BEST PRACTICES - MAPPING HEALTH DATA

Working with line data

Line data frequently represent street networks in a GIS and are commonly used to investigate healthcare resource access and allocation as well as using distance to roads in exposure scenarios. Line data can be used to map the most efficient routes for healthcare service delivery, determine time or travel distance between populations and a set of locations within a healthcare system or they can be used for creating funding models (i.e. the cost-benefit of funding a new facility in a remote area versus additional travel time/cost to reach one).

There are a number of techniques for assigning population areas to a street network file in GIS. Three commonly used methods are Euclidian distance, network analysis and travel time analysis. There are things you need to think about when working with line data:

Data projection
Calculating distance ‘as the crow flies’ (or Euclidean Distance)
Network distance
Travel time analysis
Software considerations

Data projection

The first thing to consider for calculating distance before using any of these methods is the projection that your map data is in. Depending on the distance being measured, results could vary depending on the chosen map projection. Typically, Albers equal area conic is the projection used for BC data. If calculations larger than BC are required, this may not be the best projection to use (see references below for further information).

Euclidean distance

The simplest indicator of accessibility is the "crow fly" distance. Crow fly distances can be calculated between two points using either Euclidean distance or as Great Circle distances. Calculating distance ‘as the crow flies’ does not account for geography, such as elevation changes or differing road conditions, that might impact the amount of time it takes for a person to access a healthcare facility. This method assumes equal access times among all population included within the specified distance.

However, analyses have also found that the outcome measures using “crow fly” distances produce similar results when the distances between the start and end points is small, such as block distances in urban areas. Each project will have a different spatial distribution of data and the context and needs of the project should be carefully considered when selected robust or simple distance analysis techniques.

Crow Fly and Network distance calculations

Calculating network distance

While ‘as the crow flies’ is a simple method for calculating distance, it does not necessarily reflect the travel and accessibility usually occur along a road network. Network analysis is another approach which uses street road networks. Network analysis can incorporate road elevation changes or differing road conditions and can also model marked speed limits or assign travel-time delays at street intersections to provide a more realistic estimate of travel time.

Travel time analysis

This calculates the quickest and/or shortest route (as long as speed is an attribute in the road network files). Isochrones, contours used to create drive-time zones around service centres, are used to estimate travel time out to potential populations.

Software for building your network data

Modern GIS software interfaces are well equipped for simulating real-world travel along a transportation network, although add-on software is required. Road network geographic files are also required. To maximize the transportation simulation you should ensure that your street network data contains a number of attributes that can be used to estimate real world impedances over the road network. These attributes commonly represent restrictions such as road type (e.g. highways, local streets, logging roads), one-way streets, or speed limits. Most GIS transportation network engines can automatically find this information in the street network attribute tables, but in some instances it will be necessary for you to manually highlight which column in your spatial data should be used to assist in navigation.

References

The following references demonstrate different travel algorithms for constructing both precision estimates and hospital catchments. Both provide examples of variations based on different travel algorithms, including crow-fly distances.

Online module on map projections by Laurie A. Garo at the University of Texas.

Chalasani VS, Denstadli JM, Engebretsen Ø, and Axhausen KW. Precision of Geocoded Locations and Network Distance Estimates. BTS Journal of Transportation and Statistics. 8 (2) 2005

Schuurman N, Fiedler R, et al. (2006). "Defining rational hospital catchments for non-urban areas based on travel-time." International Journal of Health Geographics 5(43): 1-11.