A test for space clustering and its application to congenital limb defects in Cardiff.

Abstract

Several methods are currently available for investigation of low intensity space-time clustering (Knox, 1963, 1964; Ederer, Myers, and Mantel, 1964; David and Barton, 1966; Mantel, 1967; Pike and Smith, 1968.) Its presence usually indicates one of two causes?either an infection or a causal factor which is local in both time and space and experienced by each individual in the cluster. There are, however, other situations in epidemiology where the effect of either time alone or space alone is localized. An annual cyclic trend is an example of the former; similarly, a condition localized only in space could exist without 'outbreaks' in the time period concerned. Admittedly if another time period in the same area were to be considered the spatial clusters might no longer exist, and in this sense we may still be considering the space-time problem. However, the main interest in this paper is a pattern in which the time component of clusters is so small that it becomes irrelevant to the analysis. The study of spatial patterns of a disease in a given area necessarily involves the population at risk. The problem is trivial if the number of cases is large and the area already divided into well-defined sub-areas whose populations are known. Analysis and subsequent interpretation is difficult when the number of observed cases with the disease is rela tively small, yet on visual inspection there appears to be spatial clustering. Notified limb defects (syn dactyly, Polydactyly, and reduction deformities) born in Cardiff between 1964 and 1966 exhibited these characteristics, and this paper presents the outcome of their analysis, primarily in respect of spatial clustering, but also with brief reference to time-clustering and space-time interaction.