Modeling GPS-based walking activity and its association with objectively measured built environment

摘要

Based on objectively measured data on both walking activity and environment, this studyuses a zero-inflated negative binomial (ZINB) model to examine the relationship betweenthe built environment and walking. Walking activity is measured through a 7-dayassessment of 715 unique study participants living in King County, WA; wearing a GPSand an accelerometer, and filling in a travel diary. Built environment variables aredefined through SmartMaps, which are localized (rasterized) measures of neighborhoodlevelurban form characteristics. The ZINB model allows for the analysis of two separatebut related phenomena: first, differentiating between locations where walking activitydoes or does not occur and examining built environment attributes associated anywalking; and second, estimating the built environment characteristics that are likely toincrease the amount of observed walking activity. Model results showed weakassociations between the built environment and walking. However, whereas any walkingactivity was significantly related to the number of destinations in the neighborhood andstreet intersection density, amounts of walking was associated with neighborhoodresidential density. The study is first to offer insights into the modeling of GPS-basedwalking activity for large populations.