Improving geosocial clustering analyses Peter English, Public Health Medicine Environment Group 18 June 2007 I can claim little expertise in this area, but the geographical clustering analysis used in this study, and in other methods that I've come across, use a fairly crude method of assessing proximity - in this case, crude distance; in some other methods using a grid of some sort.In practice, when trying to decide whether cases are likely to be linked, the geo-social clustering is often more complicated. Sometimes two locations are very close together as the crow flies, but separated by a main road, railway, or river; and in practice might as well be miles apart. Conversely, places some distance apart might be connected by excellent transport links, and should be treated as relatively cloer than crude distance implies.The best system that I've come across for capturing some of this geosocial complexity is the "isochrone". The isochrone a perimeter, which can be displayed on a map, showing the maximum distance that can be travelled from a point in a specified time (usually by car or ambulance, given typical traffic conditions). Some information systems can generate isochrones; and as an epidemiologist, they have far greater face value as an objective method that a computer can use, than crude distance, when trying to assess the probability of cases being linked. I am not aware that isochrones have been used in any cluster analysis systems; or whether there are better methods available. Competing interests I have no competing interests.