Physical inactivity was recently identified as the fourth leading cause of the death in the United States, increasing risks for more than 20 diseases (29) and responsible for 190,000 deaths annually (9). Adult prevalence estimates of meeting physical activity guidelines (i.e., 30 min of at least moderate-intensity physical activity on most days of the week) range from <5% based on objectively measured physical activity bouts of at least 10 min (38) to <50% based on self-reported leisure physical activities (4). Rather than traditional education or programmatic approaches targeting individuals, many health authorities now recommend the application of multilevel ecological models to identify environmental and policy changes that would increase physical activity of whole populations for long periods (21).
Research to identify most promising environmental attributes to target in interventions has grown rapidly during the past decade. There is consistent evidence that built environment factors are related to physical activity (19,31). Moreover, environmental attributes tend to be specific to physical activity domain (e.g., proximity of destinations is more related to transportation walking, whereas aesthetics appears more related to leisure walking) (31). Major community-wide interventions are using this body of evidence to guide their approaches (1).
However, there are several limitations to and critiques of the research on environmental correlates of physical activity. Ecological models emphasize that behavior is influenced by multiple levels, but most studies focus on only one level rather than seek maximal explanatory value by measuring correlates at multiple levels, for instance, simultaneously examining individual-level factors, such as attitudes about physical activity, and environmental factors such as the presence of walkable destinations. Ideally, studies would focus on most or all the potential levels of influence (25). For example, in physical activity research, this might include examining individual, family, peer and other social, neighborhood environment, and local and state policy factors, although such studies would, by necessity, be very large. The few studies that have included and compared psychosocial and environmental variables generally concluded that psychosocial correlates were much stronger (16). However, lack of environmental variability and lack of comprehensive and objective built environment measures were an acknowledged limitation in some studies. Most studies have also had limited environmental measures and have relied on only or mostly perceived environment variables (26) or included environmental variables not conceptually matched with specific domains of physical activity (e.g., examining total amount of walking, rather than distinguishing between utilitarian and leisure walking [2,17]). In addition, environmental measures have often been aggregated across large areas such as whole counties or cities (10), therefore potentially mis-specifying the environment immediately around an individual’s residence (although see exceptions ). Most studies have used only self-reported physical activity, whereas a preferred approach would be to include both an objective measure of total physical activity and reported measures of domain-specific activities. Because almost all built environment studies are cross-sectional, they have also been criticized for not adjusting for residential self-selection bias (3).
Studies that assess variables from multiple levels of ecological models, including objective measures of environment and behavior whenever possible, and evaluate associations with multiple physical activity outcomes are therefore needed. The Neighborhood Quality of Life Study (NQLS) was designed to address many of the previous gaps and limitations in built environment and physical activity research. Previous NQLS analyses documented that adults living in neighborhoods characterized by relatively high walkability (i.e., based on a composite of objectively measured residential density, street connectivity, land use mix, and retail site design) engaged in more physical activity (measured objectively and by self-report) than adults living in lower walkability neighborhoods, regardless of neighborhood income and after adjusting for individual-level demographics (37). The present study used NQLS data to conduct a more comprehensive analysis that 1) disaggregates “walkability” from a neighborhood level down to examine individual residence-based observation-specific built environment factors, 2) examines whether these more fine-grained factors are related to physical activity, 3) explores an expanded set of physical activity–related environment factors (e.g., proximity to parks), and 4) includes the most consistently supported individual-level psychosocial and neighborhood and individual demographic correlates of physical activity (e.g., social support, self-efficacy) in models along with environmental factors. It was hypothesized that environmental factors are related to adults’ objectively measured overall physical activity, even after controlling for demographic and psychosocial factors. It was further hypothesized that environmental factors such as land use mix and street connectivity are more related than other environmental and perhaps even sociodemographic factors to transportation-related walking. In contrast, perceived aesthetics, safety, and pedestrian amenities are likely more related to leisure walking (31).
Study Design and Neighborhood Selection
Details about the study design and neighborhood selection process in NQLS are provided elsewhere (13,37). In brief, NQLS was a cross-sectional and observational epidemiologic study in which neighborhoods were selected in the King County–Seattle, WA, and Baltimore, MD–Washington, DC, regions of the United States, which differed on built environment factors thought to be related to “walkability” and census-based median household income. Census block groups were first evaluated for residential density, land use mix, intersection density, and retail floor area ratio (FAR), and the z-score values of each (relative to the region) were summed to create a weighted walkability index for each census block group (see Frank et al. ). The seventh–tenth deciles and the first–fourth deciles represented “high” versus “low” walkability areas, respectively; block groups in the fifth and sixth deciles were omitted to create a separation between categories. Block groups also were categorized into “high income” and “low income” based on 2000 Census median household income data. Block groups with median household incomes <$15,000 or >$150,000 were excluded to avoid outliers in neighborhood incomes. The second–fourth deciles constituted the “low-income” category, the seventh–ninth deciles made up the “high-income” category, and the fifth and sixth deciles were omitted to create separation between the categories. Neighborhoods were defined as adjacent block groups that shared the same walkability classification (low or high) and median household income (low or high). Through this process, 32 neighborhoods, made up of 219 census block groups, were selected.
On the basis of marketing company mailing lists, households in these neighborhoods were mailed study introductory letters. Households with known phone (land) lines were contacted by phone soon after the introductory letter was expected to arrive. One adult (20–65 yr old) per household was attempted to be recruited. If the initially targeted adult refused or was ineligible, another adult in the household was invited to participate. Participants residing in a group living establishment (e.g., nursing home, dormitory), unable to complete written surveys in English, and those with a medical condition or disability that interfered with the ability to walk were excluded. There were 2199 participants (1287 in King County, WA, and 912 in the Baltimore–Washington, DC region) who completed an initial survey, of whom 2121 provided accelerometer data. The study participation rate (i.e., returned a survey/eligible contacts) was 26% overall. This study was approved by the institutional review boards at the authors’ institutions, and participants provided written consent for participation. The sampling frame was designed to maximize differences in specific built environment factors for moderately high- and moderately low-income participants. It was not designed to represent the entire distribution of people in each region across built environment or demographic factors.
Procedures and Measures
After returning a signed informed consent, participants were mailed the measurement instruments, including the first survey (see http://sallis.ucsd.edu/measures.html for the complete survey) and the accelerometer ($20 compensation for completing). Participants completed a second survey and wore the accelerometer again approximately 6 months later to control for season effects ($30 compensation). The present analyses include only information from the first survey except for self-reported walking measures and accelerometry measures across both time points. King County participants were recruited and assessed from May 2002 to November 2003. Participants in the Baltimore–Washington, DC, region were recruited and assessed from December 2003 to June 2005.
Age, gender, ethnicity (recategorized as non-Hispanic white or nonwhite), education (five levels from less than high school to graduate degree), number of adults and children in the household, length of time at current address, number of motor vehicles per adults in household, marital status (recategorized as married/living together or other), household income, and job status (yes/no job or unpaid work out of the home) were collected by the survey forms. Demographic data of median resident age, percent nonwhite, and median household income were collected at the census block group level. Demographic characteristics of the sample and the census block groups in which they resided are provided in Table 1.
Participants reported on reasons for selecting their neighborhood (13) and various psychosocial factors related to physical activity (35) on the survey. On the basis of previous factor analysis (12), responses from 3 of 11 residential selection items (labeled “reasons for moving” into one’s neighborhood) were averaged to form an index of the importance of physical activity–relevant characteristics in selecting where to live. Residential selection items had response options from 1 (not at all important) to 5 (very important). Items deemed relevant to physical activity included “closeness to public transportation,” “desire for nearby shops and services,” and “ease of walking” (internal consistency Chronbach α = 0.76). Other residential selection items not included in this index included “affordability/value,” “closeness to open space,” “closeness to job or school,” “sense of community,” “safety from crime,” “quality of schools,” “closeness of recreational facilities,” and “access to freeways.” Average social support for physical activity by family (3 items) and friends (3 items) (33), perceived benefits (10 items) and barriers (15 items) to physical activity (20), self-efficacy for moderate (3 items) and vigorous (3 items) physical activity (36), and enjoyment of moderate (3 items) and vigorous (3 items) physical activity (newly developed by NQLS team) were derived from survey responses (Table 2).
Self-reported neighborhood environment
Four subscales of the Neighborhood Environment Walkability Survey (NEWS) (32) were used to characterize perceived attributes of the neighborhood for which objective measures were not available. This included perceived neighborhood walking/cycling facilities (6 items), aesthetics (6 items), pedestrian/traffic safety (11 items), and safety from crime (4 items). Participants also reported on the proximity of 18 recreation facilities (e.g., park, health club), and these were tallied (34). Higher scores on all the NEWS scales were presumed to indicate more favorable environments for physical activity. Good test–retest reliability and validity have been reported for the NEWS in multiple studies (6,32). Information on item wording, response formats, and scoring can be found at: http://sallis.ucsd.edu/measures.html.
Objective built environment
Data from the county-level tax assessor, regional land use at the parcel level (data from 2002 in King County and 2003 in Baltimore–Washington, DC, region), and street networks (data from 2001 for King County and 2000 in Baltimore–Washington, DC, region) were integrated into a Geographic Information System and used to create participant-specific built environment measures for the 1-km street network buffer around each participant’s residence. Details on the creation of many of the built environment metrics are provided elsewhere (13). Briefly, net residential density was calculated based on the number of residential units relative to the amount of residential land within the buffer. Land use mix examined the evenness of distribution of the square area or floor space dedicated to residential, entertainment (including restaurants), and retail mixed land uses. Intersection density was calculated based on the number of intersections per land area. Retail FAR was calculated as the square footage of retail/commercial relative to the total land area dedicated to retail/commercial within the buffer. This latter measure serves as a proxy for site design and captures the degree to which retail uses are set toward the curb or behind parking. For example, a three-story retail building covering the entire lot has a retail FAR of 3. In contrast, a single-story retail building surrounded by parking has a retail FAR of <1. Because the distribution of retail FAR was highly skewed, categories were created with cut points of 0.33 and 0.67 to ensure sufficient sample per category and for ease of interpretation.
Parcel-level land use data were used to determine the total number of parks within or intersecting the 1-km buffer around each participant. Street network distance to the nearest park was also calculated. Using printed Yellow Pages and Internet-based phone directories, private recreation facilities (e.g., gyms, dance, and martial arts studios) within each county were identified and geocoded. The count of private recreation facilities within 1 km of each participant was calculated. See Table 2 for descriptives for self-reported and objective environment measures.
Participants were instructed to wear accelerometers for 7 d at two measurement time points approximately 6 months apart. Accelerometers (model 7164 or 71256; ActiGraph, Pensacola, FL) were provided to be worn around the waist above the right side of the hip. Accelerometers recorded movement in 1-min epochs. On return, accelerometers were screened to ensure device functioning (e.g., no unexpected repeated or out-of-range values and memory address problems) and for valid hours (no more than 30 consecutive “zero” counts within the hour) on valid days (≥10 h). Five valid days or a minimum of 66 valid hours were considered acceptable during the screening process, and respondents not meeting these thresholds were asked to rewear the accelerometer to attempt to meet these thresholds.
Accelerometer counts per minute were converted into moderate-to-vigorous physical activity (MVPA) minutes using standard adult cut points (15). Average MVPA minutes per valid day were included in the analysis. For data analysis, a valid day contained at least eight valid hours.
Self-reported minute spent in transportation-related walking (e.g., walking to the store) and leisure walking were derived from International Physical Activity Questionnaire long version (http://www.ipaq.ki.se/ipaq.htm (8)). Participants reported the number of days in the past week they walked at least 10 min to get from place to place and the usual minutes per day. Weekly minutes of transportation walking was calculated as days × minutes per day. Separate items queried about leisure walking frequency and duration, and weekly minutes of leisure walking was calculated.
For continuous variables, distributions were examined for normality. When appropriate, a log transformation was used if normality could be better approximated for highly skewed distributions. Frequency distributions were examined for categorical variables. Multicollinearity was assessed using tolerance levels with a criterion of 0.10 or less indicating collinearity. Mixed-effects regression models were fitted to account for the multilevel data structure. All models were adjusted for repeated measures (accelerometry and self-reported walking) over time, site (Seattle, Baltimore region) and season, and subjects nested within census block groups, and census block groups nested within neighborhoods. A hierarchy of initial model building consisted of adding the individual-level demographic variables first, the census-level demographic variables second, and the psychosocial variables third. As each group was added to the model, a backward stepwise regression was carried out to eliminate variables that were not predictive of physical activity at P < 0.05. That is, one variable was removed at a time until all remaining variables were significant. Once the initial model was constructed, the variables that had been removed were reintroduced into the model to determine whether one or more of these variables might be retained or if variables that had been retained might be removed (results from the initial models are not shown, but are available upon request from the first author). Although significance testing was used to develop a hierarchy of importance, some variables were retained even if not significant because they were considered integral to study design.
The next step in the analysis was to examine each environment variable alone adjusted for the retained demographic and psychosocial variables. The final step was to subject the set of environment variables to the same stepwise process described above for the demographic and psychosocial variables. For those remaining in the final models, for ease of interpretation, the regression coefficient for a categorical variable or the regression coefficient multiplied by the variable’s SD for a continuous variable is reported. Because units vary considerably for the continuous variables, using a multiple of the SD provides an interpretation as the change in minutes of MVPA or log of walking for every 1-SD increase in the continuous variable. Consequently, the modified β’s are comparable for continuous variables. However, for categorical variables, the meaning of the unmodified betas still represents an average difference between the comparison and reference levels.
The above analysis steps were conducted separately for accelerometer-measured MVPA, self-reported transportation walking, and self-reported leisure walking. All analyses were carried out using SAS 9.1.3 software (SAS, Cary, NC).
Most of the objectively measured environment factors were related to MVPA after adjusting for individual-level demographic and psychosocial factors and census block group demographic factors. Higher residential density, retail FAR, land use mix, and number of proximal private recreation facilities and parks were significantly related to MVPA, with higher intersection density marginally related to MVPA. None of the perceived environment factors were related to MVPA, and distance to nearest park was also unrelated to MVPA.
In the final MVPA model (Table 3), the individual-level demographic factors of being white/non-Hispanic, having only a high school degree, and working outside the home were related to higher MVPA. In contrast, being older, female, married or living with a partner, and having more motor vehicles per adults in the household were associated with lower MVPA, as was having a higher percentage of nonwhite residents in one’s census block group. Having a lower income and longer length of residence were marginally related to lower MVPA. The significant individual-level psychosocial factors included residential selection for more easy access to places, fewer perceived barriers to physical activity, and greater self-efficacy for moderate and vigorous physical activity. The environmental factor in the final model for MVPA was the objectively measured retail FAR, with higher retail FAR related to higher MVPA (Table 3).
Self-reported transportation walking
The same objectively measured environment factors were related to transportation walking (self-reported) as were related to accelerometer-measured MVPA when each was examined individually after adjusting for demographic and psychosocial factors. The only perceived environment factor significantly related to transportation walking was perceived safety from crime, which was negatively related to transportation walking.
In the final transportation walking model (Table 4), length of residence was inversely related to such walking. Being within the lowest household income category was positively related to transportation walking. Significant positive psychosocial correlates of transportation walking were residential selection for easy access to places, family and friend social support for physical activity, and enjoyment of moderate-intensity physical activity. Objectively measured environment factors were related to more walking for transportation and included higher intersection density, retail FAR, and number of proximal private recreation facilities. None of the perceived environment factors were related to transportation walking in this final model (Table 4).
Self-reported leisure walking
In the initial models, there were few objectively measured environment factors, only higher intersection density and greater retail FAR, related to leisure walking after demographic and psychosocial factor adjustment. Higher perceived neighborhood aesthetics was positively related to leisure walking.
In the final leisure walking model (Table 5), being older, female, and white/non-Hispanic was associated with higher reported leisure walking. Conversely, having children <18 yr old in the home, living longer at one’s current address, and having more adults in the home was related to less leisure walking. Many of the psychosocial factors were related to leisure walking, including residential selection of easy access to places, greater family and friend social support for physical activity, lower barriers to physical activity, and greater self-efficacy for moderate-intensity physical activity. The only environmental factor significantly related to leisure walking was higher perceived neighborhood aesthetics, although objectively measured retail FAR was marginally related to leisure walking (Table 5).
The present findings support the unique importance of specific built environment factors within a 1-km network distance from one’s residence as correlates of objectively measured physical activity and self-reported walking, even after accounting for demographic and psychosocial factors. There were marked differences in the significant environmental correlates based on the type of physical activity. Objectively measured moderate-to-vigorous physical activity was best accounted for among environmental factors by the retail FAR around an individuals’ residence. Retail FAR was also a significant correlate of self-reported transportation walking. Noteworthy was the lack of associations between perceived environment and objectively measured physical activity and transportation walking after controlling for demographic and psychosocial factors. In contrast, perceived neighborhood aesthetics was the only significant environmental correlate of self-reported leisure walking. The pattern of demographic correlates was inconsistent across physical activity type. However, residential selection for easy access to places and social support, self-efficacy for physical activity, enjoyment, and barriers to physical activity were associated with most physical activity metrics.
Each increase in level of retail FAR was associated with significantly higher MVPA. Individuals living in areas with the highest levels of retail FAR category averaged 6.7 min·d−1 more objectively measured MVPA than individuals living in areas with low retail FAR. Retail FAR incorporates elements of mixed use (i.e., residences are close enough to retail locations for walking to be possible), as well as the design or structure of the areas around retail buildings. In the United States, retail FAR tends to be nonexistent in residential-only neighborhoods or low in areas with some mix of residential and retail, in which retail space is surrounded by parking lots or other car-oriented amenities that discourage pedestrian access (e.g., much of the big box retail in the United States). In contrast, high retail FAR is characterized by retail developed in denser clusters where land value has been “bid up” because of the increased proximity to employment and regional transportation services. High retail FAR is often characteristic of urban designs before the dominance of automobile transportation. A critical aspect of retail FAR is its relation to the amount and location of parking. Where retail FAR is higher, parking is typically structured, and space to accommodate the car does not overwhelm safe access on foot to retail and other nonresidential destinations (e.g., access directly from sidewalk into store). Retail FAR has rarely been examined as a correlate of physical activity (18). This is perhaps because additional high-quality data are needed to construct this variable because it relies on knowing both the land use of a parcel as well as the total floor space dedicated to retail within the parcel. Whereas multiple objective built environment factors were related to MVPA when considered individually, retail FAR emerged as the only environmental correlate, objective or subjective, in the final MVPA model.
That retail FAR was also a significant correlate of self-reported walking for transportation is consistent with similar correlates found in prior research (31). Higher intersection density, a metric of street connectivity, was also associated with more transportation walking. Places with high street connectivity have gridlike patterns and short block lengths, allowing individuals to travel more directly between places, which is an important consideration in the choice to walk. In contrast, suburban areas with long block lengths, cul-de-sacs, and smaller streets feeding larger main arterial streets have low street connectivity. High retail FAR is an indicator of nonresidential destinations designed to support pedestrian access. High street connectivity is synergistic and indicates shorter and multiple routes to those destinations.
The final model for leisure walking included perceived aesthetics as the only significant environmental correlate, and this is consistent with prior evidence for recreation or leisure walking (23,31). Walking for leisure or exercise is intentional and often planned, unlike utilitarian walking, which may not be a choice (e.g., no other means of transportation). Particularly for individuals with high psychosocial values (e.g., high perceived benefits) and intentions to be active, favorable neighborhood aesthetics may encourage the decision to walk in the neighborhood, whereas low neighborhood aesthetics may discourage activity in the neighborhood or encourage alternative physical activity. It is logical that neighborhood aesthetics would be correlated with leisure walking and less with utilitarian walking because of the discretionary nature of this form of physical activity, although some have suggested that aesthetics is an important factor to consider in disadvantaged neighborhoods with regard to walking for transport (5). Leisure activities are more selectively determined; therefore, people are more likely to choose this activity over other activities when they find their community to be a pleasing place to experience on foot. When specifically examining leisure or exercise walking, most studies, like the present one, have failed to find objectively measured neighborhood environment correlates (27). The lack of associations of other perceived environment factors, including crime and traffic safety and pedestrian infrastructure, with leisure walking was unexpected. Perhaps inclusion of psychosocial variables accounted for the variance sometimes explained by these environmental variables, and prior evidence on the link between crime safety and physical activity is inconsistent (11).
In this study, park proximity metrics were unrelated to overall physical activity and walking. Prior evidence for the amount of and proximity to park and public open space being a contributor to overall physical activity and walking was mixed (22). Even in unadjusted models, distance to the nearest park was not significantly related to any activity measures. The number of nearby parks was related to MVPA and transportation walking, but not leisure walking in unadjusted models. Park proximity may not be the most appropriate metric of how parks influence physical activity, and more work is needed to identify which aspects of parks are influential (e.g., quality, amenities).
The residential self-selection factor, with higher scores indicating more interest in having easy access to places (e.g., near public transit, ease of walking), was related to overall physical activity and both walking domains. This result is consistent with other studies, with many also finding that environmental factors are related to physical activity even after adjustment for self-selection (3,28). Evaluations of policy and environmental change have the potential to better disentangle the effect of residential preference and environmental influence. The length of residence was also consistently negatively related to overall MVPA and both types of walking. Length of residence was highly associated with age (r = 0.56) and the number of available motorized vehicles (r = 0.24) in this sample; however, both of these other factors were entered in all models. Age and number of motor vehicles remained significant in the MVPA model (negatively), and age remained in the leisure walking model (positive). Length of residence may be a good proxy for home rental versus ownership, a factor not entered into the models, with respective lengths of residence of 49 versus 139 months in this sample. Home owners, particularly house owners, likely live in less residentially dense areas (although residential density was also included in the models), thus reducing the likelihood of overall MVPA and transportation-related walking. Home owners are also likely to have permanent and nearby storage for motorized vehicles (e.g., garages, car ports) and thus could be more likely to use them for transportation. There could also be an effect of disposal income and/or an interaction between income and motor vehicle availability that is being captured by length of residence. For example, more transient renters are individuals with lower incomes and no access to their own motor vehicle and thus more dependent on walking and/or taking public transportation—the latter often requiring walk trips to/from transit stops. Thus, the observed length of residence findings could be due to residual confounding effects of age, vehicle availability, home ownership, or other variables. Clear interpretations are difficult because of the interrelations among variables, requiring different research strategies to understand their independent influence.
In unadjusted models, most of the objectively measured built environment factors, including residential density, were related to accelerometer-derived MVPA and to transportation walking. Although some have found that residential density is related to transportation walking in particular (7), it could be that higher levels of residential density than those studied in NQLS (although rare in the United States relative to some European and Asian cities) are needed to demonstrate stronger associations with walking and total physical activity independent of other environmental factors. This was in contrast to the unadjusted model for leisure walking, in which most of the environmental factors were not significant correlates, with the exception of intersection density and retail FAR. Many previous studies have found that density becomes nonsignificant after other features of the built environment are introduced into models (14). Density is needed to create markets for retail, parks, and other destinations, and it predicts transit usage.
Although varying somewhat across physical activity domains, psychosocial factors were common correlates across activity types. Self-efficacy, or confidence in being active, and lower perceived barriers to physical activity were related to MVPA and leisure walking. Social support from family and friends for physical activity, although relatively low on average, was positively correlated with transportation and leisure walking. The demographic correlates of MVPA were similar to those found previously, with older, female, and nonwhite adults having less MVPA (35), although nonwhite adults in the 2003–2004 National Health and Nutrition Examination Survey had higher objectively measured physical activity than whites (38). In the present study, the positive relationship with household income was also expected based on prior research (16), as was the finding that individuals in households with more motor vehicles had lower MVPA. The relatively higher amounts of MVPA among individuals receiving only a high school education was unexpected, although less well-educated NQLS participants could have been involved in more work-related physical activity compared to more well-educated participants. The demographic and psychosocial findings in the present study highlight the continued importance of evaluating these variables when testing ecological models of health behaviors.
An important finding was that an environmental variable, namely, retail FAR, was a stronger correlate than any psychosocial variable for all three outcomes. The estimated effect size for high retail FAR was 6.7 min of MVPA compared with the highest effect size of 4.6 min among psychosocial variables (barriers), with other psychosocial variables having effect sizes <2 min. The strong association between retail FAR in relation to self-reported walking for transport is consistent with one study that included retail FAR in a larger neighborhood walkability index (12). However, the present finding seems to be in contrast with another study that included both psychosocial and built environment variables (16) that reported weak environmental correlates, although retail FAR was not among the examined built environment factors. A marked difference is likely based on the design of NQLS, in which regions and neighborhoods were selected to represent wide differences in objective built environment factors. In addition, few prior studies measured retail FAR, and few, if any, had floor space metrics in their parcel data. The built environment is a three-dimensional phenomena; therefore, a 65-story building may have more than a million square feet of development and attract many pedestrians but sit on 1 acre of land. Relying only on land area and other two-dimensional factors can mis-specify built environment and its three-dimensional nature, missing potent correlates of walking for transportation.
Previous studies likely had restricted range in environmental variables, so they could have underestimated associations. Retail FAR, our most potent and consistent correlate of physical activity reflects both destinations within walking distance and pedestrian-oriented design. Other environmental variables that were significant in models (i.e., aesthetics, intersection density, private recreation facilities) had similar effect sizes as psychosocial variables. The present results suggest that both environmental and psychosocial factors may be powerful influences on physical activity and support the principle of ecological models that important correlates, and possibly influences, on behavior can be found at every level.
Study strengths included objective measures of physical activity (at two time points approximately 6 months apart) and built environment, the examination of environmental characteristics within a kilometer of participants’ residences, and the inclusion of individual-level demographic and psychosocial factors known to be related to physical activity. Limitations included most critically the cross-sectional design, the inability to randomly assign participants to environments and therefore an inability to completely eliminate the possibility of confounding by residential self-selection. There is also considerable potential for sample bias, particularly given the response rate, which, likely because of the required accelerometer wearing, was lower than for survey-only studies. Participants were volunteers, from households with landline phones, English speaking, and able and willing to complete surveys and accelerometer wearing. The sample was generally well-educated (65.1% having at least a college degree), although these regions of the United States are known to have more well-educated populations. The sample was somewhat more affluent relative to the block groups in which they lived (only ∼43% were below median income for their block group), perhaps as a result of being older (sample mean was 45.2 yr, but the median age of the block groups was 35.9 yr). The current study did not include more fine-grained built environment features, such as the presence and quality of sidewalks. The study also did not capture the level of service of transit, an important predictor of walking and achieving recommended levels of physical activity (24). The retail FAR measure was likely picking up some of the variation in regional location that would otherwise be captured in measures of transportation accessibility such as transit level of service.
In summary, neighborhood built environmental factors, even after adjusting for demographic and psychosocial factors, were correlates of transportation walking and total MVPA. The strongest and most consistent environmental factors, retail FAR is modifiable, controlled, and prescribed by locally enforced land use regulations. The present findings provide an empirical rationale for policies that support developments with many features associated with higher retail FAR, including retail destinations near residential areas, less surface and more structured parking, buildings at the curb, and building vertically as opposed to horizontally to achieve a critical mass of destinations and activities. Taken collectively, these factors may be a powerful intervention to promote physical activity.
The Neighborhood Quality of Life Study was supported by the National Heart, Lung, and Blood Institute of the US National Institutes of Health (HL67350).
The authors do not have any conflicts of interest or professional relationships with companies or manufacturers or any other interests that would benefit from the results of this study. Authors Sallis, Cain, and Conway are currently affiliated with the University of California at San Diego, San Diego, CA.
The results of this study are solely the responsibility of the authors and do not constitute endorsement by the American College of Sports Medicine.
1. Bors P, Dessauer M, Bell R, et al.. The Active Living by Design national program: community
initiatives and lessons learned. Am J Prev Med. 2009; 37 (6 suppl 2): S313–21.
2. Burton NW, Turrell G, Oldenburg B, Sallis JF. The relative contributions of psychological, social, and environmental variables to explain participation in walking
, moderate-, and vigorous-intensity leisure-time physical activity. J Phys Act Health. 2005; 2: 181–96.
3. Cao X, Handy SL, Mokhtarian PL. The influences of built environment and residential self-selection on pedestrian behavior: evidence from Austin, TX. Transportation. 2006; 33: 1–20.
4. Carlson SA, Densmore D, Fulton JE, Yore MM, Kohl HW 3rd. Differences in physical activity prevalence and trends from 3 U.S. surveillance systems: NHIS, NHANES, and BRFSS. J Phys Act Health. 2009; 6 (suppl 1): S18–27.
5. Cerin E, Leslie E, Owen N. Explaining socio-economic status differences in walking
for transport: an ecological analysis of individual, social and environmental factors. Soc Sci Med. 2009; 68 (6): 1013–20.
6. Cerin E, Saelens BE, Sallis JF, Frank LD. Neighborhood Environment Walkability
Scale: validity and development of a short form. Med Sci Sports Exerc. 2006; 38 (9): 1682–91.
7. Coogan PF, White LF, Adler TJ, et al.. Prospective study of urban form and physical activity in the Black Women’s Health Study. Am J Epidemiol. 2009; 170 (9): 1105–17.
8. Craig CL, Marshall AL, Sjostrom M, et al.. International Physical Activity Questionnaire: 12-country reliability and validity. Med Sci Sports Exerc. 2003; 35: 1381–95.
9. Danaei G, Ding EL, Mozaffarian D, et al.. The preventable causes of death in the United States: comparative risk assessment of dietary, lifestyle, and metabolic risk factors. PLoS Med. 2009; 6 (4): e1000058.
10. Ewing R, Schmid T, Killingsworth RE, Zlot A, Raudenbush S. Relationship between urban sprawl and physical activity, obesity, and morbidity. Am J Health Promot. 2003; 18 (1): 47–57.
11. Foster S, Giles-Corti B. The built environment, neighborhood crime and constrained physical activity: an exploration of inconsistent findings. Prev Med. 2008; 47 (3): 241–51.
12. Frank LD, Saelens BE, Powell KE, Chapman JE. Stepping towards causation: do built environments or neighborhood and travel preferences explain physical activity, driving, and obesity? Soc Sci Med. 2007; 65 (9): 1898–914.
13. Frank LD, Sallis JF, Saelens BE, et al.. The development of a walkability
index: application to the Neighborhood Quality of Life Study. Br J Sports Med. 2010; 44: 924–33.
14. Frank LD, Stone BJ, Bachman W. Linking land use with household vehicle emissions in the Central Puget Sound: methodological framework and findings. Transp Res Part D: Transport Environ. 2000; 5 (3): 173–96.
15. Freedson PS, Melanson E, Sirard J. Calibration of the Computer Science and Applications, Inc. accelerometer. Med Sci Sports Exerc. 1998; 30 (5): 777–81.
16. Giles-Corti B, Donovan RJ. The relative influence of individual, social and physical environment determinants of physical activity. Soc Sci Med. 2002; 54: 1793–812.
17. Giles-Corti B, Timperio A, Bull F, Pikora T. Understanding physical activity environmental correlates: increased specificity for ecological models. Exerc Sports Sci Rev. 2005; 33 (4): 175–81.
18. Handy S. Critical assessment of the literature on the relationships among transportation, land use, and physical activity. In: Transportation Research Board and Institute of Medicine Committee on Physical Activity Health Transportation and Land Use, editor. TRB Special Report 282: Does the Built Environment Influence Physical Activity? Examining the Evidence. Washington (DC): The National Academies Press; 2005. p. 102.
19. Heath GW, Brownson RC, Kruger J, et al.. The effectiveness of urban design and land use and transport policies and practices to increase physical activity: a systematic review. J Phys Act Health. 2006; 3 (suppl 1): S55–76.
20. Hovell MF, Sallis JF, Hofstetter CR, et al.. Identifying correlates of walking
for exercise: an epidemiologic prerequisite for physical activity promotion. Prev Med. 1989; 18: 856–66.
21. IOM (Institute of Medicine) and National Research Council. Local Government Actions to Prevent Childhood Obesity. Washington (DC): The National Academies Press; 2009. p. 139.
22. Kaczynski AT, Potwarka LR, Saelens BE. Association of park size, distance, and features with physical activity in neighborhood parks. Am J Public Health. 2008; 98 (8): 1451–6.
23. Kondo K, Lee JS, Kawakubo K, et al.. Association between daily physical activity and neighborhood environments. Environ Health Prev Med. 2009; 14 (3): 196–206.
24. Lachapelle U, Frank LD. Transit and health: mode of transport, employer-sponsored public transit pass programs, and physical activity. J Public Health Pol. 2009; 30 (suppl 1): S73–94.
25. Li F, Fisher KJ, Brownson RC, Bosworth M. Multilevel modelling of built environment characteristics related to neighbourhood walking
activity in older adults. J Epidemiol Community
Health. 2005; 59 (7): 558–64.
26. Lin L, Moudon AV. Objective versus subjective measures of the built environment, which are most effective in capturing associations with walking
? Health Place. 2010; 16 (2): 339–48.
27. Lovasi GS, Moudon AV, Pearson AL, et al.. Using built environment characteristics to predict walking
for exercise. Int J Health Geogr. 2008; 7: 10.
28. Owen N, Cerin E, Leslie E, et al.. Neighborhood walkability
and the walking
behavior of Australian adults. Am J Prev Med. 2007; 33 (5): 387–95.
29. Physical Activity Guidelines Advisory Committee. Physical Activity Guidelines Advisory Committee Report, 2008. Washington (DC): US Department of Health and Human Services; 2008. p. 683.
30. Rodriguez DA, Evenson KR, Diez Roux AV, Brines SJ. Land use, residential density, and walking
. The multi-ethnic study of atherosclerosis. Am J Prev Med. 2009; 37 (5): 397–404.
31. Saelens BE, Handy SL. Built environment correlates of walking
: a review. Med Sci Sports Exerc. 2008; 40 (suppl 7): S550–66.
32. Saelens BE, Sallis JF, Black JB, Chen D. Neighborhood-based differences in physical activity: an environment scale evaluation. Am J Public Health. 2003; 93 (9): 1552–8.
33. Sallis JF, Grossman RM, Pinski RB, Patterson TL, Nader PR. The development of scales to measure social support for diet and exercise behaviors. Prev Med. 1987; 16 (6): 825–36.
34. Sallis JF, Johnson MF, Calfas KJ, Caparosa S, Nichols JF. Assessing perceived physical environmental variables that may influence physical activity. Res Q Exerc Sport. 1997; 68 (4): 345–51.
35. Sallis JF, Owen N. Physical Activity and Behavioral Medicine. Thousand Oaks (CA): Sage; 1999. p. 210.
36. Sallis JF, Pinski RB, Grossman RM, Patterson TL, Nader PR. The develoment of self-efficacy scales for health-related diet and exercise behaviors. Health Educ Res. 1988; 3: 283–92.
37. Sallis JF, Saelens BE, Frank LD, et al.. Neighborhood built environment and income: examining multiple health outcomes. Soc Sci Med. 2009; 68 (7): 1285–93.
38. Troiano RP, Berrigan D, Dodd KW, et al.. Physical activity in the United States measured by accelerometer. Med Sci Sports Exerc. 2008; 40 (1): 181–8.