Background
Population disparities in physical activity and health are well documented, and promoting physical activity among the most vulnerable segments of the population is regarded as one of the top public health agendas. However, existing empirical studies provide limited insights into how similarly or differently built and social environmental conditions are associated with physical activity for different sub-populations. Further, the strength and nature of this relationship change based on the purposes or types of physical activity being considered.

Objectives
This paper presents and tests two theoretical frameworks, the Active Living for Transportation (ALT) and Active Living for Recreation (ALR), conceptualizing correlates of ALT and ALR for particular sub-populations. They are derived from the social ecological model and the Behavioral Model of Environment (BME).

The focus is to examine the associations that the built and social environments of the neighborhood have with physical activity for two of the most vulnerable sub-populations: high health risk and low-income populations. This paper points to how (in)equitably environmental resources for active living are distributed in areas with different sub-populations live, and to the need for future policy and design interventions to be more targeted and tailored to the sub-population specific needs.

Methods
Data: Physical activity and socio-demographic background data came from a survey of 438 randomly sampled adults living in Seattle. Environmental variables were measured both subjectively as perceived social and physical neighborhood environments and objectively in GIS, using detailed and disaggregated GIS measures taken around each survey respondent’s home location.

Variables: Latent factors and observed variables were used. Latent factors corresponded to the constructs of theoretical frameworks, including transportation and recreation physical activity, health risk, economic challenge, social environment, and physical environment-transportation, recreation and area. They were captured by three to six observed variables identified from factor analysis and previous studies. The health risk factor, for example, was defined by BMI, perceived health status and age. The origin/destination and route constructs of the BME were re-specified into transportation and recreation factors, while the area construct remained as it is.

Analyses: Descriptive statistics and bivariate analyses focused on the observed variables, examining the associations between the sub-population variables (health risk and economic challenge) and physical activity as well as the social and physical environments. Structural Equation Model (SEM) was selected to test the validity of the overall theoretical frameworks, the ALT and ALR, due to its ability to efficiently combine theory-based with analytical approaches to test hypotheses.

Results
Bivariate Analysis: Lower income populations lived in areas with more routine destinations such as restaurants and grocery stores as well as higher densities. They were more active for transportation but less active for recreation than higher income groups. Those with higher health risks were less active for recreation and perceived their neighborhood to be less socially supportive. Moderate and vigorous physical activity was associated with reduced health risks. Those who had activity limitations reported lower levels of recreational activities. The physical environment had a strong association with transportation physical activity. Even the physical environment–recreation variables were more strongly related with transportation than with recreation activities.

SEM: The SEM results were consistent with the bivariate analysis results. Both ALT and ALR models were valid and useful in framing environment-physical activity research for sub-populations, demonstrated by acceptable levels (>0.9) of goodness of fit values. The SEMs showed no significant associations between physical environments and health risk. However, social environment was significantly associated with recreation activity that people with high health risks do. Health risk was negatively related with transportation and recreation physical activity. Among the three physical environment latent factors, only the area factor (captured by density, housing type, and traffic volume) was related with transportation physical activity. None was associated with recreation physical activity.

Conclusions
This study showed that social and built environmental correlates of physical activity varied by different sub-populations. Explicit distinctions between different purposes and types of physical activity, and between different socio-economic backgrounds of the study populations are essential in future research. Future policy and design interventions must be tailored toward the specific needs of the target populations.

This study relied on cross-sectional data and was conducted in urban areas. Future research needs to include longitudinal studies, various environmental settings and diverse at-risk sub-populations, to further understand the mechanism through which different social and physical environmental factors interact with health and economic conditions.