Share this article on:

Effect Of Neighborhood-unit Definition On The Relationship Between Physical Activity And The Built Environment: 3667 Board #114 June 3 800 AM - 930 AM

Salvo, Deborah1; Durand, Casey P.2; Evans, Alexandra E.1; Perez, Adriana1; Kohl, Harold W. III FACSM1

Medicine & Science in Sports & Exercise: May 2017 - Volume 49 - Issue 5S - p 1050
doi: 10.1249/01.mss.0000519882.14652.dc
G-33 Free Communication/Poster - Research Methods Saturday, June 3, 2017, 7:30 AM - 11:00 AM Room: Hall F

1The University of Texas Health Science Center at Houston - School of Public Health (Austin), Austin, TX. 2The University of Texas Health Science Center at Houston - School of Public Health (Houston), Houston, TX.

(No relationships reported)

PURPOSE: Substantial evidence demonstrates that built environment features, like density, connectivity, land-use, pedestrian/transit infrastructure, and recreational facilities, can influence physical activity. However, inconsistent findings remain in terms of significance, direction and strength. The purpose of this paper was to determine if the lack of a standardized definition for a neighborhood unit contributes towards these inconsistencies.

METHODS: Published literature (PUBMED & SCOPUS) was abstracted to identify studies examining the relation between physical activity and Geographic Information Systems (GIS)-based built environment measures. Data were abstracted to determine the various definitions of neighborhood units used for GIS built environment measures. Each tested association was coded per the presence or absence of a significant finding. Logistic regression was used to estimate the odds of reporting a significant association (p<0.05) between GIS built environment measures and physical activity outcomes, by neighborhood unit definition. Models adjusted for study sample size.

RESULTS: Among 165 articles (published articles since Jan 2013), 26.8% used Euclidean buffers of varying radii (400-3000m) to define neighborhoods, 28.4% used network buffers, and 44.8% used administrative units of different shapes and sizes (e.g., census tracts). Relative to studies using large administrative units to represent a neighborhood, those using buffers of 400-500m (OR: 3.2, 95% CI: 1.4, 5.8), and 800-1000m (OR: 2.9, CI: 1.3, 7.1), had greater odds of reporting a significant association between GIS built environment measures and physical activity outcomes. Among those using buffers, no significant differences were found between Euclidean vs. network buffers (OR: 1.07, 95% CI: 0.46, 4.29).

CONCLUSIONS: Researchers aiming to accurately estimate the effect of the neighborhood built environment on physical activity should consider using 400-1000m buffer-based GIS indicators. Using network vs. Euclidean buffers may not be essential for characterizing the neighborhood environment for physical activity research. Future analyses should examine differences by physical activity measures (objective vs. subjective) and by built environment constructs.

Supported by NIH R01DK101593

© 2017 American College of Sports Medicine