This population-based study of public high school students in Boston, MA adds to the accumulating body of evidence on the association between neighborhood safety and adolescent overweight. Our data suggest that feeling unsafe in one's neighborhood may be associated with an increased risk for overweight. However, this finding was only statistically significant among those within the Other race/ethnicity group. The Other group is comprised mainly of Asians and South Asians (65%), but also includes non-Hispanic bi- or multi-racial students, American Indians and Alaska Natives, Native Hawaiians and Other Pacific Islanders, and other students whose race did not fit into any of the other categories (e.g. those who were Guyanese, Belizean, or Brazilian). Thirty-eight percent of the students in the Other category were immigrants, compared to 34% of Hispanics, 31% of Blacks, and 12% of Whites.
Our findings are consistent with some previous studies that have found an association between neighborhood safety and adolescent overweight status [29
], but inconsistent with others that did not find an association [18
]. Few studies have examined the association between neighborhood safety and adolescent overweight stratified by race/ethnicity [32
]. Our findings highlight the importance of considering the moderating effects of race/ethnicity in the association between neighborhood safety and overweight status. The magnitude of the association varied by race/ethnicity and the association was statistically significant for only one racial/ethnic group. We were surprised that the association held for only those within the Other race/ethnicity group; students in the Other group were the least likely to feel unsafe in their neighborhoods and were the least likely to be at-risk or overweight. We expected to find associations among Blacks and Hispanics, given that Black and Hispanic adolescents report particularly high levels of neighborhood violence exposure [45
] and given that these adolescents are more likely to be overweight compared to Whites [10
]. Interestingly, the magnitude of the association found among those within the Other race/ethnicity group was weaker in our study compared to the two other studies finding an association [29
], perhaps a result of using prevalence ratios as the measure of effect (since these studies relied on odds ratios which likely produce inflated effect estimates when there is a relative high disease prevalence [33
], such as adolescent overweight) [10
] and/or because both studies found associations with traffic-related neighborhood safety (though the studies examined other safety-related variables). Because we examined general feelings of neighborhood safety, we can only speculate on which of the multiple aspects of neighborhood safety (e.g. violence/crime, traffic and road hazards, neighborhood disorder) might influence overweight status, but our interviews with students during the pilot testing phase strongly suggest that neighborhood violence is their primary neighborhood safety concern.
There are several pathways by which neighborhood safety might be related to adolescent overweight. One possible interpretation of our findings is that adolescents' concerns about neighborhood safety might decrease their willingness to engage in outdoor physical activity (e.g. walking and playing sports in their neighborhood), promote their use of non-ambulatory transportation options (e.g. use of buses, subways, and automobiles), and/or encourage sedentary behaviors (e.g. television watching, playing computer games, and playing video games in the home), all of which could contribute to being overweight [50
]. Residing in an unsafe neighborhood might also increase stress (causing a release of cortisol) and result in overweight [52
]. Evidence indicates that exposure to neighborhood violence, which is a potentially chronic traumatic stressor, is associated with increased cortisol secretion in adolescents [56
Additionally, since we do not know whether students who perceived their neighborhoods to be unsafe actually live in unsafe neighborhoods and we are not aware of any study in this area that has conducted a validity check of perceptions of neighborhood safety, we explored whether students who reported feeling unsafe were more likely to have experienced neighborhood violence. We found that those students who felt unsafe in their neighborhoods were more likely than those who felt safe to believe that gang violence was a serious problem in their neighborhood or school and to have seen someone in their neighborhood attacked with a weapon (other than a firearm) in the past 12 months. However, they were not significantly more likely to have seen someone beaten up in their neighborhood in the past 12 months. These findings potentially indicate that witnessing a physical assault may not be a salient dimension of perceived neighborhood safety for adolescents, while neighborhood gang violence and seeing someone in their neighborhood assaulted with a weapon are important aspects of perceived neighborhood safety for them. This is an important contribution to the literature.
There is a need for additional research to clarify the role that neighborhood safety (including neighborhood violence) plays in the adolescent overweight epidemic and to understand salient aspects of perceptions of neighborhood safety. Perceived neighborhood safety is a complex, multidimensional psychosocial construct. Because there is no consensus on the definition of neighborhood safety in health research, qualitative research to explore the dimensions of neighborhood safety and to determine which dimensions of safety are most salient to adolescents at risk of overweight is warranted; this may vary by race/ethnicity. As all of the studies examining neighborhood safety and overweight in adolescents were cross-sectional, researchers should examine this association with prospective cohort designs. In addition, experimental research, e.g., cluster randomized trials (e.g. where neighborhoods might be randomly assigned to an intervention that improves safety) and natural experiments (e.g. new policies promoting police presence in certain neighborhoods to enhance safety), could be conducted (as they are the strongest evidence for temporality) to understand the effects of neighborhood safety on adolescent overweight. Neighborhood environmental interventions [57
] and residential mobility experiments [58
] hold promise to reduce the prevalence of adolescent overweight. Beyond examining perceived neighborhood safety, research can also examine objective neighborhood safety (e.g. crime statistics to ascertain one's proximity to neighborhood violence) in relation to overweight. Including both subjective and objective reports of neighborhood safety in the same study, although these concepts are likely interconnected, can be beneficial. Each measure might capture distinct neighborhood features; thus, this strategy might ensure optimal measurements of neighborhood safety features.
This study has implications for primary and secondary prevention of adolescent overweight through the development of contextually-relevant interventions and policies. Adolescents in our sample report fear in their neighborhood and high levels of exposure to neighborhood violence, as others [45
] have shown. This is concerning on its own and also because neighborhood safety may be a factor in adolescent overweight. Our study underscores the importance of policy-level overweight prevention strategies via reducing neighborhood safety concerns. A relevant policy intervention is crime prevention through environmental design [61
], which would involve changes to the physical environment (such as elimination of hiding spots, landscaping trees and shrubs, and increased surveillance via increased lighting, closed-circuit television/surveillance cameras in public spaces, and/or security guards). Problem-oriented policing, i.e. increased local police attention in "hot spots" or high-crime locations, is another method used to reduce and prevent crime and violence [61
]. Other potential strategies to reduce neighborhood safety concerns are revitalizing neighborhood watch programs to monitor criminal activity and liaising with police to enhance the protection of places where one can engage in physical activity (e.g. parks and recreation facilities). Furthermore, interventions that offer adolescents' safe havens (such as after-school programs) [63
], those that are focused on community development (e.g. ensuring neighborhood resources [such as organizations, services, and employment opportunities]) [64
], and building collective efficacy among community members [66
] could prove beneficial to reduce neighborhood violence. Lastly, it is imperative that behavior change programs (e.g. behavioral weight loss programs) recognize the neighborhood social context of the participants. Physicians, for example, should recommend indoor physical activity for overweight prevention and weight management for adolescents who reside in unsafe neighborhoods.
These findings should be interpreted in light of the limitations of our study. First, we relied on cross-sectional data; thus, the study does not inform us about the direction of causation (e.g. whether the exposure preceded the outcome). However, despite the well-known limitations of cross-sectional data, our study hypotheses and directionality have intuitive appeal and were based on conclusions from past research. Additionally, we did not evaluate specific dimensions of neighborhood safety (as previously mentioned) and we did not evaluate objective measures of neighborhood safety (e.g. crime statistics or statistics on the number of sex offenders); we were particularly interested in understanding perceived neighborhood safety rather than the actual occurrence of neighborhood crime or violence. Third, we relied on self-reported height and weight data for BMI, which has the potential for misclassification because of inaccurate reporting. Past research, however, has found that adolescents can provide valid reports of height and weight [67
]. Though the gold standard is to collect objectively measured height and weight data, this was not practical nor a central focus of the parent study. Residual confounding might also be a concern, as the survey might have excluded important confounding variables associated with both the independent variable and the dependent variable (e.g. household income, parental education and residential stability might be confounders), but we were unable to account for these variables in the adjusted regression analyses because they were not asked in the BYS data collection instrument. Due to expected high rates of non-response, we did not ask these questions. Finally, this study was conducted in one specific geographically-defined population; thus, these findings might only be generalizable to adolescents in comparable urban locations.