A geostatistical approach to the analysis of pattern in rare disease.

The incidences of human disease vary from place to place, and some show distinct patterns. Patterns in rare diseases, such as childhood cancer, are not readily discernible, however, and this makes it difficult to relate their distribution to factors in the environment in seeking possible causes of the disease. Provided cases are geographically indexed, data on a disease can be analysed geostatistically. Variograms can be computed to determine the strength and spatial scale of any pattern, and to summarize the variation, and then the risk of developing the disease can be estimated by kriging. This has been done for the incidence rates of childhood cancer from 1980 to 1984 in the West Midlands Health Authority Region of England. A novel modification was made to compute the variogram of the risk from that of the frequencies, taking into account the binomial nature of the data. The incidence of the disease appears patchy. The results show that this patchiness is spatially correlated and not purely random. The risk appears greatest in the rural south west of the Region and in some of the suburbs around the conurbation. This approach using geostatistics seems very promising and will be developed further as more data for this and other diseases become available.