Validating Area-Level Behavioral Profiles for Cancer Research

Census-derived socioeconomic measures have wide acceptance among social science researchers both as small-area proxies for individual-level information when direct measurement is not feasible, as well as important constructs in their own right in specifically ecological models investigating the effect of the social environment and contextual resources. Many extensions of demographic data are utilized successfully in commercial areas such as marketing, including "geodemographics," the segmentation and definition of small geographic areas based on both demographics and consumer behavior. However, the utility of commercially prepared area-level behavioral data for health research is largely unknown. The aims of this exploratory investigation are to test the utility of area-level behavioral consumer profile data in explaining geographic disparities in cancer burden, and to refine area-level behavioral data methodologies for future studies.

Specifically, we will present how we 1) used commercially prepared area-level consumer profile data to build a geographic coverage for Maryland on specific cancer-related behaviors and a composite cancer risk profile, 2) tested the construct validity of these measures by statistically modeling their similarity to high-quality individually reported data from the CLUE studies of Washington County residents age 21 and older, and 3) examined the predictive validity of these measures, when added to socioeconomic and healthcare access measures, on excess cancer burden for four major cancer sites (breast, colorectal, prostate, and lung), using cases from the Maryland Cancer Registry. This methodology, if validated, offers a potential tool for moving from highly confidential and reactive methods of behavioral surveillance to less costly, widely applicable methodologies.