|Walkable and Bikable Communities (WBC) Project
Urban Form Lab Health Promotion Research Center University of Washington
SIP-18 supported by the Centers for Disease Control and Prevention
CDC Project Officer: Dr. Thomas Schmid
Principal Investigator: Dr. Anne Vernez Moudon
Principal Research Assistant: Chanam Lee
Co-Principal Investigators: Drs. Allen Cheadle, Cheza Collier,
Donna Johnson, University of Washington,
and Robert Weathers, Seattle Pacific University
Consultants: Dr. Jean-Yves Courbois and Phil Hurvitz, University of Washington
The project is to support the development of audit instruments that gauge aspects of physical environments that facilitate walking and biking. The focus is on identifying environmental factors best predicting walking and biking.
Data for the analysis came from two sources: a telephone survey and GIS databases. A 27-minute survey was administered to 833 adults. The respondents were randomly selected from areas in King County with pre-specified levels of population density and proximity to services that support walking and biking, yet with different levels of development with respect to their non-motorized transportation infrastructure (e.g., presence or absence of a complete sidewalk system, trails, etc.). The survey included self-reported measures of walking, biking and general physical activity, subjective assessments of the physical environment, attitudes toward transportation and the environment, household characteristics, and demographic factors. To the extent possible, the survey used validated questions and a one-week recall period. The response rate was only 31%, however the respondents had characteristics comparable to those of Behavioral Risk Factor Surveillance Survey subjects.
More than two hundred variables measuring environmental factors were generated from transportation network and parcel-level databases with GIS. These featured proximity measures to 24 individual destinations that are likely to attract walking and biking, as well as to groups of destinations in close proximity to each other. In addition, 1- and 3-km airline and network buffer measures were extracted around respondents’ residential locations.
Both survey and environmental variables were first prioritized for their theoretical importance, and then a three-step process was used for data analysis: (1) base model development including only the self reported survey variables, (2) one by one testing of GIS-derived environmental variables, and (3) final model development adding important environmental variables to the base model. These steps were applied for modeling both walking and biking. Multinomial logit models are used for walking, estimating the likelihood of walking moderately, and walking sufficiently relative to not walking at all, and the likelihood of walking sufficiently relative to walking moderately. Biking models use the logistic regression method.
Model results are currently being analyzed. The relative strength of the variables identified in the final model serve to assign weights to the environmental factors included in the audit instruments. Two types of audit instruments will be examined: one for easy application by community groups and lay people; and another for professionals requiring more detailed analysis of environmental data. The approach to instrument design will also help outline a possible certification process for communities that support physical activity, and specifically walking and biking.