The main finding of the present study was that for 13-15 year old Belgian adolescent boys and girls the average activity level and the mean minutes moderate to vigorous physical activity per day were associated with neighborhood walkability, but this association was moderated by neighborhood SES. Only in low-SES neighborhoods, adolescent boys and girls living in high-walkable neighborhoods performed more moderate to vigorous physical activity (+ 7.4 min/day) and achieved a higher average activity level (+ 39.9 counts/min) than adolescent boys and girls living in low-walkable neighborhoods. Among adolescents living in high-SES neighborhoods, no association was found between neighborhood walkability and accelerometer-based physical activity. For the self-reported variables-active transport to and from school, walking and cycling for transport during leisure time and sport during leisure time-no association with neighborhood walkability was found. These results are an indication that the main conclusion of the PLACE, NQLS, BEPAS and SNAP studies, that adults living in high-walkable neighborhoods reach higher levels of physical activity, cannot be extended to the overall group of 13-15 year old Belgian adolescents.
The contrasting results of the BEPAS-Y and the BEPAS adult [32
] studies, conducted in the same neighbourhoods, are in particular noteworthy. The BEPAS adult study documents large differences in accelerometer-based moderate to vigorous physical activity across walkability groups, in both low- and high-SES neighborhoods. Given the similarity of environments and measures, questions can be raised about the different responses of adolescents and adults to their built environment.
Neighbourhood "walkability" refers to the ability to walk or cycle to nearby destinations. The key elements of "walkability" are determined based on adult research investigating the specific neighborhood environmental attributes that are associated with higher levels of walking and cycling for transport [38
]. Walkability has been consistently found to be associated with walking for transport on three continents so far [30
], as well as with cycling for transport [32
]. In comparison with adult physical activity behavior, adolescent physical activity behavior is characterized by a larger variation in types of physical activity. Active transport forms only a fraction of adolescents' total physical activity behavior. Given this, the possibility exists that the construct "walkability" is not as relevant for adolescents. In the current evidence-base, the specific built environmental attributes that characterize a walkable neighborhood, are less consistently associated with adolescents' physical activity [29
]. In some adolescent studies connectivity even showed an inverse association with physical activity [29
]. Including connectivity in the walkability-index for adolescents could therefore negate some of the effects.
Considering the multidimensional character of neighborhood environments, it could be that among youth, other built environment attributes are the key elements of neighborhoods conducive to an active lifestyle. It could also be that family rules could interact with built environment and social environment features in influencing adolescent physical activity. Thus, further studies are needed to determine the specific environmental attributes characterizing the ability for youth to be active in a neighborhood. Parallel to research in adults, those attributes can be used to create an index that refers to the ability of youth to be active in their neighborhood, e.g. an "activability" index.
In the development of an "activability" index for youth it may be advisable to take into account the diverse physical activity domains adolescents are participating in. As stated by Giles-Corti et al. [55
] environmental attributes are expected to have effects that vary by physical activity domain. In particular, walkability is expected to be related to active transportation, and a recent review confirms that proximity to parks and recreation facilities was one of the most consistent correlates of adolescent physical activity [29
]. The present study did not include measures of recreation environments expected to be related to leisure time physical activity. Thus, an "activability" index likely needs to include recreation environment attributes as well. Due to large variations in built environments across countries and continents, an "activability" index may need to be tailored to each context. Research to identify built environmental attributes that have beneficial effects on physical activity in adults and youth is of utmost importance to inform policy makers and urban planners in making well-considered decisions concerning built environmental redevelopments of existing neighborhoods and planning of new neighborhoods.
Surprisingly, previous studies that found an interaction between neighborhood walkability and neighborhood SES showed a stronger association between neighborhood walkability and physical activity in high-income neighborhoods [31
]. Part of the explanation for the interaction found in the present study may be attributed to economic factors. As stated by Stalsberg et al. [56
] the observed SES moderating effect could be explained by varying ability to deal with the financial outlay that certain activities require (e.g. sport material, membership fees). Most adolescents living in high-SES neighborhoods have the ability to participate in activities that require financial outlay, in contrast to adolescents living in low-SES neighborhoods. Consequently, adolescents from high-SES neighborhoods are less dependent on their neighborhood environment to be active than their peer group from low-SES neighborhoods. This explanation is supported by the objectively measured levels of physical activity. The average activity level and mean number of minutes moderate to vigorous physical activity were lowest in the low-SES/low-walkable neighborhoods. Since there is evidence that low-SES adolescents often have lower levels of physical activity [57
] and that SES shows a clear inverse relationship with levels of overweight and obesity [59
], the present study's finding involve several implications for public policy and the results may be important for future environmental interventions or governmental initiatives. Improving income and education in low-SES communities is an ultimate solution. However, interim improvements can be based on emerging findings in the US that low-SES neighborhoods have numerous environmental deficits that are not reflected in walkability [61
]. Low-income neighborhoods, regardless of walkability, were disadvantaged in access to recreation facilities, walking/cycling facilities, aesthetics, pedestrian/traffic safety, and crime safety [62
]. Thus, if such environmental disparities are documented in Belgium, interventions could be targeted at low-income neighborhoods to provide more recreation facilities, improve walking/cycling facilities, enhance aesthetics by planting trees and removing graffiti, increase safety of street crossing, and institute more effective crime control methods. Walking for transport during leisure time was the only outcome for which an association with neighbourhood SES was found. Adolescent boys and girls living in low-SES neighbourhoods reported twice as much walking during leisure time than adolescent boys and girls living in high-SES neighbourhoods (13.4 (13.8) min/day vs. 6.5 (8.2) min/day). As stated by Ross et al. (2008), neighborhoods characterized by a higher level of poverty may have a culture in which people are outside on the streets, walk to visit someone, talk out on the street or just hang out on the street [53
]. This culture may encourage adolescents to walk for transport during leisure time. Another possible explanation is that lower-SES households have less access to automobiles. The lack of association between neighborhood SES and the other physical activity variables may be explained by a finding by Voorhees et. al. [36
]. In that study no association was found between SES and accelerometer-based physical activity in adolescent girls. However, some qualitative differences in types and location of activities between low- and high-SES girls were found. Lower-SES girls achieved higher levels of moderate to vigorous physical activity at home whereas higher-SES girls were more likely to participate in moderate to vigorous physical activity at school or community facilities. Furthermore, low-SES girls reported less moderate to vigorous physical activity in organized activity and were more involved in informal and spontaneous activities in comparison with high-SES girls. The Voorhees et al. (2009) results indicate that low-SES girls do more of their physical activity near home, suggesting that neighborhood environments could be more important for low-SES adolescents.
Limitations and strengths
Limitations of the present study included the cross-sectional study design which does not permit causal inferences. Second, despite stratified recruitment in higher- and lower-SES neighborhoods, education levels of parents were generally high. This likely reflects both high education levels of the university city and some recruitment bias. Third, neighborhoods were defined using existing statistical sectors. In that way artificial neighborhoods were created. A possible consequence of this method is that those artificial neighborhoods are not compatible with how respondents would define their own neighborhood. Neighborhood-level indicators of walkability and SES are relatively crude indicators, and individual buffers may more accurately reflect the environments that adolescents are exposed to. Strengths of the present study included the use of both objective and self-reported measures of physical activity. The average activity level, expressed in accelerometer counts/minute, gives an indication of the total level of physical activity. This outcome is based on the raw data provided by the accelerometer and is not dependent of processing differences. The FPAQ was used to collect information about adolescents' physical activity behavior in diverse domains. Questionnaire answers may be biased by social desirability, but to reduce over reporting, interviewers were trained to question high reports. Second, GIS databases were used to identify neighborhoods that maximized variation in walkability, and Belgian census data were used to identify neighborhoods of low and high-SES. Third, this study included a large sample of adolescent boys and girls. Finally, the study design and protocol was similar to PLACE, BEPAS, NQLS and SNAP studies in adults and importantly was conducted in the same neighborhoods as the BEPAS Adult study.