5.4 Representing the patient travel system

This section of the course discusses different GIS approaches to modelling patients’ geographic accessibility to health services. Because there are numerous modes of transport that patients might use to access health services, very different GIS models of patient travel have been developed. Here, we consider GIS models for journeys made on foot, journeys by private car, and journeys by scheduled public transport.

A simple GIS model for patients’ journey time to a health facility on foot can be constructed using a Euclidean distance function. Such functions calculate the straight-line distance from each cell in a raster grid to a specified point (in this case, a health facility), which can then be multiplied by an average walking speed to produce an estimate of journey time. Such a model assumes patients can traverse the landscape in any direction at a constant speed. A more realistic approach is to incorporate landscape features that determine journey time such as gradient, barriers, and the footpath network by defining an impedance or cost surface. This is a raster grid that represents the journey time or cost associated with traversing each cell across a modelled landscape. Rather than every cell having a uniform walking speed associated with it, individual cells are assigned different values of impedance base on their gradient or land cover. Cells that form part of a footpath or road network can be assigned low impedances, whilst cells away from paths or on steep terrain can be assigned very high impedances. Cells representing a complete barrier such as a river can be masked out, completely blocking travel across them. This cost surface is then input into a shortest-path algorithm within the GIS. These algorithms determine the shortest route between each cell and the point (e.g. health facility) of interest. By knowing which cells a patient must travel through, and the impedances associated with each, the algorithm also calculates the journey time from each cell to the health facility. The box below illustrates how a cost surface might be constructed and used with a shortest-path algorithm to estimate journey time.

Journeys made by car are generally represented in a different way in a GIS to those made on foot. Pedestrian models often use raster surfaces to allow journeys to be modelled across the land surface as a whole. In contrast, we know that journeys made by car are likely to be restricted to the road network. This important difference means that car journeys are generally modelled using a vector network representation in a GIS rather than a raster surface. In a network representation, each vector segment of road is assigned an average speed, and the time required to travel along each segment is calculated. Journey time between two points is then estimated using network analysis functions within the GIS. These functions determine the fastest route between the two points in question across the network, and calculate the time required to travel along that route.

Journeys made by public transport can also be represented in a GIS, although in a different way again to either private car or pedestrian journeys (Tanser et al., 2006). Public transport journeys are constrained not only to a given road or rail network, but also by a set of pre-defined journey schedules between specified destinations, as defined by the service timetable. If timetables of public transport services are known and can be queried within a GIS, then the expected journey time between public transport hubs (bus stops, train stations etc) can be determined. Journeys made using multiple transport services can be modelled by cross-referencing between different timetables. An important aspect of modelling journeys by public transport is that the complete door-to-door journey often involves sections made on foot. As such, a pedestrian journey model can be used in conjunction with a public transport model to provide a more realistic representation (Martin et al., 2006).

---

---