7.2 Catchment models

One straightforward means of planning the location of health service facilities is to develop maps of catchment areas. A catchment area represents the geographic extent of the population served by a given facility. There are various techniques for estimating such areas using GIS, which are sometimes referred to as catchment models. There are two types of catchment areas: mandated service areas, the area that a health facility is required to serve by a health ministry and ‘natural’ service areas, the area that the users of a given health facility reside in. Maps of catchment areas can be used by planners to inform decisions about health services and where they are located. For example, in modelling ‘natural’ catchment areas, GIS can help identify areas of population that are currently poorly served by existing health facilities.

Whilst it is not uncommon to find manually drawn catchment maps for health facilities (typically representing mandated service areas), it is also possible to delineate catchment areas automatically using GIS, based on a variety of techniques with increasing degrees of complexity:

Thiessen polygons : the most simple technique involves calculating the area that is closer to a given health facility than any other facility in an area. This approach, also known as Voronoi tessellation, uses straight-line (Euclidean) distances in calculating catchment areas. In reality, of course, patients make use of transport networks to reach health facilities and there are likely to be barriers to movement (such as rivers, for example) so a straight-line approach is unrealistic.

 
Cost surfaces and networks : a more complex technique is to develop a cost surface, which represents the degree of difficulty of moving across a given area. A cost surface is essentially a raster grid in which each grid cell contains a ‘friction’ value, representing the degree of difficulty of movement. It is possible to calculate the path with the least ‘friction’ from a given patient’s address to a given health facility. By identifying the health facility that has the least-cost path from a given address, it is possible to calculate the catchment area for a specific facility. For example, Schuurman et al (2006) used cost surfaces to define catchment areas for health facilities in British Columbia .

Network analysis can also be used to represent the transport network. In a network approach, travel times are attached to sections of road and travel times are calculated from patient addresses to health facilities along this road network. The approach can be thought of as an equivalent, vector-based procedure to the cost surface raster approach.

Cost surfaces and networks thus take into account the presence of transport networks. Typically, catchment maps derived from cost surfaces, networks, and as Thiessen polygons have sharp boundaries. In reality, catchment areas rarely have such ‘sharp’ boundaries and the population from a given area may well make use of several different health facilities. Furthermore, in choosing between different facilities, patients will respond to the characteristics of particular facilities. A hospital with a high success rate in a particular surgical operation is thus likely to attract more patients for that operation than facilities elsewhere.

Patient records : An alternative approach to identifying catchments is to make use of existing patient records. For example, Peters and Hall (1999) identified ambulance catchment areas by examining call-out patterns from past emergency calls. Based on these records, they were able to map the catchment areas in which particular stations had responded to 75% or more of calls.

There are variants on all of these techniques for mapping catchments. Exactly how a catchment is defined can have a substantial impact when calculating physician: population ratios and in calculating disease rates per head of population.