Research
Validation of travel times to hospital estimated by GIS
-
* Corresponding author: Robin Haynes R.Haynes@uea.ac.uk
1 School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ, UK
2 Respiratory Diseases Department, Centre for Infections, Health Protection Agency, 61 Colindale Avenue, London NW9 5EQ, UK
International Journal of Health Geographics 2006, 5:40 doi:10.1186/1476-072X-5-40
Published: 19 September 2006Abstract
Background
An increasing number of studies use GIS estimates of car travel times to health services, without presenting any evidence that the estimates are representative of real travel times. This investigation compared GIS estimates of travel times with the actual times reported by a sample of 475 cancer patients who had travelled by car to attend clinics at eight hospitals in the North of England.
Methods
Car travel times were estimated by GIS using the shortest road route between home address and hospital and average speed assumptions. These estimates were compared with reported journey times and straight line distances using graphical, correlation and regression techniques.
Results
There was a moderately strong association between reported times and estimated travel times (r = 0.856). Reported travel times were similarly related to straight line distances. Altogether, 50% of travel time estimates were within five minutes of the time reported by respondents, 77% were within ten minutes and 90% were within fifteen minutes. The distribution of over- and under-estimates was symmetrical, but estimated times tended to be longer than reported times with increasing distance from hospital. Almost all respondents rounded their travel time to the nearest five or ten minutes. The reason for many cases of reported journey times exceeding the estimated times was confirmed by respondents' comments as traffic congestion.
Conclusion
GIS estimates of car travel times were moderately close approximations to reported times. GIS travel time estimates may be superior to reported travel times for modelling purposes because reported times contain errors and can reflect unusual circumstances. Comparison with reported times did not suggest that estimated times were a more sensitive measure than straight line distance.