Presentation at the 2005 Active Living Research Annual Conference
A large proportion of U.S. adolescents do not meet physical activity guidelines, and interventions have had limited success. Understanding correlates of physical activity could inform improved intervention approaches. Most studies of correlates have examined only psychological and social variables. Environmental changes could affect entire populations, but relatively few studies of environmental correlates of physical activity have been conducted. Studies of adults demonstrate physical activity is consistently associated with the "walkability" of communities (defined by mixed land use, connected streets, higher residential density), but few such studies of youth have been reported. Studies of both adults and youth have documented recreational environment correlates with physical activity, but few studies have included both walkability and recreational environment variables.
Associations between physical activity and environmental indicators of neighborhood walkability and access to recreational facilities were examined in a diverse sample of adolescents.
799 adolescents were recruited from primary care clinics, and the current cross-sectional analyses used baseline data only. The age range was 11-15 years, 53% were female, 43% were non-white, and 45% were considered at risk for overweight or overweight as defined by a body mass index ?85th percentile for age and sex. Physical Activity (PA) was measured by the Actigraph accelerometer, a small monitor worn on the waist. In laboratory and field settings, Actigraphs have been shown to be valid for quantifying children's activity levels. Adolescents wore Actigraphs up to 7 days, and the outcome variable was minutes of moderate plus vigorous physical activity (3+ METs) averaged across valid days of monitoring (MVPA). MVPA was transformed using the box-cox transformation to meet the normality assumptions. Community design variables were adopted from the urban planning and design literature to quantify community environment characteristics related mainly to walking or cycling for transportation. Geographic Information System (GIS) software and parcel-level databases were used to create environmental variables within buffers defined by the street network around each adolescent's home. The variables computed for .5 and 1 mile buffers were land use mix (based on an entropy method), retail floor area ratio (a measure of intensity of use of retail land), intersection density (connectivity), and residential density. A walkability index was created as a sum of z-scores of each component. Recreational environment variables included distance to nearest private recreational facility, public park, trail, or school, as well as the number of these types of facilities within 0.5 and 1 mile of each participant's home. Socio-demographic variables were assessed by self- or parent-report: sex, age, ethnicity, highest parent education. All analyses were stratified by sex. Bivariate analyses were performed using Spearman's correlation. Significant or marginal variables were then entered in multivariate linear regression models. After all independent variables were added, forward selection was used to determine which demographic and socio-demographic variables confounded the relationship between physical activity and environmental variables.
Most of the significant environmental correlations with MVPA were based on the 1 mile buffer, so only those results are presented. For females, significant recreational environment variables were number of private recreational facilities (r=.11, p<.04) and the number of parks (r=.14, p<.007). Age was a significant negative correlate (r=-.47, p<.001) and whites were more active than non-whites (r=.12, p<.03). For females, the only significant environmental variable in the final multiple regression model was number of private recreational facilities (p<.003). For males, retail floor area ratio was significant (r=.12, p<.04), walkability was marginally significant (r=.10, p<.09), and age was a significant correlate (r=-.44, p<.0001) of MVPA. Retail floor area ratio was the only significant environmental variable in the multiple regression (p<.02).
In bivariate analyses, characteristics related to neighborhood design and recreational facilities significantly explained physical activity in female adolescents. When adjusted for age and ethnicity, only private recreational facilities remained significant. Retail floor area ratio was significantly related to adolescent males' MVPA, even adjusting for age, suggesting that developing retail stores more for pedestrians than drivers may stimulate physical activity. Reasons for sex differences in findings are unclear and need to be verified in other studies. Study strengths included objective measurements of multiple environmental variables and physical activity, as well as sex-specific analyses. The study was limited by one geographic location. It might be useful to explore associations in high risk subgroups such as low income and ethnic minority adolescents. Present results provide limited evidence that community design and recreational environmental variables are related to MVPA in adolescents.