Tests and development of perceived and objective built environment measures for physical activity research

Heather A Starnes, Purdue University


In the first study, a commonly used measure of perceived neighborhood environment walkability was tested for factorial invariance by demographic, behavioral, and environmental factors. A sample of 2,919 Nurses' Health Study participants (mean age=73.0 ± 6.9 years) in California, Massachusetts, and Pennsylvania completed mailed surveys in 2008. In a series of multi-group confirmatory factor analyses, invariance was tested by age, presence of walking limitations, frequency of neighborhood walking, population density within a buffer around home, and state of residence. Several measurement model fit indices were examined, including change in the comparative fit index (Delta CFI), to evaluate four increasingly restrictive hypotheses of configural, metric, scalar, and residual invariance. All levels of invariance by age, walking limitations, and neighborhood walking were supported (Delta CFI less than -.01). However, there was a lack of support for scalar invariance by state and population density (Delta CFI = -.02 and -.05, respectively). The sources of scalar non-invariance were identified in two subscales, infrastructure for walking and access to destinations. Overall, evidence for the invariance of the measure of perceived neighborhood environment walkability provides support for its use in older women of different ages, with different degrees of walking limitations, and frequency of neighborhood walking. However, the finding of partial scalar invariance, suggests that researchers should be cautious when examining certain neighborhood environment subscales among residents living in different states and areas with varied levels of population density. In the second study, novel methods in a geographic information system were examined for assessing daily lifespace and built environment exposure and their relation to daily physical activity. Analyses were conducted with an existing data set of global positioning system (GPS) and accelerometer data collected from 148 adult trail users (mean age 43.9 ± 12.9 years, 72.7% white, 51.4% female) in Massachusetts in 2004-2005. Geographic location and physical activity data were collected simultaneously for a period of one to four days. Daily lifespace (i.e., geographic area individuals circulate within during a given day) of 448 days of monitoring were assessed using three geographic information system (GIS) approaches, convex hull polygon, standard deviational ellipse, and line-based buffer. Comparisons were made between the three GIS approaches in the measures of lifespace area and length, population density, land use mix, and greenness and their associations with demographic characteristics of participants and daily moderate-to-vigorous physical activity (MVPA). Variation in lifespace characteristics was also examined by weekend and weekday. Spatial and built environment characteristics of daily lifespaces were strongly correlated (R = .76 to .99) between the three approaches and also yielded similar patterns of associations with individual demographic factors, weekend days, and physical activity. Male gender, white race and weekend days were associated with lower land use mix, lower population density, and higher greenness in daily lifespaces. Male gender and white race were also associated with longer length of lifespace in kilometers. Land use mix and population density were positively associated with MVPA, and greenness was inversely associated with MVPA. The findings from these two studies provide evidence for the utility of novel statistical and geographic information systems methods to test and develop measures of both the perceived and objective built environment. Major strengths of this study included: 1) invariance tests of a commonly used perceived built environment measure across three states and in older women; 2) development of daily built environment measures that account for the spatial and temporal dynamics of human movement; and 3) examination of relationships between daily built environment exposures and objective daily MVPA. (Abstract shortened by UMI.)




Troped, Purdue University.

Subject Area

Public health

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server