zbMATH — the first resource for mathematics

Case-control analysis of risk around putative sources. (English) Zbl 0964.62100
Lawson, Andrew (ed.) et al., Disease mapping and risk assessment for public health. Chichester: Wiley. 271-286 (1999).
From the introduction: Many epidemiological investigations on environmental hazards focus on a priori putative sources (see, for example, the studies on childhood leukaemia and nuclear reprocessing plant sites). The pattern of risk as a function of distance from the source is used as an approximation for the true distribution of risk by exposure levels and therefore such study design has a poor value as a proof of a direct causal relationship. This kind of geographical analysis may be used either to describe and evaluate the magnitude of the problem, or to surrogate the unknown exposures by distance measures.
To estimate this risk pattern aggregate data can be used, but the results could be affected by the ecological bias since individual risk factors are not taken into account. Theoretically the successful modelling of the risk pattern as a function of distance from putative sources depends on the availability of individual data and on information on major risk factors and predictors of the disease under study. Individual information is usually of high quality but at greater cost than routinely collected aggregate data, and in epidemiological research special sampling designs are used to ensure an acceptable cost-effectiveness balance: case-control studies are one example of a highly efficient sampling strategy. In geographical analyses, the sample of controls is used in any situation where the distribution of the population by small areas is unknown or when it is too difficult or expensive to gather.
The present chapter focuses on the analysis of individual data generated by a case-control sampling design aimed to assess the risk gradient as a function of the distance from a putative source. The data are an example of a heterogeneous Poisson point process in the plane, in which the intensity of cases in a given location depends on the number of people at risk, on the distance from source and on individual covariates.
For the entire collection see [Zbl 0942.00010].
62P10 Applications of statistics to biology and medical sciences; meta analysis