The timing of geographic power

In many studies on the spatial risk of disease, investigators use geographic locations at the time of disease diagnosis in spatial models to search for individual areas of elevated risk. However, these studies often fail to find a significant spatial signal. This may be due to the misspecification of the timing and location of pertinent exposures. Environmental exposures related to cancer risk vary over space and time, and many cancers have long latencies. When these factors are considered in conjunction with a mobile population, it is likely that the spatial signal related to relevant historic environmental exposures is obscured. To investigate this hypothesis, we conducted simulation studies to characterize the effect of residential mobility on the ability of generalized additive models to detect areas of significantly elevated historic environmental exposure. We generated data based on the residential histories of participants in the National Cancer Institute Surveillance, Epidemiology, and End Results non‐Hodgkin lymphoma study, and varied the duration and intensity of the environmental exposure. Results showed that the probability of detection, mean spatial sensitivity, and mean spatial specificity of models decreased steadily as the time since relevant exposure increased. This suggests that for diseases with long latencies, spatial areas of high risk due to high‐intensity exposure of relatively short duration will be difficult to detect over time when using residential locations at the time of diagnosis in mobile study populations.