Assessing bias associated with geocoding of historical residence in epidemiology research

Abstract

The use of geocoded historical residence as proxy for retrospective assessment of exposure in early life is increasing in epidemiological studies of chronic health outcomes. Dealing with historical residence poses challenges, primarily due to higher uncertainties associated with data collection and processing. A possible source of bias is connected with the exclusion of subjects, who cannot, for various reasons, be geocoded. We evaluated the potential bias that may arise due to incomplete geocoding, using birth residence data collected as part of a population-based case-control study of breast cancer in western New York state. We found that geocoded and non-geocoded populations did not differ in the distribution of most risk factors compared, and that the geocoding status did not modify the spatial patterns of the study populations. However, the results emphasize the need for epidemiological studies to consider the potential biases that may be introduced by geocoding of historical residence when investigating retrospectively chronic disease and early-life exposure.