This project involves assembling a dataset that will allow us, in subsequent work, to examine the impact of place, defined at two different levels of spatial resolution, on cancer course in the elderly. We will build a novel, three-level, hierarchical dataset about an inception cohort of elderly Medicare beneficiaries diagnosed with cancer in 11 US regions during 1998-2002. There will be 550,000 individual patients drawn from the SEER cancer registry at level I, linked by Census tract of residence to information about their neighborhoods (N=8,790) at level II, and linked by their ZIP code to information about their health service areas (HSAs).