|
Key words: Clusters, Spatial Pattern, Epidemiology, Testing
The detection of spatial patterns (clusters) in disease databases has been of interest to epidemiologists for many years. Over time, many methods for detecting clusters have been developed and tested (e.g., Cusick and Edwards, 1990). However, many epidemiologists remain unconvinced by these demonstrations and continue to use home grown and often geographically naive methods. For example, the EUROHAZCON study (Dolk et al., 1998) investigates the rates of congenital birth abnormalities around landfill sites by considering circles centred on the landfill sites and extending 7 km with a break at 3 km. No justification for the 3 km or 7 km distances chosen is given by the authors, nor is there discussion of the how robust the results are with respect to changes in these distances. Elliott et al. (1992) in similarly studies of cancer near to waste incinerators in the UK report a similar technique of fixed radii around the points of interest. A final recent example is Cousens et al. (1999) who consider the geographical distribution of new variant CJD by the use of fixed circles centred on rendering plants of up to 50 km. Several methods have, however, been proposed to take account of the geography of the problem (see Alexander and Boyle (1996) for a good review). The paper will investigate by the use of synthetic data sets the risks inherent in assuming that clusters must be centred on a site (or sites) that the researcher feels is (are) responsible for the disease without considering the whole spatial extent of the problem. The data sets will be investigated using a series of fixed radii circles around "suspected" locations to investigate the spatial robustness of the methods commonly used by practising epidemiologists and compare the results to those found using focused whole map tests such as GAM (Openshaw, 1996; Openshaw, Turton and Macgill, 1999) and the Besag and Newell (1991) method. The results will be used to develop recommendations as to best practice in cancer cluster detection and will include links to a web site allowing users to test the methods online.
References
Alexander, F.E. and Boyle, P. (Eds), 1996, Methods for Investigating Localised Clustering of Disease. Lyon, France, IARC Scientific Publication No. 135.
Besag, J. and Newell, J., 1991, The detection of clusters in rare diseases. Journal of the Royal Statistical Society, Series A 154, pp. 143-155.
Cousens, S., Linsell, L., Smith, P., Chandrakumar, M., Wilesmith, J., Knight, R., Zeidler, M., Stewart, G. and Will, R., 1999, Geographic distribution of variant CJD in the UK (excluding Northern Ireland). Lancet 353. pp. P18-21.
Cusick, J. and Edwards, R., 1990, Tests for spatial clustering in heterogeneous populations. Journal of the Royal Statistical Society, Series B 52, pp. 73-104.
Dolk, H., Vrijheid, M., Armstrong, B., Abramsky, L., Bianchi, F., Garne, E., Nelen, V., Robert, E., Scott, J.E.S., Stone, D. and Tenconi, R., 1998, Risk of congenital anomalies near hazardous-waste landfill sites in Europe: the EUROHAZCON study. Lancet 352, pp. 423-427.
Elliott, P., Hills, M., Beresford, J., Kleinschmidt, I., Jolley, D., Pattenden, S., Rodrigues, L., Westlake, A. and Rose, G., 1992, Incidence of cancer of the larynx and lung near incinerators of waste solvents and oils in Great Britain. Lancet 339, pp 854-858.
Openshaw, S., 1996, Using a geographical analysis machine to detect the presence of spatial clusters and the location of clusters in synthetic data. In Alexander, F.E. and Boyle, P. (Eds), Methods for Investigating Localised Clustering of Disease. Lyon, France, IARC Scientific Publication No. 135, pp. 68-87.
Openshaw, S., Turton, I., and Macgill, J., 1999, Using the Geographical Analysis Machine to analyse census limiting long term illness. Geographical & Environmental Modelling 3.1, pp. 83-99.