Commentary and Review Series 1-92
Correcting for catchment area nonresidency in studies based on tumor-registry data
Sposto R, Preston DLSummary
We discuss the effect of catchment area nonresidency on estimates of cancer incidence from a tumor-registry-based cohort study and demonstrate that a relatively simple correction is possible in the context of Poisson regression analysis if individual residency histories or the probabilities of residency are known. A comparison of a complete data maximum likelihood analysis with several Poisson regression analyses demonstrates the adequacy of the simple correction in a large simulated data set. We compare analyses of stomach-cancer incidence from the Radiation Effects Research Foundation tumor registry with and without the correction. We also discuss some implications of including cases identified only on the basis of death certificates.
We discuss the effect of catchment area nonresidency on estimates of cancer incidence from a tumor-registry-based cohort study and demonstrate that a relatively simple correction is possible in the context of Poisson regression analysis if individual residency histories or the probabilities of residency are known. A comparison of a complete data maximum likelihood analysis with several Poisson regression analyses demonstrates the adequacy of the simple correction in a large simulated data set. We compare analyses of stomach-cancer incidence from the Radiation Effects Research Foundation tumor registry with and without the correction. We also discuss some implications of including cases identified only on the basis of death certificates.