Secondary GIS built environment data for health research: guidance for data development.

Built environment (BE) data in geographic information system (GIS) format are increasingly available from public agencies and private providers. These data can provide objective, low-cost BE data over large regions and are often used in public health research and surveillance. Yet challenges exist in repurposing GIS data for health research. The GIS data do not always capture desired constructs; the data can be of varying quality and completeness; and the data definitions, structures, and spatial representations are often inconsistent across sources. Using the Small Town Walkability study as an illustration, we describe (a) the range of BE characteristics measurable in a GIS that may be associated with active living, (b) the availability of these data across nine U.S. small towns, (c) inconsistencies in the GIS BE data that were available, and (d) strategies for developing accurate, complete, and consistent GIS BE data appropriate for research. Based on a conceptual framework and existing literature, objectively measurable characteristics of the BE potentially related to active living were classified under nine domains: generalized land uses, morphology, density, destinations, transportation system, traffic conditions, neighborhood behavioral conditions, economic environment, and regional location. At least some secondary GIS data were available across all nine towns for seven of the nine BE domains. Data representing high-resolution or behavioral aspects of the BE were often not available. Available GIS BE data - especially tax parcel data - often contained varying attributes and levels of detail across sources. When GIS BE data were available from multiple sources, the accuracy, completeness, and consistency of the data could be reasonable ensured for use in research. But this required careful attention to the definition and spatial representation of the BE characteristic of interest. Manipulation of the secondary source data was often required, which was facilitated through protocols.