Background and Purpose
In spite of progress on the issue, obesity remains among the most challenging health issues of our time.  Features of the built environment can contribute to obesity by increasing the real or opportunity costs of healthy food choices and physical activity.  The term “obesogenic environment” describes geographic areas that promote obesity across multiple domains—too much fast food, not enough fresh food, and not enough support for physical activity (1).  However, the effect of obesogenic environments on the actual weight status of those exposed to them has not been conclusively established.   Most research has focused on either single objective built environment indicators, subjective ratings of walkability, and/or conventional residential buffers to characterize neighborhood-level risk. In this study, we used multiple objective GIS measures to generate latent profiles of neighborhood risk, with neighborhood defined activity spaces rather than residences.

Objectives
Using a sample of 460 adults in Southern California, we sought to discover a typology of obesogenic risk in the built environment, and determine whether these categories of places predicted physical activity and weight status.

Methods
We used Wave 1 data from 460 adult respondents in two control groups from Healthy PLACES, a natural experiment which has the overarching goal of examining the effects of “smart growth” community design principles on obesity outcomes.  Study participants had at least one school-aged child who also participated in the study.  We used 7 measures of the built environment derived from a GIS: vegetation index, residential/commercial land use mix, fast food restaurants, parks, street connectivity, and traffic accidents involving pedestrians or cyclists.  To define neighborhood of likely exposure, we created stadium-shaped 1-mile buffers around the line connecting adults’ homes and the school at which their child is enrolled.  Since most adults spend a large amount of their time outside the immediate vicinity of home, but within a few miles of it (2), we took these buffers as a proxy for the local geography to which the study participants are likely to be exposed on a regular basis.  We also used self-reported data on age, gender, and educational attainment; anthropometric measures; and physical activity collected using an accelerometer over a 7-day study period.

Analysis proceeded in two phases.  Our first analytic step was to enter the set of 7 continuous area measures into a latent profile analysis.  A latent variable modeling approach can be used to identify unobserved subgroups among a set of continuous characteristics, in this case, characteristics of the built environment that have been linked to obesity.

In the second stage of analysis, we entered the categories of neighborhood identified in the first step as independent variables in a regression model.  Our outcomes in separate models were moderate-to-vigorous physical activity (MVPA) measured by accelerometer, body mass index (BMI), and waist circumference.  Final models were adjusted for age, gender, and educational level, an indicator of socioeconomic status.

Results
Latent profile analysis identified four distinct unobserved environmental profiles.  We expected neighborhoods classified as Profile 2 to be the most obesogenic, with low greenness, high proportion commercial land use, and the highest rate of pedestrian and bike accidents.  Neighborhoods fitting in Profile 3 seemed the least obesogenic, with high greenness, low proportion commercial, low pedestrian and bike accidents, and moderate street connectivity.  Bivariate analysis confirmed that participants in Profile 2 had the highest weight status and lowest MVPA; and those exposed to Profile 3 had the healthiest weight status and highest MVPA.  These differences were significant.

Membership in a Profile 3 neighborhood was significantly associated with lower BMI and nearly 20 minutes per day of additional MVPA, compared to membership in a Profile 2 context.  However, the results did not remain significant after adjusting for gender, age, and education.

Conclusions
Our primary aims were to find latent profiles of built environment obesity risk factors, and test whether these profiles were associated with increased risk for obesity.  Using LPA, we identified four distinct profiles of neighborhood obesogenic risk.  Furthermore, these context types were marginally predictive of obesity and physical activity among the adults who experienced them.  Our results suggest that latent profile analysis can uncover latent clusters of risk factors for obesity in the built environment.  Our results also suggest that the experience of built environment as a factor in obesity is complex and multidimensional.  Further research should focus on the interrelationships between many environmental exposure factors, and the possibility that they interact with one another.

Implications for Practice and Policy
As we work to change the obesity risk environment, we should consider that environments are multifactorial.  Strategic small changes in more than one dimension of the built environment may be able to shift the overall obesogenic profile of an area.  Also, we should be mindful of the fact that people do not spend all of their time at home; they experience a range of places and may self-select environments based on their own priorities.

References

1. Saelens BE, Sallis JF, Frank LD, Couch SC, Zhou C, Colburn T, et al. Obesogenic neighborhood environments, child and parent obesity: the Neighborhood Impact on Kids study. Am J Prev Med 2012;42(5):e57-64.
2. Jones M, Pebley AR. Redefining Neighborhoods Using Common Destinations: Social Characteristics of Activity Spaces and Home Census Tracts Compared. In press, Demography 2013.

Support / Funding Source
This research was supported by the National Cancer Institute (Grant T32CA009492-280).