., Tong, S., 2007. Spatiotemporal relationship between particle
air pollution and respiratory emergency hospital admissions in Brisbane, Australia.
Sci. Total Environ. 373, 57-67.
Chock, D. P., Winkler, S. L., Chen, C., 2000. A study of the association between daily
mortality and ambient air pollutant concentrations in Pittsburgh, Pennsylvania. J.
Air Waste Manage. Assoc. 50, 1481-1500.
Cifuentes, L. A., Vega, J., Kopfer, K., Lave, L. B., 2000. Effect of the fine fraction of
particulate matter versus the coarse mass and other pollutants on daily mortality in
Santiago, Chile. J. Air Waste Manage. Assoc.50, 1287-1298.
Finkelstein, M. M., Jerrett, M., Sears, M. R., 2005. Environmental inequality and
circulatory disease mortality gradients. J. Epidemiol. Community Health 59, 481-
487.
118
Fisher, G., Rolfe, K. A., Kjellstrom, T., Woodward, A., Hales, S., Sturman, A. P.,
Kingham, S., Petersen, J., Shrestha, R., King, D., 2002. Health effects due to motor
vehicle pollution in New Zealand: Report to the Ministry of Transport, 1-72.
Gehring, U. et al., 2006. Long-term Exposure to Ambient Air Pollution and
Cardiopulmonary Mortality in Women. Epidemiology 17, 545-551.
Glover, J. D., Hetzel, D. M., Tennant, S. K., 2004. The socioeconomic gradient and
chronic illness and associated risk factors in Australia. Aust. New Zealand Health
Policy. 1-8.
Hajat, S., Armstrong, B., Wilkinson, P., Busby, A. Dolk, H., 2007. Outdoor air
pollution and infant mortality: analysis of daily time-series data in 10 English cities.
J. Epidemiol. Community Health 61, 719-722.
Hastie, R.T., Tibshirani, R., 1990. Generalized Additive Models. Chapman & Hall,
London.
Hodgson, G., Mathur, S., 2000. Multiple causes of death: Diabetes as a case study. The
10th Biennial conference of the Australian Population Association Population and
Globallsation: Australia in 21st Centry, Melbourne Australia.
Isaaks, E. H., Srivastava, R. M., 1989. An Introduction to Applied Geostatistics. Oxford
University Press, New York, pp. 249-77.
Krewski, D., Burnett, R. T., Goldberg, M. S., Hoover, B. K., Siemiatycki, J., Jerrett, M.
et al., 2003. Overview of the reanalysis of the Harvard Six Cities Study and
American Cancer Society Study of Particulate Air Pollution and Mortality. J.
Toxicol. Environ. Health Part A 66, 1507-1552.
Lee, J. T., Kim, H., Hong, Y. C., Kwon, H. J., Schwartz, J., Christiani, D. C., 2000. Air
pollution and daily mortality in seven major cities of Korea, 1991-1997.
Environ.Res. 84, 247-254.
119
Maheswaran, R., Craglia, M., 2004. GIS in Public Health Practice. CRC Press, Boca
Raton, pp. 125-166.
McCullagh, P., Nelder, J. A., 1989. Generalized Linear Models. Chapman & Hall,
London.
Ministry of Health, 2002. Operational Standard for Ethics Committees. Wellington:
Ministry of Health.
Mitchell, G., Dorling, D., 2003. An environmental justice analysis of British air quality.
Environ. Plan. A 35, 909-929.
Moolgavkar, S. H., 2003. Air pollution and daily mortality in two U.S. counties: season-
specific analyses and exposure-response relationships. Inhal Toxicol. 15, 877-907.
Moore, H., Jorm, L., 2001. Measuring health inequalities in New South Wales. NSW
Public Health Bull, 12, 120-125.
Murray, C. J. L., Lopez, A., 1996. The global burden of disease: a comprehensive
assessment of mortality and disability from disease, injuries and risk factors in 1990
and projected to 2020. Distributed by Harvard University Press, Cambridge, MA,
990 pp.
Neidell, M.J., 2004. Air pollution, health, and socio-economic status: the effect of
outdoor air quality on childhood asthma. J. Health Econ. 23, 1209-1236.
O’Sullivan, D., Unwin, D., 2003. Geographic Information Analysis. Wiley, Hoboken,
N.J. pp. 77-134.
Peng, R. D., Dominici, F., Pastor-Barriuso, R., Zeger, S. L., Samet, J. M., 2005.
Seasonal analyses of air pollution and mortality in 100 US cities. Am. J. Epidemiol.
161, 585-94.
120
Pope, C. A. III, Burnett, R. T., Thun, M. J., Calle, E. E., Krewski, D., Ito, K. et al.,
2002. Lung cancer, cardiopulmonary mortality, and long-term exposure to fine
particulate air pollution. JAMA 287, 1132–1141.
Pope, C. A. III, Burnett, R. T., Thurston, G. D., Thun, M. J., Calle, E. E., Krewski, D. et
al., 2004. Cardiovascular mortality and long-term exposure to particulate air
pollution: epidemiological evidence of general pathophysiological pathways of
disease. Circulation 109, 71–77.
QEPA (Queensland Environmental Protection Agency), 2007.
http://www.epa.qld.gov.au
Roberts, S., 2004. Interactions between particulate air pollution and temperature in air
pollution mortality time series studies. Environ. Res. 96, 328-337.
Scoggins, A., Kjellstrom, T., Fisher, G., Connor, J., Gimson, N., 2004. Spatial analysis
of annual air pollution exposure and mortality. Sci. Total Environ. 321, 71-85.
Selvin, S., 2001. Epidemiologic analysis: a case-oriented approach. Oxford University
Press, New York, pp. 161-185.
Simpson, R., Williams, G., Petroeschevsky, A., Best, T., Morgan, G., Denison, L. et al.,
2005. The short-term effects of air pollution on hospital admissions in four
Australian cities. Aust. N. Z. Publ. Health 29, 213-221.
Taylor, R., Page, A., Morrell, S., Carter, G., Harrison, J., 2004. Socio-economic
differentials in mental disorders and suicide attempts in Australia. Br. J. Psychiatry
185, 486-493.
Turrell, G., Mathers, C. D., 2000. Socioeconomic inequalities in all-cause mortality in
Australia: 1985-97 and 1995-97. Int J Epidemiol. 29, 231-239.
Vedal, S., Brauer, M., White, R., Pethau, J., 2003. Air pollution and daily mortality in a
city with low levels of pollution. Environ. Health Perspect. 111, 45-51.
121
WHO (World Health Organization), 2005. Particulate matter air pollution: how it harms
health. Fact sheet EURO/04/ 05. Berlin, Copenhagen, Rome, 14 April 2005.
Wu, J., Winer, A. M., Delfino, R. J., 2006. Exposure assessment of particulate matter
air pollution before, during, and after the 2003 Southern California wildfires.
Atmos. Environ. 40, 3333-3348.
Yang, C. Y., Chang, C. C., Chuang, H. Y., Tsai, S. S., Wu, T. N., Ho, C. K., 2004.
Relationship between air pollution and daily mortality in a subtropical city: Taipei,
Taiwan. Environ. Int. 30, 519-523.
Zanobetti, A., Schwartz, J., 2007. Particulate Air Pollution, Progression, and Survival
after Myocardial Infarction. Environ. Health Perspect. 115, 769-775.
Zhang, Y., Huang, W., London, S. J., Song, G., Chen, G., Jiang, L., Zhao, N., Chen, B.,
Kan, H., 2006. Ozone and daily mortality in Shanghai, China. Environ. Health
Perspect. 114, 1227–1232.
122
CHAPTER 7: GENERAL DISCUSSION
In each of the previous manuscripts (Chapters 4-6), there is its own discussion section,
where the major findings are discussed in relation to relevant literature; its interpretation;
the strengths and limitations of the analysis; and the public health implications of the
research findings. This chapter discusses the key overall findings from these three
manuscripts. It also discusses future research directions, the strengths and limitations of
the study, and finally makes some conclusions.
7.1 AN OVERVIEW OF KEY FINDINGS FROM THIS STUDY
The findings have been described in each of the previous results chapters. A general
overview is provided below.
We evaluated the long-term air pollution (the daily maximum 1-hour average or daily
24-hour average concentrations of NO2, O3, and PM10) trends using a polynomial
regression model in two monitoring sites (Eagle Farm and Rocklea) in Brisbane,
Australia, between 1980 and 2003 (Chapter 4). The results show that there were
significant up-and-down features for NO2 and O3 concentrations in both sites during
1980 - 1993 and the similar trend was observed for PM10 concentrations in Rocklea
during 1985 – 2003. In Eagle Farm, PM10 gradually decreased from 1994 to 1997 and
then levelled off although there were fluctuations. However relatively stable trends were
observed with NO2, O3 and PM10 concentrations for both sites in recent years. In
123
summary, Rocklea recorded a substantially higher number of days with concentrations
above the relevant daily maximum 1-hour or 24-hour standards than that in Eagle Farm.
Because this station is located in an open area of the former Department of Primary
Industries Animal Husbandry Research farm and is surrounded by light industry and
residential areas. So if there is no effective method to reduce air pollution around
Rocklea area, the residents’ health is likely to be affected. Additionally, the results
show that there is a significant spatial variation of air pollution in these areas. Thus, it is
necessary to examine the spatial features of air pollutants and their health effects in
Brisbane.
Since there has been growing concern about the spatial variation in the relationship
between air pollution and health outcomes, we focused on the analyses at a SLA level
instead at a city level. The GIS and related mapping technologies were used to display
and assess the relationship between air pollution and health outcomes (mortality) in
Chapters 5 and 6.
We examined the spatial variation of SO2 (the daily maximum 1-hour average)
concentrations and mortality in Brisbane, Australia, 1999 – 2001, using GIS techniques
at a statistical local area level (Chapter 5), because ambient SO2 pollution is still a health
concern in many metropolitan areas in Australia (Australian Government Department of
the Environment and Heritage, 2004), few data are available on the spatial patters of
ambient SO2 pollution and mortality, the nature and magnitude of the impact of long-
term exposure to SO2 on mortality remain to be determined. The statistically significant
and positive association between long-term exposure to SO2 and cardiorespiratory
mortality were observed in this study (Table 7.1).
124
Table 7.1 Summary of the relationship between long-term exposure to air pollution and
cardiorespiratory mortality in Brisbane, Australia
Odds Ratio
Studies Disease / Pollutant Single pollutant
(95% CI) Multiple pollutants
(95% CI) SO2 and mortality RD 1.154 (1.083-1.229) 1.069 (0.999-1.143) in Brisbane, Australia CVD 1.115 (1.085-1.147) 1.044 (1.014-1.076) 1999 – 2001 CRD 1.122 (1.093-1.151) 1.048 (1.020-1.077)
Gaseous pollution and NO2 0.990 (0.977-1.003) 0.993 (0.980-1.007)
CRD mortality in Brisbane, O3 1.008 (1.000-1.015) 1.005 (0.997-1.013)
Australia 1996 - 2004 SO2 1.035 (1.009-1.062) 1.031 (1.004-1.058)
The further study about the association of long-term exposure to gaseous air pollutants
(including the daily maximum 1-hour average concentrations of NO2, O3 and SO2) with
cardiorespiratory mortality in Brisbane, Australia, 1996 – 2004 was presented in
Chapter 6. The pollutant concentrations were estimated using GIS techniques at a
statistical local area (SLA) level. Logistic regression was used to investigate the impact
of NO2, O3 and SO2 on cardiorespiratory mortality after adjusting for potential
confounding effects of age, sex, calendar year and socio-economic indexes for areas
(SEIFA). The results indicate that long-term exposure to gaseous air pollutants in
Brisbane, even at the levels lower than most of cities (especially SO2, the means were
little difference between chapter 5 & chapter 6 for same monitoring sites), was
associated with cardiorespiratory mortality (Table 7.1). Therefore, spatial patterns of
gaseous air pollutants and their impact on health outcomes need to be assessed for an
evaluation of long-term effects of air pollution in metropolitan areas.
125
7.2 ALTERNATIVE EXPLANATIONS
7.2.1 Chance
Many different health outcomes (e.g., respiratory, cardiovascular and cardiorespiratory
mortality) were examined and a consistent relationship between long-term exposure to
air pollution and adverse health outcomes was observed in this study. These findings are
unlikely to be explained by chance because of the following considerations:
Firstly, all the statistical significance tests were driven by the hypotheses. We analysed
all available data, and presented a whole spectrum of research findings.
Secondly, there is consistent evidence that long-term exposure to air pollution affects
population health. Such evidence was not only observed for different health outcomes,
but also across different geographic areas.
Finally, rigorous statistical approaches were developed to examine the association
between air pollution and mortality. The logistic regression model for different study
periods produced similar results, which strengthened the validity of the research
findings.
Therefore, the key findings in this study are unlikely to be explained by chance.
7.2.2 Bias
7.2.2.1 Information bias
126
Information bias is likely to occur in the process of data collection, although the disease
surveillance system and air quality monitoring system in Australia are generally
regarded as quite good. The failure to report cases is an important source of information
bias. A case that occurred “elsewhere” but was neglected can cause a problem of under
reporting. For instance, the case register in Brisbane may fail to include residents in
Brisbane who developed the disease in Sydney or other places during their visit there,
even though the number of such cases may be small.
Measurement error in exposure status is another source of information bias. The
measurement of air pollution levels in different monitoring sites is using different
technology. For example, the differential optical absorption spectroscopy (DOAS)
system installed in 1998 at the Brisbane CBD (sci) site was the first of its type in
Australia. DOAS instrumentation operates on a well-established scientific principle,
Beer-Lambert's absorption law, which relates the quantity of light absorbed to the
number of gas molecules in the light path. This technology is used to measure a number
of different pollutants along a single light beam which may be up to 800 metres long
(QEPA, 2007). And other monitoring sites are still using original monitoring system
(except Springwood also used DOAS). As the mortality data (or air pollution data) in
Brisbane were collected by different staff in different locations (or monitoring sites),
over the long observation periods, bias is possible. However, information bias of such
kind is unlikely to have a significant impact on the results of this study because the data
quality is unlikely to change remarkably on the annual basis.
7.2.2.2 Confounding
127
As an ecological study, a number of potential confounders may affect the assessment of
the relationship between air pollution and mortality. These potential confounders
include local health promotion expenditure, air pollution control measures, socio
behaviours and housing conditions, obesity rates and nutrition intake, which might vary
across SLAs, or over time. Temperature or seasonal changes are also potential
confounders. These factors may impact on the incidence of mortality, but the data on
these factors were unavailable at the SLA level for most of the study period in Brisbane.
Moreover, in a small SLA there could be an influence from neighbour SLAs. For
example, as boundaries between SLA often share the same atmosphere, the domestic
activities (e.g., backyard burning and BBQ) will influence the air pollution level on both
sides of border.
7.3 COMPARISON WITH OTHER STUDIES
Previous studies (Ren & Tong, 2006; Ren et al., 2006; R Simpson et al., 2000; R W
Simpson et al., 1997) have examined the association between air pollution and health
outcomes using air pollution data from a single monitoring site or average values from a
few monitoring sites to represent the whole population of the study area. There has been
an increasing concern that the relationships between air pollution and mortality may vary with
geographical area (Burnett et al., 2001; L Chen et al., 2007; Krewski et al., 2003; Maheswaran
& Craglia, 2004; Peng et al., 2005a; Scoggins et al., 2004).
In this study, the IDW method was used to estimate the gaseous air pollutant
concentrations of O3, NO2 and SO2 for each SLA (Isaaks & Srivastava, 1989; O'Sullivan
128
& Unwin, 2003). This method can provide a more complete spatial picture of air quality
than monitoring data which is only available from a few stations. It is necessary for
examining the effects of long-term exposure to air pollution because no single site can
represent air quality of a whole city, particularly for a big city.
Both short-term and long-term exposures to air pollution have an effect on mortality and
hospital admissions. However, relatively few studies have addressed the long-term
effects of air pollution on mortality, and most of the previous studies have shown the
short-term effects of PM on morbidity/mortality (Fisher et al., 2002; Pope et al., 2002;
Pope et al., 2004). Because long-term exposure to PM results in a substantial reduction
in life expectancy, evidently the long-term effects have greater significance to public
health than the short-term effects (WHO, 2005). Studies on the short-term effects of air
pollution suffer some important limitations. In particular, whether increases in mortality
associated with air pollution reflect a short-term shift in deaths that would otherwise
have occurred a few days, weeks or months later. The study (Zanobetti et al., 2003)
found that most of the effects of air pollution are not simply advanced by a few weeks
and that effects persist for more than a month after exposure. Two studies by Pope et al.
(2002, 2004) have shown an association between long-term exposure to particulate air
pollution and mortality from cardiovascular disease and lung cancer. The strongest
association was found for PM2.5 (Pope et al., 2002).
In this study we found the long-term exposure to SO2, even at low levels, affected
cardiorespiratory mortality even after controlling for confounding factors. Pope (2002)
also found positive associations between long-term exposure to SO2 and
cardiorespiratory mortality. There are a few other studies which examined the
129
association between SO2 and mortality (Arribas-Monzon et al., 2001; Fischer et al.,
2003; H Kan & Chen, 2003; H Kan et al., 2004; Venners et al., 2003; T W Wong et al.,
2002). However, most of the previous studies didn’t address the spatial issue.
Scoggins et al (2004) used urban airshed modelling and GIS-based techniques to
quantify long-term exposure to ambient air pollution levels and its impact on mortality.
After controlling for confounding factors, they found the association between annual
average NO2 and circulatory and respiratory mortality. In this study we used the similar
spatial approach method, but did not find the association between NO2 and
cardiorespiratory mortality. In their study they selected 24-hour average NO2
concentrations in one year (1999) with four years (1996-1999) mortality data. In our
study, we selected three years (Chapter 5) and nine years (Chapter 6) maximum 1-hour
NO2 concentrations with same period mortality data.
Mortality rates (overall, premature, and from specific causes) and years of life lost have
been shown to increase with decreasing socio-economic status (Turrell & Mathers,
2001). Measures of morbidity (adverse birth outcomes and behavioural disturbances),
health behaviours (smoking prevalence and physical inactivity), and risk factors
(obesity and high cholesterol levels) are also significantly higher in people in the lower
socio-economic groups compared to higher ones. There were similar gradients in
inequalities among socio-economic groups in rates of mortality, morbidity, unhealthy
behaviours and risk factors in NSW (Moore & Jorm, 2001). Therefore, a composite
indicator was used to measure socio-economic status in this study. To some extent, this
would have reflected of other possible confounding factors such as smoking and
occupational exposure to air pollution. Smoking is a risk factor that has a clear social
gradient, with disadvantaged people more likely to be smokers. Furthermore, workers
130
from deprived areas are more likely to occupy positions in the less skilled occupational
categories, which might have higher exposure to air pollution (Ministry of Health,
2002).
7.4 THE STRENGTHS AND LIMITATIONS
This study has five major strengths. Firstly, to our knowledge, few epidemiological
studies systematically examine spatial distributions of cardiorespiratory mortality
associated with long-term exposure to gaseous air pollution in a metropolitan setting.
Secondly, the datasets used in this study are quite comprehensive; there were less than
5.1% missing values of air pollution data in the available monitoring sites during the
study period. Thirdly, SO2 level in this study was much lower than the National
Standard, but we still observed its effects on the CRD mortality. Fourthly, we have
controlled for most of the known confounding factors including age, sex, SEIFA and
other air pollutants at the SLA level. Finally, research outcomes from this study may
have important implications for public health decision-making in the control and
prevention of the adverse health effects of low level exposure to air pollution.
There are also several limitations in this study, which include possible confounding and
biases in the study. For example, seasonality in air pollution and mortality was not
examined as we only used annual average of data. Measurement errors in exposures of
air pollution might be inevitable to some extent. Population changes occur all the time
and it may impact on the research outcomes. In this study, we used the latest census
data in 2001 because it included sex and age group by each SLA. Like other ecological
studies, misclassification might occur to both health outcomes and exposure. For
131
example, exposure at people’s homes and work places may differ. However, most of the
deaths occurred among the elderly who were likely to stay at home most of the time.
Hence, the extent of misclassification bias may be limited. However, the findings of this
study should be interpreted cautiously, and need to be confirmed by further research.
7.5 PUBLIC HEALTH IMPLICATIONS OF RESEARCH FINDINGS
This study found that long-term exposure to gaseous air pollution (especially SO2), even
at low levels, affected health outcomes (e.g., cardiorespiratory mortality). It may have
two major implications in the planning and implementation of public health
interventions as follows:
1). GIS and spatial analytic approach developed at SLA levels in this study may
be useful in the surveillance of air pollution and health outcomes (e.g.,
morbidity/mortality) to identify and monitor high-risk areas over different
periods of time.
2). The findings of this study suggest that the long-term exposure to air
pollution, even at low levels, is a significant hazard to population health.
Therefore, spatial patterns of air pollutants and their impact on health outcomes
need to be assessed for an evaluation of the health effects of air pollution in
metropolitan areas. Additionally, decision makers can use this information to
identify the priority environmental health issues in relation to air pollution at
local, state and national levels.
132
7.6 DIRECTIONS FOR FUTURE RESEARCH
This study examined the association of long-term exposure to air pollution with
mortality using GIS and logistic regression models. Future research should focus on an
improvement of the relevant methodology in the following areas:
Further studies on the long-term relationship between air pollution and health outcomes
should focus on the advanced methods to reduce measurement errors of exposure. For
example, it is desirable to undertake a prospective cohort study using comprehensive air
pollution monitoring data. In this study, we only used IDW interpolation method for the
annual average concentrations of air pollution at a SLA level. Other spatial methods (eg.,
Kriging, and airshed model) can be used to create the interpolation values of air
pollution in the same SLA areas. The accuracy of spatial monitoring data can be
assessed and adjusted by personal monitoring information. For instance, a proportion of
cohort subjects (eg, 5-10%) can wear a personal monitor for 4 months (a month/season),
and the spatial and personal monitoring data can be compared.
Bayesian hierarchical and generalised estimate equation models have been increasingly
used in air pollution studies (Biggeri et al., 2005; Peng et al., 2005; Roberts & Martin,
2006). However, few studies have been undertaken to examine the similarities and
differences in the outcomes between these models. Further research is needed to
compare the results of the relationship between air pollution and health outcomes using
these different models.
133
A few previous studies have shown that weather conditions and age modify the
association between air pollution and health outcomes (Ren & Tong, 2006; Ren et al.,
2006). It is necessary to further explore the modifications of these and other
characteristics in future studies.
7.7 CONCLUSIONS AND RECOMMENDATIONS
This study examined the relationship between air pollution and health outcomes. GIS
and relevant mapping technologies were used to display the spatial patterns of air
pollution and cardiorespiratory mortality at a SLA level. The results of this study show
that long-term exposure to gaseous air pollutants in Brisbane, even at the levels lower
than most of cities (especially SO2), was associated with cardiorespiratory mortality.
These findings may have important public health implications in the control and
prevention of air pollution-related health effects, since now many countries and
governments have paid more attention to control wide spread air pollution and protect
our environment and human health.
The following recommendations are made through this study:
• The current air pollution monitoring systems should be improved (e.g., reducing
measurement errors or missing records). To keep up with the growing
population in Brisbane, more air pollution monitoring sites are needed (e.g.,
south-western areas).
134
• As motor vehicles cause the most urban air pollution, health education is
necessary to let everyone know and reduce traffic-related air pollution at local
areas (QEPA, 2007).
• GIS and spatial analysis techniques (e.g., airshed model and spatiotemporal
modelling) should be adopted to identify and monitor the communities at high
risk for air pollution-related morbidity/mortality, and to provide valuable
information for researchers and public health decision-making.
135
APPENDIX I
Regression coefficients for cardiorespiratory mortality associated with annual average
SO2 concentrations in Brisbane by SLA (1999 - 2001) (Table 5.3)
Unadjusted Adjusted* Gender Mortality Disease Estimate SE P Estimate SE P All Respiratory 0.143 0.032 0.000 0.066 0.034 0.054 Cardiovascular 0.109 0.014 0.000 0.043 0.015 0.005 Cardiorespiratory 0.115 0.013 0.000 0.047 0.014 0.001 Male Respiratory 0.061 0.046 0.182 -0.013 0.049 0.790 Cardiovascular 0.113 0.022 0.000 0.056 0.023 0.015 Cardiorespiratory 0.103 0.020 0.000 0.043 0.021 0.040 Female Respiratory 0.229 0.046 0.000 0.151 0.049 0.002 Cardiovascular 0.105 0.019 0.000 0.036 0.020 0.073
Cardiorespiratory 0.123 0.018 0.000 0.053 0.019 0.005 *Adjusted for age, disadvantage index and other air pollutants (i.e., PM10, NO2 and O3). Regression coefficients for cardiorespiratory mortality associated with annual average
gaseous pollutants in Brisbane by SLA (1996 - 2004) (Table 6.2)
Signal pollutanta Mutiple pollutantsb Pollutant Estimate SE P Estimate SE P
NO2 -0.010 0.007 0.144 -0.007 0.007 0.344 O3 0.008 0.004 0.053 0.005 0.004 0.188
SO2 0.034 0.013 0.009 0.030 0.013 0.024 aAdjusted for the confounding effects of age, sex, calendar year and SEIFA bAdjusted for the confounding effects of age, sex, calendar year, SEIFA and other gaseous air pollutants
136
APPENDIX II An example of using direct standardization method to calculate cardiorespiratory mortality (CRM) disease of SMR only include five SLAs, not for whole Brisbane (2001) is showed in below tables:
Table 1. Population data by age, sex and total for each SLA (2001) SLA No SLA Name M_AGE1 M_AGE2 M_AGE3 M_AGE4 F_AGE1 F_AGE2 F_AGE3 F_AGE4 TOTAL
1001 Acacia Ridge 823 2,102 224 109 766 2,161 236 194 6,615
1004 Albion 112 786 49 44 195 856 79 89 2,210 1007 Alderley 358 1689 116 136 343 1733 150 229 4,754 1012 Algester 742 2517 148 91 728 2638 182 118 7,164 1015 Annerley 647 3101 218 215 646 3120 234 424 8,605
Total pop 2,682 10,195 755 595 2,678 10,508 881 1,054 29,348
Table 2. Number of CRM by age, sex and total for each SLA (2001) SLA No SLA Name M_AGE1 M_AGE2 M_AGE3 M_AGE4 F_AGE1 F_AGE2 F_AGE3 F_AGE4 TOTAL
1001 Acacia Ridge 0 0 2 4 0 1 3 8 18
1004 Albion 0 2 1 4 0 0 0 0 7 1007 Alderley 0 1 1 5 0 0 1 6 14 1012 Algester 0 2 3 3 0 1 0 2 11 1015 Annerley 0 2 4 17 0 1 1 42 67
Total CRM 0 7 11 33 0 3 5 58 117
Table 3. Calculate CRM rate (/105) by age, sex and total for each SLA (2001) SLA No SLA Name M_AGE1 M_AGE2 M_AGE3 M_AGE4 F_AGE1 F_AGE2 F_AGE3 F_AGE4 TOTAL
1001 Acacia Ridge 0.0 0.0 892.9 3,669.7 0.0 46.3 1,271.2 4,123.7 272.1
1004 Albion 0.0 254.5 2,040.8 9,090.9 0.0 0.0 0.0 0.0 316.7 1007 Alderley 0.0 59.2 862.1 3,676.5 0.0 0.0 666.7 2,620.1 294.5 1012 Algester 0.0 79.5 2,027.0 3,296.7 0.0 37.9 0.0 1,694.9 153.5 1015 Annerley 0.0 64.5 1,834.9 7,907.0 0.0 32.1 427.4 9,905.7 778.6
Total CRM rate 0.0 68.7 1457.0 5546.2 0.0 28.5 567.5 5502.8 398.7
Table 4. Calculate expected CRM number (CRM/105 divided by total pop by age and sex) for each SLA (2001) SLA No SLA Name M_AGE1 M_AGE2 M_AGE3 M_AGE4 F_AGE1 F_AGE2 F_AGE3 F_AGE4 TOTAL
1001 Acacia Ridge 0 0 7 22 0 5 11 43 88
1004 Albion 0 26 15 54 0 0 0 0 95 1007 Alderley 0 6 7 22 0 0 6 28 68 1012 Algester 0 8 15 20 0 4 0 18 65 1015 Annerley 0 7 14 47 0 3 4 104 179
Table 5. Calculate SMR of CRM rate (/105) for each SLA
SLA No SLA Name SMR 1001 Acacia Ridge 300 1004 Albion 325 1007 Alderley 231 1012 Algester 221 1015 Annerley 610
M: male; F: female, AGE1: 0 – 14; AGE 2: 15 – 64; AGE 3: 65 – 74 and AGE 4: 75 +
137
BIBLOGRAPHY
ABS. (2001) ABS (Australian Bureau of Statistics), 2001
http://www.abs.gov.au/ausstats/[email protected]/4079a1bbd2a04b80ca256b9d00
208f92/ab51cb9754fe337aca256bbf0000dbfc!OpenDocument Accessed 7
February, 2005
Ackermann-Liebrich, U. and Rapp, R. (1999) Epidemiological effects of oxiders of
nitrogen, especially NO2 in air pollution and health
Aga, E., Samoli, E., Touloumi, G., Anderson, H. R., Cadum, E., Forsbergz, B.,
Goodman, P. and Goren, A. (2003) Short-term effects of ambient particles on
mortality in the elderly: results from 28 cities in the APHEA2 project, European
Respiratory Journal, 21, 28s-33s.
Arribas-Monzon, F., Rabanaque, M. J., Martos, M. C., Abad, J. M., Alcala-Nalvaiz, T.
and Navarro-Elipe, M. (2001) Effects of air pollution on daily mortality in the
city of Zaragoza, Spain, 1991-1995, Salud Publica Mex, 43, 289-97.
Atkinson, R. W., Anderson, H. R., Sunyer, J., Ayres, J., Baccini, M., Vonk, J. M.,
Boumghar, A., Forastiere, F., Forsberg, B., Touloumi, G., Schwartz, J. and
Katsouyanni, K. (2001) Acute effects of particulate air pollution on respiratory
admissions: results from APHEA 2 project. Air Pollution and Health: a
European Approach, Am J Respir Crit Care Med, 164, 1860-6.
Australian Government Department of the Environment and Heritage. (2004) Australian
Government Department of the Environment and Heritage, State of the Air:
Community Summary 1991-2001
138
http://www.environment.gov.au/atmosphere/airquality/publications/status/index.
html Accessed 8 January, 2007.
Bailey, T. C. and Gatrell, A. C. (1995) Interactive spatial data analysis, New York.
Ballester, F., Tenias, J. M. and Perez-Hoyos, S. (2001) Air pollution and emergency
hospital admissions for cardiovascular diseases in Valencia, Spain, Journal of
epidemiology and community health, 55, 57-65.
Bell, M. L., McDermott, A., Zeger, S. L., Samet, J. M. and Dominici, F. (2004) Ozone
and short-term mortality in 95 US urban communities, 1987-2000, The journal
of the American Medical Association, 292, 2372-8.
Biggeri, A., Baccini, M., Bellini, P. and Terracini, B. (2005) Meta-analysis of the Italian
studies of short-term effects of air pollution (MISA), 1990-1999, Int J Occup
Environ Health, 11, 107-22.
Botto, C., Escalona, E., Vivas-Martinez, S., Behm, V., Delgado, L. and Coronel, P.
(2005) Geographical patterns of onchocerciasis in southern Venezuela:
relationships between environment and infection prevalence, Parassitologia, 47,
145-50.
Briet, O. J., Galappaththy, G. N., Konradsen, F., Amerasinghe, P. H. and Amerasinghe,
F. P. (2005) Maps of the Sri Lanka malaria situation preceding the tsunami and
key aspects to be considered in the emergency phase and beyond, Malar J, 4, 8.
Briggs, D. J., Collins, S., Elliott, P., Fischer, P., Kingham, S., Lebret, E., Pryl, K.,
Vanreeuwijk, H., Smallbone, K. and Vanderveen, A. (1997) Mapping Urban Air
Pollution Using GIS - a Regression-Based Approach, International Journal of
Geographical Information Science, 11, 699-718.
Briggs, D. J. and Elliott, P. (1995) The use of geographical information systems in
studies on environment and health, World Health Stat Q, 48, 85-94.
139
Brunekreef, B., Dockery, D. W. and Krzyzanowski, M. (1995) Epidemiologic studies
on short-term effects of low levels of major ambient air pollution components,
Environ Health Perspect, 103 Suppl 2, 3-13.
Brunekreef, B. and Holgate, S. T. (2002) Air pollution and health, Lancet, 360, 1233-
42.
Burnett, R., Ma, R., Jerrett, M., Goldberg, M. S., Cakmak, S., Pope, C. A., 3rd and
Krewski, D. (2001) The spatial association between community air pollution and
mortality: a new method of analyzing correlated geographic cohort data,
Environmental health perspectives, 109, 375-80.
Carreras-Sospedra, M., Dabdub, D., Rodriguez, M. and Brouwer, J. (2006) Air quality
modeling in the South Coast Air Basin of California: what do the numbers really
mean?, J Air Waste Manag Assoc, 56, 1184-95.
Chan, Y. C., Simpson, R. W., Mctainsh, G. H., Vowles, P. D., Cohen, D. D. and Bailey,
G. M. (1997) Characterisation of chemical species in PM2.5 and PM10 aerosols
in Brisbane, Australia., Atmospheric Environment, 31, 3773-85.
Chan-Yeung, M. N. (2000) Air pollution and health, Hong Kong Med J, 6, 390-8.
Chen, K. S., Ho, Y. T., Lai, C. H., Tsai, Y. A. and Chen, S. J. (2004) Trends in
Concentration of Ground-Level Ozone and Meteorological Conditions during
High Ozone Episodes in the Kao-Ping Airshed, Taiwan., Journal of the Air and
Waste Management Association, 54, 36-48.
Chen, L., Mengersen, K. and Tong, S. (2007) Spatiotemporal relationship between
particle air pollution and respiratory emergency hospital admissions in Brisbane,
Australia, The Science of the total environment, 373, 57-67.
140
Chen, L., Verrall, K. and Tong, S. (2006) Air particulate pollution due to bushfires and
respiratory hospital admissions in Brisbane, Australia, International journal of
environmental health research, 16, 181-91.
Choi, K. M., Christakos, G. and Wilson, M. L. (2006) El Nino effects on influenza
mortality risks in the state of California, Public Health, 120, 505-16.
Christakos, G. and Serre, M. L. (2000) Spatiotemporal analysis of environmental
exposure-health effect associations, J Expo Anal Environ Epidemiol, 10, 168-87.
Chung, K., Yang, D. H. and Bell, R. (2004) Health and GIS: toward spatial statistical
analyses, J Med Syst, 28, 349-60.
Cicero-Fernandez, P., Torres, V., Rosales, A., Cesar, H., Dorland, K., Munoz, R., Uribe,
R. and Martinez, A. P. (2001) Evaluation of human exposure to ambient PM10
in the metropolitan area of Mexico City using a GIS-based methodology, J Air
Waste Manag Assoc, 51, 1586-93.
Cifuentes, L., Borja-Aburto, V. H., Gouveia, N., Thurston, G. and Davis, D. L. (2001)
Assessing the health benefits of urban air pollution reductions associated with
climate change mitigation (2000-2020): Santiago, Sao Paulo, Mexico City, and
New York City, Environ Health Perspect, 109 Suppl 3, 419-25.
Cliff, A. D. and Ord, J. K. (1981) Spatial Processes: Models & Applications, Pion
Limited, London.
Cressie, N. and Read, T. R. C. (1989) Spatial Data Analysis of Regional Counts,
Biometrical Journal, 31, 699-719.
Croner, C. M., Sperling, J. and Broome, F. R. (1996) Geographic information systems
(GIS): new perspectives in understanding human health and environmental
relationships, Stat Med, 15, 1961-77.
141
Dab, W., Medina, S., Quenel, P., Le Moullec, Y., Le Tertre, A., Thelot, B., Monteil, C.,
Lameloise, P., Pirard, P., Momas, I., Ferry, R. and Festy, B. (1996) Short term
respiratory health effects of ambient air pollution: results of the APHEA project
in Paris, J Epidemiol Community Health, 50 Suppl 1, s42-6.
Daniels, M. J., Dominici, F., Samet, J. M. and Zeger, S. L. (2000) Estimating particulate
matter-mortality dose-response curves and threshold levels: an analysis of daily
time-series for the 20 largest US cities, Am J Epidemiol, 152, 397-406.
Daniels, M. J., Dominici, F., Zeger, S. L. and Samet, J. M. (2004) The National
Morbidity, Mortality, and Air Pollution Study. Part III: PM10 concentration-
response curves and thresholds for the 20 largest US cities, Res Rep Health Eff
Inst, 1-21; discussion 23-30.
Denison, L., Simpson, R., Petroeschevsky, A. and Williams, G. (2001) Ambient air
pollution and daily hospital admissions in Melbourne 1994-1997. 789
Dominici, F., Peng, R. D., Bell, M. L., Pham, L., McDermott, A., Zeger, S. L. and
Samet, J. M. (2006) Fine particulate air pollution and hospital admission for
cardiovascular and respiratory diseases, Jama, 295, 1127-34.
English, P., Neutra, R., Scalf, R., Sullivan, M., Waller, L. and Zhu, L. (1999)
Examining associations between childhood asthma and traffic flow using a
geographic information system, Environ Health Perspect, 107, 761-7.
Environmental Protection Agency (2007)
http://www.epa.qld.gov.au/environmental_management/air/air_quality_monitori
ng/southeast_queensland_monitoring_stations/#BrisbaneCBD.
EPA. (2007) EPA (Environmental Protection Agency), 2007.
http://www.epa.qld.gov.au/environmental_management/air/air_quality_monitori
142
ng/southeast_queensland_monitoring_stations/#BrisbaneCBD. Accessed 20 July
2007.
ESRI. (2006) GIS and Mapping Software website http://www.esri.com/index.html
Accessed 22 May, 2006
ESRI. (2007)
http://webhelp.esri.com/arcgisdesktop/9.2/body.cfm?tocVisable=1&ID=3926&T
opicName=Using%20kriging Accessed 20 June 2007.
Firket, J. (1936) Fog along the Meuse Valley, Trans Faraday Society, 32, 1192-1197.
Fischer, P., Hoek, G., Brunekreef, B., Verhoeff, A. and van Wijnen, J. (2003) Air
pollution and mortality in The Netherlands: are the elderly more at risk?, The
European respiratory journal. Supplement, 40, 34s-38s.
Fisher, G., Rolfe, K. A., Kjellstrom, T., Woodward, A., Hales, S., Sturman, A. P.,
Kingham, S., Petersen, J., Shrestha, R. and King, D. (2002) Health effects due to
motor vehicle pollution in New Zealand: Report to the Ministry of Transport
Goovaerts, P. (2006) Geostatistical analysis of disease data: accounting for spatial
support and population density in the isopleth mapping of cancer mortality risk
using area-to-point Poisson kriging, Int J Health Geogr, 5, 52.
Hajat, S., Armstrong, B., Wilkinson, P., Busby, A. and Dolk, H. (2007) Outdoor air
pollution and infant mortality: analysis of daily time-series data in 10 English
cities, J Epidemiol Community Health, 61, 719-22.
Health Effects Institute (2004) Health Effects of Outdoor Air Pollution in Developing
Countries of Asia: A Literature Review
Hoek, G., Brunekreef, B., Goldbohm, S., Fischer, P. and van den Brandt, P. A. (2002)
Association between mortality and indicators of traffic-related air pollution in
the Netherlands: a cohort study, Lancet, 360, 1203-9.
143
Hoek, G., Fischer, P., Van Den Brandt, P., Goldbohm, S. and Brunekreef, B. (2001)
Estimation of long-term average exposure to outdoor air pollution for a cohort
study on mortality, J Expo Anal Environ Epidemiol, 11, 459-69.
Howie, J., Tong, S., Verrall, K., Gerber, R. and Wolff, R. (2005) Air Pollution and
Cardiopulmonary Diseases in Australia: A Review of Epidemiological
Evidence, Environmental health perspectives, 5, 23-35.
Isaaks, E. H. and Srivastava, R. M. (1989) An Introduction to Applied Geostatistics.,
Oxford University Press, New York.
Jo, W. K., Yoon, I. H. and Nam, C. W. (2000) Analysis of air pollution in two major
Korean cities: trends, seasonal variations, daily 1-hour maximum versus other
hour-based concentrations, and standard exceedances., Environmental Pollution,
110, 11-18.
Joines, J. D., Hertz-Picciotto, I., Carey, T. S., Gesler, W. and Suchindran, C. (2003) A
spatial analysis of county-level variation in hospitalization rates for low back
problems in North Carolina, Soc Sci Med, 56, 2541-53.
Jolley, D. J., Jarman, B. and Elliott, P. (1992a) Geographical & environmental
epidemiology methods for small-area studies, Oxford University Press.
Jolley, D. J., Jarman, B. and Elliott, P. (1992b) Socio-economic confounding.
Jolley, D. J., Jarman, B. and Elliott, P. (1996) Socio-economic confounding.
Kan, H. and Chen, B. (2003) Air pollution and daily mortality in Shanghai: a time-
series study, Archives of environmental health, 58, 360-7.
Kan, H. and Chen, B. (2004) Particulate air pollution in urban areas of Shanghai, China:
health-based economic assessment, Sci Total Environ, 322, 71-9.
144
Kan, H., Jia, J. and Chen, B. (2004) The association of daily diabetes mortality and
outdoor air pollution in Shanghai, China, Journal of environmental health, 67,
21-6.
Kan, H. D. and Chen, B. H. (2002) Impact of long-term exposure to air particulate
matter on life expectancy and survival rate of Shanghai residents, Biomed
Environ Sci, 15, 209-14.
Kan, H. D., Chen, B. H. and Jia, J. (2003) A case-crossover study of ambient air
pollution and daily mortality in Shanghai, Zhonghua Liu Xing Bing Xue Za Zhi,
24, 863-7.
Katsouyanni, K., Touloumi, G., Samoli, E., Gryparis, A., Le Tertre, A., Monopolis, Y.,
Rossi, G., Zmirou, D., Ballester, F., Boumghar, A., Anderson, H. R., Wojtyniak,
B., Paldy, A., Braunstein, R., Pekkanen, J., Schindler, C. and Schwartz, J. (2001)
Confounding and effect modification in the short-term effects of ambient
particles on total mortality: results from 29 European cities within the APHEA2
project, Epidemiology, 12, 521-31.
Kitron, U. (1998) Landscape ecology and epidemiology of vector-borne diseases: tools
for spatial analysis, J Med Entomol, 35, 435-45.
Kjellstrom, T. E., Neller, A. and Simpson, R. W. (2002) Air pollution and its health
impacts: the changing panorama, Med J Aust, 177, 604-8.
Krewski, D., Burnett, R., Jerrett, M., Pope, C. A., Rainham, D., Calle, E., Thurston, G.
and Thun, M. (2005) Mortality and long-term exposure to ambient air pollution:
ongoing analyses based on the American Cancer Society cohort, Journal of
toxicology and environmental health. Part A, 68, 1093-109.
Krewski, D., Burnett, R. T., Goldberg, M. S., Hoover, B. K., Siemiatycki, J., Jerrett, M.,
Abrahamowicz, M. and White, W. H. (2003) Overview of the reanalysis of the
145
Harvard Six Cities Study and American Cancer Society Study of Particulate Air
Pollution and Mortality, Journal of toxicology and environmental health. Part A,
66, 1507-51.
Kunzli, N., Kaiser, R., Medina, S., Studnicka, M., Chanel, O., Filliger, P., Herry, M.,
Horak, F., Jr., Puybonnieux-Texier, V., Quenel, P., Schneider, J., Seethaler, R.,
Vergnaud, J. C. and Sommer, H. (2000) Public-health impact of outdoor and
traffic-related air pollution: a European assessment, Lancet, 356, 795-801.
Kuzma, J. W. (1992) Basic statistics for the health sciences, Mayfield, California.
Le Tertre, A., Medina, S., Samoli, E., Forsberg, B., Michelozzi, P., Boumghar, A.,
Vonk, J. M., Bellini, A., Atkinson, R., Ayres, J. G., Sunyer, J., Schwartz, J. and
Katsouyanni, K. (2002) Short-term effects of particulate air pollution on
cardiovascular diseases in eight European cities, J Epidemiol Community
Health, 56, 773-9.
Lee, J. T., Kim, H., Hong, Y. C., Kwon, H. J., Schwartz, J. and Christiani, D. C. (2000)
Air pollution and daily mortality in seven major cities of Korea, 1991-1997,
Environ Res, 84, 247-54.
Mage, D., Wilson, W., Hasselblad, V. and Grant, L. (1999) Assessment of human
exposure to ambient particulate matter, J Air Waste Manag Assoc, 49, 1280-91.
Maheswaran, R. and Craglia, M. (2004) GIS in Public Health Practice., CRC Press,
Boca Raton.
MapInfo Corporation (2003) MapInfo professional software 7.0, New York.
Matejicek, L. (2005) Spatial modelling of air pollution in urban areas with GIS: a case
study on integrated database development, Advances in Geosciences, 4, 63-68.
146
Matejicek, L., Engst, P. and Janour, Z. (2006) A GIS-based approach to spatio-temporal
analysis of environmental pollution in urban areas: a case study of Prague's
environment extended by LIDAR data, Ecological Modelling, 199, 261-277.
Maynard, N. G. and Conway, G. A. (2006) A view from above: use of satellite imagery
to enhance our understanding of potential impacts of climate change on human
health in the Arctic, Alaska Med, 49, 38-43.
Ministry of Health (1954) Mortality and morbidity during the London fog of December
1952. Reports on Public Health and Medical Subjects 95
Ministry of Health. (2002) Operational Standard for Ethics Committees. Wellington:
Ministry of Health. http://www.hrc.govt.nz/assets/pdfs/publications/operational-
standard-for-ethics-committees-updated-edition.pdf Accessed 20 May, 2007
Moore, H. and Jorm, L. (2001) Measuring health inequalities in New South Wales, N S
W Public Health Bull, 12, 120-125.
Morgan, G., Corbett, S. and Wlodarczyk, J. (1998a) Air pollution and hospital
admissions in Sydney, Australia, 1990 to 1994, Am J Public Health, 88, 1761-6.
Morgan, G., Corbett, S., Wlodarczyk, J. and Lewis, P. (1998b) Air pollution and daily
mortality in Sydney, Australia, 1989 through 1993, Am J Public Health, 88, 759-
64.
Murray, C. J. L. and Lopez, A. (1996) The global burden of disease: a comprehensive
assessment of mortality and disability from disease, injuries and risk factors in
1990 and projected to 2020., Distributed by Harvard University Press,
Cambridge, MA,.
Odoi, A., Martin, S. W., Michel, P., Middleton, D., Holt, J. and Wilson, J. (2004)
Investigation of clusters of giardiasis using GIS and a spatial scan statistic, Int J
Health Geogr, 3, 11.
147
O'Sullivan, D. and Unwin, D. (2003) Geographic Information Analysis., John Wiley &
Sons, Inc., Hoboken, N.J.
Pan, X., Yue, W., He, K. and Tong, S. (2007) Health benefit evaluation of the energy
use scenarios in Beijing, China, Sci Total Environ, 374, 242-51.
Peng, R. D., Dominici, F., Pastor-Barriuso, R., Zeger, S. L. and Samet, J. M. (2005)
Seasonal analyses of air pollution and mortality in 100 US cities, Am J
Epidemiol, 161, 585-94.
Peng, R. D., Dominici, F., Pastor-Barriuso, R., Zeger, S. L. and Samet, J. M. (2005a)
Seasonal analyses of air pollution and mortality in 100 US cities, American
journal of epidemiology, 161, 585-94.
Peng, R. D., Dominici, F., Pastor-Barriuso, R., Zeger, S. L. and Samet, J. M. (2005b)
Seasonal analyses of air pollution and mortality in 100 US cities, Am J
Epidemiol, 161, 585-94.
Petroeschevsky, A., Simpson, R. W., Thalib, L. and Rutherford, S. (2001) Associations
between outdoor air pollution and hospital admissions in Brisbane, Australia,
Archives of environmental health, 56, 37-52.
Pikhart, H., Bobak, M., Gorynski, P., Wojtyniak, B., Danova, J., Celko, M. A., Kriz, B.,
Briggs, D. and Elliott, P. (2001) Outdoor sulphur dioxide and respiratory
symptoms in Czech and Polish school children: a small-area study (SAVIAH).
Small-Area Variation in Air Pollution and Health, Int Arch Occup Environ
Health, 74, 574-8.
Pooley, F. and Mille, M. (1999) Composition of air pollution particles
Pope, C. A., 3rd (2000a) Epidemiology of fine particulate air pollution and human
health: biologic mechanisms and who's at risk?, Environ Health Perspect, 108
Suppl 4, 713-23.
148
Pope, C. A., 3rd (2000b) What do epidemiologic findings tell us about health effects of
environmental aerosols?, J Aerosol Med, 13, 335-54.
Pope, C. A., 3rd, Burnett, R. T., Thun, M. J., Calle, E. E., Krewski, D., Ito, K. and
Thurston, G. D. (2002) Lung cancer, cardiopulmonary mortality, and long-term
exposure to fine particulate air pollution, Jama, 287, 1132-41.
Pope, C. A., 3rd, Burnett, R. T., Thurston, G. D., Thun, M. J., Calle, E. E., Krewski, D.
and Godleski, J. J. (2004) Cardiovascular mortality and long-term exposure to
particulate air pollution: epidemiological evidence of general pathophysiological
pathways of disease, Circulation, 109, 71-7.
Powe, N. A. and Willis, K. G. (2004) Mortality and morbidity benefits of air pollution
(SO2 and PM10) absorption attributable to woodland in Britain, Journal of
environmental management, 70, 119-28.
QEPA. (2001) Ambient air quality monitoring in Queensland: 2001 annual summary
and trend report
http://www.epa.qld.gov.au/environmental_management/air/air_quality_monitori
ng/air_quality_reports/ Accessed 10 June 2005.
QEPA. (2007) Queensland Environmental Protection Agency
http://www.epa.qld.gov.au/environmental_management/air/air_quality_monitori
ng/southeast_queensland_monitoring_stations/#BrisbaneCBD. Accessed 20 July
2007.
Ren, C. and Tong, S. (2006) Temperature modifies the health effects of particulate
matter in Brisbane, Australia, International journal of biometeorology, 51, 87-
96.
149
Ren, C., Williams, G. M. and Tong, S. (2006) Does particulate matter modify the
association between temperature and cardiorespiratory diseases?, Environmental
health perspectives, 114, 1690-6.
Rob, M. A. (2003) Applications of Geographical Information Systems in Understanding
Spatial Distribution of Asthma, Informing Science, 6, 89-99.
Roberts, S. and Martin, M. A. (2006) Applying a moving total mortality count to the
cities in the NMMAPS database to estimate the mortality effects of particulate
matter air pollution, Occup Environ Med, 63, 193-7.
Rutherford, S., Simpson, R., Williams, G., Mitchell, C. and McCall, B. (2000)
Relationships between environmental factors and lung function of asthmatic
subjects in south east Queensland, Australia, J Occup Environ Med, 42, 882-91.
Samet, J. M., Dominici, F., Curriero, F. C., Coursac, I. and Zeger, S. L. (2000c) Fine
particulate air pollution and mortality in 20 U.S. cities, 1987-1994, N Engl J
Med, 343, 1742-9.
Samet, J. M., Dominici, F., Zeger, S. L., Schwartz, J. and Dockery, D. W. (2000a) The
National Morbidity, Mortality, and Air Pollution Study. Part I: Methods and
methodologic issues, Res Rep Health Eff Inst, 5-14; discussion 75-84.
Samet, J. M., Zeger, S. L., Dominici, F., Curriero, F., Coursac, I., Dockery, D. W.,
Schwartz, J. and Zanobetti, A. (2000b) The National Morbidity, Mortality, and
Air Pollution Study. Part II: Morbidity and mortality from air pollution in the
United States, Res Rep Health Eff Inst, 94, 5-70; discussion 71-9.
Schrenk, H., Heimann, H., Clayton, G., Gafafer, W. and Wexler, H. (1949) Air
Pollution in Donora, PA: Epidemiology of the Unusual Smog Episode of
October 1948, Preliminary Report
150
Schwela, D. (2000) Air pollution and health in urban areas, Rev Environ Health, 15, 13-
42.
Scoggins, A., Kjellstrom, T., Fisher, G., Connor, J. and Gimson, N. (2004) Spatial
analysis of annual air pollution exposure and mortality, The Science of the total
environment, 321, 71-85.
Selvin, S. (2001) Epidemiologic analysis: a case-oriented approach., Oxford University
Press, New York.
Semaan, S., Sternberg, M., Zaidi, A. and Aral, S. O. (2007) Social capital and rates of
gonorrhea and syphilis in the United States: spatial regression analyses of state-
level associations, Soc Sci Med, 64, 2324-41.
Sim, M. (2002) Bushfires: are we doing enough to reduce the human impact?, Occup
Environ Med, 59, 215-6.
Simpson, R., Denison, L., Petroeschevsky, A., Thalib, L. and Williams, G. (2000)
Effects of ambient particle pollution on daily mortality in Melbourne, 1991-
1996, J Expo Anal Environ Epidemiol, 10, 488-96.
Simpson, R., Williams, G., Petroeschevsky, A., Best, T., Morgan, G., Denison, L.,
Hinwood, A. and Neville, G. (2005) The short-term effects of air pollution on
hospital admissions in four Australian cities, Australian and New Zealand
journal of public health, 29, 213-21.
Simpson, R. W., Williams, G., Petroeschevsky, A., Morgan, G. and Rutherford, S.
(1997) Associations between outdoor air pollution and daily mortality in
Brisbane, Australia, Arch Environ Health, 52, 442-54.
Sullivan, R. and Beudeker, B. (1999) New Directions in Air Quality Impact Assessment
in Australia, clean Air, 33, 32-35.
151
Talbot, T. O., Kulldorff, M., Forand, S. P. and Haley, V. B. (2000) Evaluation of spatial
filters to create smoothed maps of health data, Stat Med, 19, 2399-408.
Thurston, G. D. and Ito, K. (2001) Epidemiological studies of acute ozone exposures
and mortality, J Expo Anal Environ Epidemiol, 11, 286-94.
Touloumi, G., Katsouyanni, K., Zmirou, D., Schwartz, J., Spix, C., de Leon, A. P.,
Tobias, A., Quennel, P., Rabczenko, D., Bacharova, L., Bisanti, L., Vonk, J. M.
and Ponka, A. (1997) Short-term effects of ambient oxidant exposure on
mortality: a combined analysis within the APHEA project. Air Pollution and
Health: a European Approach, Am J Epidemiol, 146, 177-85.
Turrell, G. and Mathers, C. (2001) Socioeconomic inequalities in all-cause and specific-
cause mortality in Australia: 1985-1987 and 1995-1997, Int J Epidemiol, 30,
231-9.
Vedal, S., Brauer, M., White, R. and Petkau, J. (2003) Air pollution and daily mortality
in a city with low levels of pollution, Environ Health Perspect, 111, 45-52.
Venners, S. A., Wang, B., Xu, Z., Schlatter, Y., Wang, L. and Xu, X. (2003) Particulate
matter, sulfur dioxide, and daily mortality in Chongqing, China, Environmental
health perspectives, 111, 562-7.
Vine, M. F., Degnan, D. and Hanchette, C. (1997) Geographic information systems:
their use in environmental epidemiologic research, Environ Health Perspect,
105, 598-605.
WHO (2000) Air Quality Guidelines for Europe, 2nd, WHO reg publ eur ser;, 91, 1-
287.
WHO. (2005) Particulate matter air pollution: how it harms health
http://www.euro.who.int/document/mediacentre/fs0405e.pdf Accessed 14 April,
2005
152
Wise, E. K. and Comrie, A. C. (2005) Meteorologically adjusted urban air quality trends
in the Southwestern United States., Atmospheric Environment, 39, 2969-2980.
Wong, E. Y., Ponce, R. A., Farrow, S., Bartell, S. M., Lee, R. C. and Faustman, E. M.
(2003) Comparative risk and policy analysis in environmental health, Risk Anal,
23, 1337-49.
Wong, T. W., Tam, W. S., Yu, T. S. and Wong, A. H. (2002) Associations between
daily mortalities from respiratory and cardiovascular diseases and air pollution
in Hong Kong, China, Occupational and environmental medicine, 59, 30-5.
World Health Organization (1997) Health and environment in sustainable development
: five years after the earth summit., Geneva : World Health Organization.
Wu, J., Winer, A. M. and Delfino R, J. (2006) Exposure assessment of particulate
matter air pollution before, during, and after the 2003 Southern California
wildfires, Atmospheric environment, 40, 3333-3348.
Yan, C., Wu, S., Shen, X., Zhang, Y., Jiang, F., Yin, J., Zhou, J., He, J., Ao, L. and Li,
R. (2002) The trends of changes in children's blood lead levels since the
introduction of lead free gasoline in Shanghai, Zhonghua Liu Xing Bing Xue Za
Zhi, 23, 172-4.
Yang, C. Y., Chang, C. C., Chuang, H. Y., Tsai, S. S., Wu, T. N. and Ho, C. K. (2004)
Relationship between air pollution and daily mortality in a subtropical city:
Taipei, Taiwan, Environ Int, 30, 519-23.
Yiannakoulias, N., Rowe, B. H., Svenson, L. W., Schopflocher, D. P., Kelly, K. and
Voaklander, D. C. (2003) Zones of prevention: the geography of fall injuries in
the elderly, Soc Sci Med, 57, 2065-73.
Yuval, D. B. (2006) High-resolution spatial patterns of long-term mean concentrations
of air pollutants in Haifa Bay area, Atmospheric Environment, 40, 3653-64.
153
Zanobetti, A., Schwartz, J., Samoli, E., Gryparis, A., Touloumi, G., Peacock, J.,
Anderson, R. H., Le Tertre, A., Bobros, J., Celko, M., Goren, A., Forsberg, B.,
Michelozzi, P., Rabczenko, D., Hoyos, S. P., Wichmann, H. E. and Katsouyanni,
K. (2003) The temporal pattern of respiratory and heart disease mortality in
response to air pollution, Environ Health Perspect, 111, 1188-93.