Examining neighbourhood confounding in a survey: an example using the National Health and Nutrition Examination Survey II

This paper demonstrates a methodology for examining whether neighbourhood stratification lessens the confounding bias of some specific risk-factor/disease associations in analysis of the National Health and Nutrition Examination Survey II (NHANES II). The fortuitous clustering of the sample design of NHANES allows us to estimate risk-factor/disease associations with and without controlling for neighbourhood effects on the same population. We briefly discuss the implications of this methodology for neighbourhood matching in case-control studies.