Geographic variation in health, at levels ranging from neighborhoods to counties to states, is well documented. Areas might differ in health status because residents differ in characteristics that influence health, such as demographic attributes, socioeconomic status (SES), or health behaviors. Areas also might differ in health status because places differ in characteristics that influence the health of residents, such as local wealth, crime, social support, environmental exposures, or medical services. We are studying the influence of place across the full clinical course of one disease domain, cancer. We employ two distinct, but spatially nested, geographic units: the residential neighborhood of the subject (the “micro residential neighborhood”–approximated, as in most prior work, by the Census tract) and the broader health care service region (the “macro health service area”- approximated by Health Service Areas, or HSAs).  We hypothesize that both spatial levels are relevant to individuals’ health outcomes, potentially affecting different aspects of patients’ illness. This work is organized around the following four specific aims:

* We are developing a new, three-level dataset with demographic and clinical information about elderly cancer patients and detailed contextual information regarding their micro residential neighborhoods and their macro health service areas. Through linkage of various disparate datasets, we are studying approximately 550,000 elderly patients with incident cancers at level I, nested within approximately 8,790 neighborhoods at level II, in turn nested in approximately 304 HSAs at level III.  

* We are evaluating associations between attributes of context (micro residential neighborhood and macro health service area) that may be relevant to cancer outcomes. We describe how, within each level of geography, attributes of areas are related, and how, between levels of geography, attributes are related.  

* We are evaluating the potential simultaneous contributions of two levels of context to the clinical cancer continuum, net of patient-level compositional contributions. We are first evaluating the contribution of the individual level, micro level, and macro level to the total variance in each of the cancer outcomes. Second, we are undertaking hypothesis-driven analyses to explore the specific sources of the variation within each level.

* We are evaluating whether the quality of the hospitals that cancer patients use mediates the effects of place on cancer course. Some of the effects of individual characteristics, micro residential neighborhoods, and macro health service areas on individuals’ cancer outcomes might be mediated through patients’ use of hospitals of better or worse “quality.”  

These multilevel investigations are guided by our overarching hypothesis that where people live –their physical, social, and health care environment– is an important determinant of their experience of cancer, at all stages of the disease, from presentation to treatment to death. Overall, our work is relevant to health policy because it speaks to the impact of health care infrastructure on disease outcomes and because it can contribute to the understanding of racial and poverty-based disparities in health outcomes and health care use.