Disease risk near point sources: statistical issues for analyses using individual or spatially aggregated data.
Authors:
- P DigglePDP Diggle
- P ElliottPEP Elliott
Open Access
- 1 December 1995
- journal article
- Published by BMJ in Journal of Epidemiology and Community Health
- Vol. 49 (Suppl 2) , S20-S27
Abstract
STUDY OBJECTIVE--To examine the statistical issues involved in the analysis of disease risk near point sources of environmental pollution, where data are held at both the individual and group (areal) level. To explore these issues with reference to possible socioeconomic confounding. DESIGN--Statistical review. SETTING--Point sources of environmental pollution. MAIN RESULTS--Except in very specific circumstances unlikely to hold in practice, aggregation of data to the areal level will lead to bias in the estimation of disease risk. CONCLUSIONS--There is no easy solution to the analysis of spatial data when some covariates (for example, age and sex of cases) are known at individual level, whereas others (for example, populations, age-sex distributions, small area deprivation indices) are known only at the areal (ecological) level. The underlying assumptions inherent in the analysis of these data need to be explicitly recognised in order to understand better the limitations of the available methodology as well as to inform interpretation of results. Ideally, the data should be kept as disaggregated as possible, to maximise the information available and minimise potential for bias.Keywords
This publication has 6 references indexed in Scilit:
- Childhood leukaemia around the Sellafield nuclear plant1996
- Socio-economic confounding1996
- Approximate Inference in Generalized Linear Mixed ModelsJournal of the American Statistical Association, 1993
- Incidence of cancers of the larynx and lung near incinerators of waste solvents and oils in Great BritainThe Lancet, 1992
- Ecological Bias, Confounding, and Effect ModificationInternational Journal of Epidemiology, 1989
- Empirical Bayes Estimates of Age-Standardized Relative Risks for Use in Disease MappingBiometrics, 1987