Exposure to Neighborhood Crime Is Associated with Lower Levels of Physical Activity and Higher Obesity Risk among Adolescent Girls,but Not Boys

Abstract

Background:

Neighborhood safety has been linked to physical activity and obesity among adolescents, but few studies have used objective measures of crime or examined associations by gender. The goal of this study was to investigate if exposure to objectively measured neighborhood crime was associated with physical activity and overweight/obesity among adolescents, and if this association varied by gender.

Methods:

Georeferenced data from the National Health and Nutrition Examination Survey (NHANES) 1999–2006 were linked to census tract-level data on crime reports from the Federal Bureau of Investigation, focusing on a sample of adolescents 12–20 years of age (n = 15,261). Our outcome variables were engaging in moderate-to-vigorous physical activity in the previous week and overweight/obesity status, defined as ≥85th percentile in the CDC's age- and sex-specific BMI-for-age growth charts (based on measured weight and height). Our exposure variable was neighborhood-level violent crime risk, categorized as low vs. high. Gender-stratified three-level multilevel logistic regression models were fitted adjusting for individual, family, and neighborhood-level demographic and socioeconomic variables and adolescents' diet quality.

Results:

In fully adjusted models, girls exposed to high crime neighborhoods had 26% lower odds of engaging in moderate-to-vigorous physical activity during the previous week [odds ratio (OR) = 0.74, confidence interval (95% CI) = 0.59–0.92] and 27% higher odds of being overweight/obese (OR = 1.27, 95% CI = 1.02–1.58), compared with adolescent girls exposed to low crime neighborhoods. No significant associations were found between exposure to neighborhood crime and physical activity and overweight/obesity among boys.

Conclusion:

Prevention efforts focused on encouraging adolescents to be physically active need to account for gender-specific barriers to exercise outdoors.

Introduction

It is well established that the neighborhoods where people live affect their health.¹ In particular, neighborhood safety has been associated with negative health behaviors and health outcomes, including a lower likelihood to engage in physical activity and an increased risk of obesity.^2–4 A recent systematic review and meta-analysis focused on longitudinal studies found evidence of an association between neighborhood safety and physical activity among children and adolescents, but no effect for obesity.⁵ This meta-analysis, however, included mostly subjective measures of neighborhood safety, with only 1 of the 22 included studies measuring neighborhood crime objectively.⁶ Although subjective perceptions of crime can influence people's behavior, it is important to know the potential impact of objectively measured crime because population-level interventions (as opposed to individual level, i.e., to reduce actual crime vs. perception of crime) have the potential of influencing individual-level behaviors at a larger scale through “contagion effects.”^7,8 Moreover, actual crime may influence the surrounding built environment, in addition to the social environment, which also affects individual-level obesity-related behaviors.⁹

In addition, the differential effect of neighborhood safety/crime on weight-related behaviors and overweight/obesity by gender has seldom been reported.^6,10,11 This is despite existing evidence of a differential effect of the neighborhood environment on women vs. men, because of a hypothesized increased susceptibility and/or exposure of women to neighborhood effects.¹² For example, Derose et al. have recently reported that women living in high-poverty neighborhoods in Los Angeles, California have lower physical activity levels than men, and this is partly because of safety concerns related to accessing outdoor parks.¹³ Similarly, Yang et al. report that although neighborhood characteristics affect boys' and girls' obesity risk, girls are more sensitive to these effects, especially older girls.¹⁴ In the abovementioned systematic review, only two of the included studies reported gender differences in the association between neighborhood safety and children's physical activity¹⁵ and obesity risk,¹⁶ although both used subjective measures of neighborhood safety.

The objectives of this study, therefore, were to (1) examine the association between objectively measured neighborhood crime and physical activity (self reported) and overweight/obesity (measured) in a national sample of US adolescents and (2) explore if this association varies by gender.

Methods

Data for this study came from the National Health and Nutrition Examination Survey (NHANES) pooled cross-sectional samples from 1999 to 2006. We focused on adolescents 12–20 years of age (n = 15,261), who were clustered within households and within neighborhoods; neighborhoods were defined by the family's census tract of residence (n = 2400). Georeferenced NHANES data¹⁷ were linked to neighborhood crime risk data, obtained from ESRI community data sources including the most recent estimates and projections of the Uniform Crime Report (UCR) of Federal Bureau of Investigation (FBI) from 2000 to 2008 Applied Geographic Solutions.¹⁸

NHANES data used included our outcome variables—physical activity and overweight/obesity—as well as adolescents' diet quality and sociodemographic information on adolescents and their families. An adolescent was classified as being physically active if she/he reported engaging in vigorous or moderate physical activity in the previous week. Based on measured weight and height, overweight/obesity status was defined as being ≥85th percentile in the CDC's age- and sex-specific BMI-for-age growth charts.¹⁹ The Healthy Eating Index (HEI), a measure of diet quality related to compliance to the Dietary Guidelines for Americans, was estimated following standard procedures²⁰; HEI scores range from 0 to 100 with diet quality classified as good if HEI >80, needs improvement if 51 ≤ HEI ≤80, or poor if HEI <51. Sociodemographic information included adolescents' age (12–14, 15–17, and 18–20 years), gender, and race/ethnicity [non-Hispanic (NH) white, NH black, Hispanic/Latino, and other]; parental education level (<9th grade, 9–12th grade, and high school graduate or more); family poverty status [income-to-poverty ratio (IPR) <1, IPR between 1 and 2, and IPR ≥2]; and urban/rural residential location.

Data on neighborhood crime risk obtained from ESRI included a series of standardized indices (relative to national risk) for a range of crimes including murder, rape, robbery, assault, burglary, and theft (including car theft), as categorized by the FBI's UCR Program.¹⁸ These indices include data from UCR and local jurisdictions, where available, and are provided annually at the block group level. Results are scaled to jurisdiction and weighted by population, then converted to indices relative to the national total (national average set at 100). Each participant in our study, then, was assigned a crime risk value aggregated at their census tract of residence for the given year of the NHANES survey. Violent crime risk was used in this analysis given the proximity of violent vs. nonviolent crime to residential address, and the unweighted nature of the overall crime risk indices. The total violent crime risk index ranged from 1 to 1036 (mean = 138.76, SD = 134). Neighborhood crime risk, which was our exposure variable, was then categorized as high crime risk (≥191) vs. low crime risk (<191) based on the 75th percentile.

Statistical Analysis

Descriptive statistics were used to characterize the sample. Three-level multilevel logistic regression models, fitted in SAS v9.0 with PROC GLIMMIX, were carried out since adolescents (level 1) were nested within households (level 2) and within neighborhoods (level 3) and our outcome variables (physical activity and overweight/obesity) were dichotomous. A pseudo-intraclass correlation coefficient (ICC)²¹ was calculated to estimate the amount of variance explained in individual-level outcomes that could be attributed to differences between neighborhoods. Separate analyses were conducted for adolescent girls and boys to assess the role of gender as an effect modifier in the association between neighborhood crime and physical activity and overweight/obesity. All analyses were weighted for sampling design and survey nonresponse, and multivariable analyses were adjusted for age, race/ethnicity, HEI score, parental education, family poverty status, survey year, number of years living in the neighborhood, urban/rural status, overweight/obesity (for model with physical activity as the outcome), and physical activity (for model with overweight/obesity as the outcome).

Results

Table 1 provides the characteristics of the sample and differences by overweight/obesity status. The prevalence of overweight/obesity among adolescents was 37%, whereas 57% reported engaging in no physical activity during the previous week. Physical activity levels were similar among overweight and nonoverweight adolescents. The proportion of overweight/obese was higher among NH black and Latino adolescents, and among those living in families with an IPR <1.

Table 1.

Characteristics of a Sample of Adolescents 12–20 Years of Age According to Overweight/Obesity Status, National Health and Nutrition Examination Survey 1999–2006

	Overweight or obese^a		Not overweight or obese		Total
	n = 5648		n = 9613		n = 15,261
	Row (%)	Mean (SD)	Row (%)	Mean (SD)	Column (%)	Mean (SD)
Age (years)
12–14	40.9		59.1		38.7
15–17	34.5		65.3		35.5
18–20	25.8		63.4		25.8
Gender
Boys	37.6		62.4		49.5
Girls	36.4		63.6		50.5
Race/ethnicity
NH white	32.0		68.0		27.6
NH black	38.7		61.3		30.6
Hispanic/Latino	39.9		62.1		37.8
Other	31.3		68.7		4.0
Parental education
<9th grade	37.1		62.9		86.1
9–12th grade	34.6		65.4		8.6
High school graduate/GED or higher	32.5		67.5		5.3
Family poverty status
IPR <1	40.8		59.2		31.7
1 < IPR <2	37.4		62.3		27.1
IPR ≥2	34.1		65.9		41.2
Urbanicity
Urban	38.1		61.9		29.5
Rural	36.6		63.4		70.5
Healthy Eating Index scores^b		57.6 (11.6)		59.5 (11.7)		58.8 (11.5)
Past-week physical activity
None	36.7		63.3		57.4
Moderate or vigorous	37.4		62.6		42.6

Proportions and means based on nonmissing values.

Defined as being ≥85th percentile in the CDC's age- and sex-specific BMI-for-age curves.

Healthy Eating Index scores ranged from 0 to 100, with a total score of >80 considered “good,” scores of 51–80 indicating “needs improvement,” and scores <51 considered “poor.”

GED, General Equivalency Diploma; IPR, income-to-poverty ratio; NH, non-Hispanic.

We observed a higher level of clustering at the neighborhood level for physical activity (ICC = 3.1%) than for overweight/obesity (ICC = 2.3%). For girls, ∼4% of differences in physical activity and overweight/obesity outcomes could be explained by differences in the neighborhood of residence, with a lower variance explained for boys (ICC for physical activity 2.5%, ICC for overweight/obesity 1.5%).

Table 2 provides the results of the multilevel logistic regression analyses. After adjusting for age, gender, race/ethnicity, HEI score, parental education, family poverty status, survey year, number of years living in the neighborhood, urban/rural status, and overweight/obesity or physical activity (depending on the model), there was a weak and nonsignificant association between exposure to high crime in the neighborhood and physical activity, and a positive and significant association between high neighborhood crime and overweight/obesity. When the sample was stratified by gender, significant associations emerged for girls only: in fully adjusted analyses, the odds of engaging in vigorous or moderate physical activity in the previous week were 36% lower for girls living in high crime neighborhoods than those living in low crime neighborhoods. In addition, girls who lived in high crime vs. low crime neighborhoods had 27% higher odds of being overweight/obese.

Table 2.

Results from Multilevel Logistic Regression Analyses Predicting Physical Activity and Overweight/Obesity Status by Neighborhood Crime Risk (High vs. Low) among a Sample of Adolescents 12–20 Years of Age from the National Health and Nutrition Examination Survey 1999–2006 (n = 15,261)

	Physical activity^a			Overweight/obesity^b
	Overall OR (95% CI)	Girls OR (95% CI)	Boys OR (95% CI)	Overall OR (95% CI)	Girls OR (95% CI)	Boys OR (95% CI)
Neighborhood crime risk
High	0.86 (0.73–1.00)	0.74 (0.59–0.92)	0.98 (0.80–1.21)	1.11 (1.04–1.31)	1.27 (1.02–1.58)	1.11 (0.94–1.31)
Low (reference)	1.00	1.00	1.00	1.00	1.00	1.00

Neighborhood crime risk index estimated based on a series of standardized indices for violent crimes relative to the national risk. Based on the 75th percentile cutoff, high crime risk was defined as having an index ≥191, with low crime risk defined as having an index <191. All analyses were adjusted by age, race/ethnicity, parental education, family poverty status, survey year, years living in neighborhood, Healthy Eating Index score, physical activity (for overweigh/obesity model), and overweight/obesity (for physical activity model).

Defined as having engaged in vigorous or moderate physical activity during the previous week.

Defined as being ≥85th percentile in the CDC's age- and sex-specific BMI-for-age curves.

CI, confidence interval; OR, odds ratio.

Bold values represent statistical significance at p<0.05.

Discussion

In this study focused on a US nationally representative sample of adolescents, we found that exposure to high levels of neighborhood crime was associated with a lower likelihood of engaging in physical activity and higher odds of being overweight/obese among girls only. A recent systematic review and meta-analysis focused on longitudinal studies found that neighborhood safety has a significant but minimal effect on children's physical activity levels, whereas no effect was found for overweight/obesity.⁵ The authors did not conduct separate analyses by gender, however, despite evidence from some of the articles included in the systematic review of gender differences in the impact of neighborhood safety on weight-related behaviors.^15,16 In addition, this meta-analysis included studies focused predominately on subjective perceptions of neighborhood safety, contrary to our study that included only objective neighborhood crime reports.

Prins et al. focused on studying sociodemographic differences in disengagement from sports in adolescence among a sample in Rotterdam, the Netherlands.¹⁵ They found that girls were more likely than boys to disengage in sports over time, and that perceived neighborhood safety partially explained this gender difference.¹⁵ Similar to our findings, Bacha et al. found that neighborhood safety, as perceived by the mother in third grade, was associated with a higher weight status in fifth grade for girls, but not for boys, in a US nonrepresentative sample.¹⁶ Interestingly, though, they found that this association was not explained by less time spent outdoors or more television viewing.¹⁶ The authors hypothesized that the stress caused by living in an unsafe neighborhood may be one of the culprits, as chronic stress has been linked to obesity²² and neighborhood stressors seem to affect women more than men.¹² On the contrary, Evenson et al. found that perceived neighborhood safety was not associated with physical activity among a sample of girls attending 6th grade in 36 schools across the United States.²³

Only a handful of studies to date have focused on the effect of objective crime measures on physical activity among adolescents while reporting gender differences, showing results comparable with ours. Gomez et al. reported that exposure to violent crime within half mile of the home, as obtained from police reports, was associated with decreased outdoor physical activity for girls but not boys among a sample of Mexican American seventh graders in San Antonio, TX.¹⁰ Similar results were reported among Hispanic girls, but not boys, in Nashville, TN.¹¹ In contrast, using data from waves I and III of the National Longitudinal Study of Adolescent Health (Add Health) and county-level reports of nonviolent and violent crime, Boone-Heinonen and Gordon-Larsen reported no gender differences in the impact of crime on physical activity.⁶ Although our study is focused on a nationally representative sample of adolescents, as is the one by Boone-Heinonen and Gordon-Larsen, our crime reports come from a smaller geography (census tract vs. county level), which may better reflect neighborhood crime exposure. In addition, our study focused only on violent crimes, whereas the study by Boone-Heinonen and Gordon Larsen included both violent and nonviolent crimes.

Two qualitative studies focused on African American girls also support our findings that girls living in violent neighborhoods are less likely to be physically active, particularly outdoors, because of safety concerns. One study based in Newark, NJ found that mothers relied on confining their susceptible daughters indoors in an effort to protect them from neighborhood violence, thus limiting their opportunities to engage in physical activity.²⁴ The other study carried out in Baltimore, MD found that girls' outdoor activity levels were heavily influenced by concerns for safety, more so than for boys.²⁵

In terms of overweight/obesity, only a handful of studies have investigated the association between exposure to objectively measured crime and overweight/obesity among children or adolescents,^4,26 with little discussion on gender differences. We found that girls exposed to high levels of crime in their neighborhoods had 27% higher odds of being overweight/obese than girls living in low-crime areas. Miranda et al. found that children 2–18 years of age exposed to high levels of neighborhood violent crime in Durham, NC had 48% increased odds of overweight/obesity, with no indication on variations based on age group or gender.⁴ Our findings show a more modest impact of neighborhood crime on overweight/obesity (27% vs. 48%), which may be explained by the differences in age groups and/or geographical representation (United States vs. Durham, NC). Another study focused on preschool-aged children in Seattle, WA found that high frequency to proximal crime was positively associated with BMI,²⁶ also with no reports on gender differences. Forsyth et al. reported some interesting findings among a diverse sample of adolescents in Minneapolis/St. Paul, MN.²⁷ The authors found that perceived neighborhood crime was associated with BMI for both boys and girls, but objectively reported crime against persons (including homicide, rape, assault, and robbery) was associated with BMI only among girls.²⁷ Unfortunately, the authors did not speculate on a reason for this discrepancy, but do report that in their sample perceived and reported crime were significantly and positively correlated, and boys and girls were exposed to similar levels of neighborhood crime.²⁷ Future studies should investigate the mechanisms involved in the apparent gender-specific effect of objectively measured neighborhood crime on weight status and weight-related behaviors.

Strengths and Limitations

Although previous research has highlighted the differential effects that neighborhood exposures have on girls compared with boys,¹⁴ this study is one of the few to evaluate the differential effects by gender of objectively measured neighborhood crime on physical activity and weight status among adolescents. Our study is based on a nationally representative sample that allows for the generalizability of our results. Both neighborhood crime and adolescents' weight and height (used to estimate BMI) were objectively measured; however, our physical activity variable was based on self-reported data. In addition, there are many ways of categorizing physical activity (any vs. none, number of minutes engaged in physical activity, type of physical activity performed, etc.). We were interested in an overall effect of neighborhood crime on engaging in any moderate/vigorous physical activity; however, our results may have been different if we would have used a different physical activity classification.

We were able to include covariates at three levels of influence, including individual, family, and neighborhood levels; however, an important sphere of influence for adolescents—the school—was not included because of lack of relevant data. Another limitation includes the cross-sectional nature of our data, limiting our ability to assess the temporality of the associations observed in this study, and the difference in measurement dates for the exposure and outcome variables. Moreover, our crime risk variable was based on violent crimes only; results may be different if nonviolent crimes were to be included.

Conclusions

Among a US representative sample of adolescents, exposure to high levels of violent crime in the neighborhood of residence was associated with lower odds of engaging in physical activity and higher odds of overweight/obesity only for girls, with no significant associations found for boys. It is increasingly evident that violence in the neighborhood is a detrimental exposure for children and adolescents' obesity-related behaviors, particularly influencing girls. Interventions focused on incentivizing physical activity among adolescents should be aware of the apparent gender disparity in accessing safe spaces for outdoor activity. Future studies should investigate if the association between exposure to neighborhood crime and overweight/obesity is mediated through physical activity or if there are additional mechanisms at play, for example stress, as suggested by Bacha et al.,¹⁶ and if these mechanisms at play are also gender specific. In addition, our results suggest that policies aimed at curbing neighborhood crime and violence would improve adolescent girls' health behaviors and outcomes, highlighting the need for cross-disciplinary collaborations when designing programs and policies for the betterment of the community.

Footnotes

Acknowledgments

This study was supported by grants from the Health Resources & Services, Administration Maternal & Child Health Epidemiology Doctoral Training Program (T03MC07649 to K.P.T.), and the Centers for Disease Control and Prevention (K01SH000002 to K.P.T.).

Author Disclosure Statement

No competing financial interests exist.

References

Diez Roux

, Mair

. Neighborhoods and health. Ann NY Acad Sci, 2010; 1186:125–145.

Robinson

, Carnes

, Oreskovic

. Spatial analysis of crime incidence and adolescent physical activity. Prev Med, 2016; 85:74–77.

Janssen

. Crime and perceptions of safety in the home neighborhood are independently associated with physical activity among 11–15 year olds. Prev Med, 2014; 66:113–117.

Miranda

, Edwards

, Anthopolos

, et al. The built environment and childhood obesity in Durham, NC. Clin Ped, 2012; 51:750–758.

, Yang

, Hoschke

, et al. Influence of neighbourhood safety on childhood obesity: A systematic review and meta-analysis of longitudinal studies. Obes Rev, 2017; 18:1289–1309.

Boone-Heinonen

, Gordon-Larsen

. Life stage and sex specificity in relationships between the built and socioeconomic environments and physical activity. J Epidemiol Community Health, 2011; 65:847–852.

Rose

. Sick individuals and sick populations. Int J Epidemiol, 2001; 30:427–432.

Cohen

, Scribner

, Farley

. A structural model of health behavior: A pragmatic approach to explain and influence health behaviors at the population level. Prev Med, 2000; 30:146–154.

Foster

, Giles-Corti

. The built environment, neighborhood crime and constrained physical activity: An exploration of inconsistent findings. Prev Med, 2008; 47:241–251.

10.

Gomez

, Johnson

, Selva

, Sallis

. Violent crime and outdoor physical activity among inner-city youth. Prev Med, 2004; 39:876–881.

11.

Van Bakergem

, Sommer

, Heerman

, et al. Objective reports versus subjective perceptions of crime and their relationships to accelerometer-measured physical activity in Hispanic caretaker-child dyads. Prev Med, 2017; 95 Suppl:S68–S74.

12.

Stafford

, Cummins

, Macintyre

, et al. Gender differences in the associations between health and neighbourhood environment. Soc Sci Med, 2005; 60:1681–1692.

13.

Derose

, Bing

, Williamson

, Cohen

. Gender disparities in park use and physical activity among residents of high-poverty neighborhoods in Los Angeles. Womens Health Issues, 2018; 28:6–13.

14.

Yang

, Jiang

, Xu

, et al. A cross-sectional study of the influence of neighborhood environment on childhood overweight and obesity: Variation by age, gender, and environment characteristics. Prev Med, 2018; 108:23–28.

15.

Prins

, Kamphuis

, van Empelen

, et al. Explaining socio-demographic differences in disengagement from sports in adolescence. Eur J Public Health, 2013; 23:811–816.

16.

Bacha

, Appugliese

, Coleman

, et al. Maternal perception of neighborhood safety as a predictor of child weight status. The moderating effect of gender and assessment of potential mediators. Int J Ped Obes, 2010; 5:72–79.

17.

Research Data Center, National Center for Health Statistics. National Health and Nutrition Examination Survey 2005–2006. Documentation, Codebook and Frequencies: Geocoding variables. Hyattsville, MD, National Center for Health Statistics, 2009.

18.

Federal Bureau of Investigation, Uniform Crime Report, 2000–2008. Applied Geographic Solutions, most current Estimates and Projections.

19.

Kuczmarski

, Ogden

, Guo

, et al. 2000 CDC growth charts for the United States: Methods and development. Vital Health Stat, 2002; 11:1–190.

20.

Guenther

, Casavale

, Reedy

, et al. Update of the Healthy Eating Index: HEI-2010. J Acad Nutr Diet, 2013; 113:569–580.

21.

Snijders

, Boskers

. Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling. Sage Publications: London, United Kingdom, 1999.

22.

Björntorp

. Do stress reactions cause abdominal obesity and comorbidities?. Obes Rev, 2001; 2:73–86.

23.

Evenson

, Scott

, Cohen

, Voorhees

. Girls' perception of neighborhood factors on physical activity, sedentary behavior, and BMI. Obesity, 2007; 15:430–445.

24.

Dias

, Whitaker

. Black mothers' perceptions about urban neighborhood safety and outdoor play for their preadolescent daughters. J Health Care Poor Underserved, 2013; 24:206–219.

25.

Ries

, Gittelsohn

, Voorhees

, et al. The environment and urban adolescents' use of recreational facilities for physical activity: A qualitative study. Am J Health Prom, 2008; 23:43–50.

26.

Gartstein

, Seamon

, Thompson

, Lengua

. Community crime exposure and risk for obesity in preschool children: Moderation by the hypothalamic-pituitary-adrenal-axis response. J Ped Psychol, 2017; 1–13. DOI: 10.1093/jpepsy/jsx116

27.

Forsyth

, Wall

, Choo

, et al. Perceived and police-reported neighborhood crime: Linkages to adolescent activity behaviors and weight status. J Adolesc Health, 2015; 57:222–228.