Data pertaining to general practice registrations are collected at the individual level, but for dissemination purposes are often aggregated at postcode level in order to protect confidentiality.

Kernel Density Estimation (KDE) techniques are particularly effective in the analysis of GP catchments areas (CAs) in urban environments because Primary Care Trusts allocate people to their closest surgeries. In the case of Southwark Primary Care Trust, the majority of patients generally live within a one mile radius of their closest surgery, although some longer distances occur when people subsequently move to another part of the Borough.

The Google Map Mashup above shows the distribution of patient registrations relative to the general practices in Southwark. KDE was applied on the postcode dataset using the attribute field containing the number of patient registered to the given GP, a bandwidth of 500 meters and a cell size of 10 meters. If an appropriate bandwidth is chosen, the resulting surface detects the presence of physical features like roads or parks. KDE performs better than classic distance statistics and geographical dispersion measures like the standard deviational ellipse, because it takes in account the direction of distribution.

50%, 75% and 95% Volume Contours were calculated from the KDE surface and it they are shown by the red, blue and black lines. 50% Volume Contour can be interpreted as the “core” CA and can be also linked with other socio-economic variables to obtain a profile of the area surrounding the GP. This is potentially very useful when targeting health campaigns.

If you want to know more about KDE in General Practice Catchment Area Analysis, please read the paper's abstract presented at GISRUK 2007:

GIBIN M., LONGLEY P., ATKINSON P., (2006). Kernel Density Estimation and Percent Volume Contours in General Practice Catchment Area Analysis in Urban Areas. GISRUK 2007, Maynooth - Ireland.

[TOP]