Using GIS to measure inequalities in access to healthcare
This link will direct you to a paper that was published by Delamater et al. in the International Journal of Health Geographics in 2012.
This paper describes how inequalities in access to healthcare in areas of Michigan were measured using two different GIS approaches. Limited access to healthcare was defined as being located more than 30 minutes from the nearest acute care hospital. The authors applied both raster and network data approaches to compare modeled estimates of areas of limited access. Assumptions made by the two GIS methodologies are reviewed, as they are important in estimating access to health services. Conceptual and practical differences were also reviewed for service area estimates produced using the two models.
While both approaches identified similar areas of limited accessibility, the raster method identified a larger area and more people with limited access than the network-based model did. Although the raster model was more sensitive to travel speed settings, the network-based model demonstrated greater sensitivity to the “population assignment method” used in Michigan.
Five categories of assumptions made in health accessibility studies are discussed, including:
The fact that all people are assumed to be similar. This is not necessarily true as, for example, disabled or elderly people can experience more difficulties in reaching health care.
There are many factors that influence travel time like time of day (traffic jams) or weather conditions (hurricanes). A choice has to be made if this variation in conditions should be considered.
Often the assumption is made that all people possess knowledge on how to access health care, which may not be the case.
A model can assume that people travel along the shortest path between locations (e.g. their home and a hospital) but this is not necessarily true. A mother will first drop off her children at school before going to see a doctor. This is often referred to as multipurpose trips.
Population is often assigned to a single point location. This location can be a house, but can also be a random point in the district a person lives in when the home address is unknown. This introduces loss of quality in a model. Another related problem is the assumption that the start location of a person is their home; this could however also be their workplace.
Making appropriate, informed assumptions is essential for modeling access to health care, and it is impossible to model without them. But for each individual research question, good choices have to be made.