4 Spatial Modelling Approaches in Air Quality
Epidemiological Studies
The major challenges involving the health effects of air pollution from industrial,
traffic and other conventional and non-conventional heating sources are being the
focus of considerable public and scientific attention. The modelling approaches used
for assessing the impact of exposure on human health largely depends on the kind of
data sources and how the data collection process is happening over space and/or
time. Data used in studies to evaluate the exposure–risk relationship generally
include disease events, measurements on pollutants or other risk factors and the
reference population at risk, along with other covariates describing socio-demographic features.
Air pollution concentrations are measured at routine intervals at a number of
monitoring sites across a region of interest producing point reference data or are
simulated by numerical models over a preset grid of locations. On the other hand,
health data are mostly collected at an aggregated level and are usually associated
with irregularly shaped administrative regions. Epidemiological investigations
where the associations between disease occurrence and environmental risk factors
over aggregated groups are studied rather than at individual levels are termed as
ecological studies (Cameletti et al. 2019).
In statistical epidemiology, health data are usually collected either as count data
or at individual level (case event data). In the latter case, two major types of studies
can be distinguished. Those based on analysing the spatial pattern of case event
locations using point processes (Grell et al. 2015; Diggle 1990, 2003; Pinto Junior
et al. 2015), and the other is cohort studies which are useful in quantifying the health
effects of long-term exposure (months or years) to pollution (Molitor et al. 2006;
Hoek et al. 2013).
There are several challenges involved in studies related to the case event data.
Firstly, the data are not easily available involving high costs to obtain due to the large
amount of data required and the anticipated expenses using the GIS and global
positioning system (GPS)-based devices for survey and data collection (Nuckols
et al. 2004). Secondly, the location of the case home address could not be possibly
related to the disease aetiology. For instance, the disease could have been contracted
at the work place because of occupational exposure or could be related to the amount
of time spent by individuals in several indoor and outdoor microenvironments
characterised by different pollutant concentrations.
In this regard some authors have proposed models for personal exposure based on
longitudinal panel studies, time-activity diary data or GPS-enabled individual exposure monitor (McBride et al. 2007; Zidek et al. 2007 and Blangiardo et al. 2011). As
a result of these drawbacks related to individual data, either time series or areal unit
study designs have been suggested to quantify the short- and long-term health
impacts of air pollution. These studies utilise population level summaries rather
than individual level data and cannot be used to quantify individual level cause and
effect.
382
M. Prashanthi Devi et al.
Epidemiological Studies
The major challenges involving the health effects of air pollution from industrial,
traffic and other conventional and non-conventional heating sources are being the
focus of considerable public and scientific attention. The modelling approaches used
for assessing the impact of exposure on human health largely depends on the kind of
data sources and how the data collection process is happening over space and/or
time. Data used in studies to evaluate the exposure–risk relationship generally
include disease events, measurements on pollutants or other risk factors and the
reference population at risk, along with other covariates describing socio-demographic features.
Air pollution concentrations are measured at routine intervals at a number of
monitoring sites across a region of interest producing point reference data or are
simulated by numerical models over a preset grid of locations. On the other hand,
health data are mostly collected at an aggregated level and are usually associated
with irregularly shaped administrative regions. Epidemiological investigations
where the associations between disease occurrence and environmental risk factors
over aggregated groups are studied rather than at individual levels are termed as
ecological studies (Cameletti et al. 2019).
In statistical epidemiology, health data are usually collected either as count data
or at individual level (case event data). In the latter case, two major types of studies
can be distinguished. Those based on analysing the spatial pattern of case event
locations using point processes (Grell et al. 2015; Diggle 1990, 2003; Pinto Junior
et al. 2015), and the other is cohort studies which are useful in quantifying the health
effects of long-term exposure (months or years) to pollution (Molitor et al. 2006;
Hoek et al. 2013).
There are several challenges involved in studies related to the case event data.
Firstly, the data are not easily available involving high costs to obtain due to the large
amount of data required and the anticipated expenses using the GIS and global
positioning system (GPS)-based devices for survey and data collection (Nuckols
et al. 2004). Secondly, the location of the case home address could not be possibly
related to the disease aetiology. For instance, the disease could have been contracted
at the work place because of occupational exposure or could be related to the amount
of time spent by individuals in several indoor and outdoor microenvironments
characterised by different pollutant concentrations.
In this regard some authors have proposed models for personal exposure based on
longitudinal panel studies, time-activity diary data or GPS-enabled individual exposure monitor (McBride et al. 2007; Zidek et al. 2007 and Blangiardo et al. 2011). As
a result of these drawbacks related to individual data, either time series or areal unit
study designs have been suggested to quantify the short- and long-term health
impacts of air pollution. These studies utilise population level summaries rather
than individual level data and cannot be used to quantify individual level cause and
effect.
382
M. Prashanthi Devi et al.