Exploring zone design methods for a small-area

ISEE Symposium August 2002
Exploring zone design methods for a small-area environmental epidemiological study
Cockings, S1, Poulstrup, A2, Martin, D1, Hansen, HL2
1 Department of Geography, University of Southampton, UK; 2 Medical Office of Health, Province of
Vejle, Denmark
Objective: To explore whether methods that have recently been developed for the design of
geographical areas for the 2001 UK Census of Population can also be employed to develop purposespecific sets of geographical units for use in environmental epidemiological studies.
Introduction: Many epidemiological investigations use data aggregated to a set of geographical units.
This is generally because individual level data are not available (often due to confidentiality
constraints) or because area-level effects are important to the relationships under investigation or
simply because we wish to calculate rates within an area. The design of such geographical areas (in
terms of their size, shape and the characteristics of the population that they contain) can have a
significant influence on the relationships observed between environmental factors, socio-economic
factors and health and on any practical or policy decisions emanating from such research. Whilst
recent studies have demonstrated the usefulness of zone design methods for the collection and
dissemination of Census data, few have explored the potential usefulness of such techniques for
designing zones specifically for health studies. The research presented here uses these zone design
methods to develop geographical areas for use in a study investigating the possible health effects of
exposure to air pollution in a city (Kolding) in Denmark.
Methods: The existence of a Central Population Register (CPR) in Denmark (where all individuals are
identified by a unique number and which is updated daily), together with geo-referencing of all
addresses, means that the place of residence of all individuals is known to a very high spatial and
temporal resolution. Residents living within the city during the time period of interest are identified
using these datasets. A set of building blocks is then developed from these points (the georeferenced addresses) by drawing (Thiessen) polygons around the points using geographical
information systems (GIS) techniques. A recently developed automated zone matching (AZM) tool is
then used to aggregate these building blocks into sets of geographical units according to usercontrolled zone design criteria. Design parameters used include: meeting minimum population
thresholds, seeking to minimise variation in population size between units, maintaining features of the
natural and built environment as boundaries of units, and maximising socio-economic homogeneity
within units.
Results: Sets of units are produced according to the varying zone design parameters. The
characteristics of the units (in terms of shape, size and characteristics of the population within them)
are compared. Selected sets of units are then employed to calculate rates of cancers within the city
and the sensitivity of these rates to the zone design parameters is explored.
Conclusion: Zone design methods and tools developed for other applications provide the
opportunity for users within the health field to design purpose-specific sets of geographical areas for
use in environmental epidemiological investigations.