Neighborhood Profiles and Body Mass Index Trajectory in Female Adolescents and Young Adults

Abstract

Purpose:

The purpose of this study is to identify distinct neighborhood profiles patterned by key structural, physical, and social characteristics and test whether living in different profiles are associated with body mass index trajectories during adolescence in racial/ethnic minority female youth.

Methods:

Participants were 1,328 sexually active female adolescents and young adults aged 14 to 23 years, predominately Hispanic and Black, enrolled in an HPV4 vaccine (Gardasil) surveillance study at a large adolescent health clinic in New York City between 2007 and 2018. Body mass index (BMI) was calculated from weight and height every six months. A comprehensive set of neighborhood structural, social, and physical characteristics from multiple national and state datasets were linked to each participant based on home address.

Results:

Latent profile analysis revealed five distinct neighborhood profiles in New York City: High Structural/High Social Advantage, Moderate Advantage/Low Crime, Low SES/High Activity, Low SES/High Social Advantage, and High Disadvantage. Results from multilevel growth curve analysis revealed that living in Low SES/High Activity neighborhoods was associated with a lower BMI at age 22 (b = −1.32, 95%CI = [−2.49, −0.16]), as well as a slower increase in BMI from age 14 to 22 years (b = −0.22, 95%CI = [−0.46, 0.02]), compared to the High Disadvantage profile.

Conclusions:

Our findings suggest that improving neighborhood structural, social, and physical environments may help promote healthy weight and reduce health disparities during adolescence and young adulthood.

Introduction

Adolescent obesity is a critical public health concern in the United States (US) due to its prevalence and high risk for comorbidities even into adulthood.^1,2 Despite the recent plateauing of obesity rates among adolescents in the US,³ socioeconomic status (SES) and racial/ethnic disparities in obesity are growing, with Black and Hispanic youth and those from low SES backgrounds having the highest rates of obesity.^3,4 Adolescence is a sensitive period for excess weight gain as youth experience a growth spurt and increased fat deposition beginning at puberty. Girls are at elevated risk across adolescence since they accumulate a higher percentage of body fat than boys.⁵ These dynamic physical changes, coupled with greater autonomy to make decisions about diet and increased time spent outside the home, highlight the importance of understanding environmental factors during the critical periods of adolescence to promote healthy weight and reduce disparities in obesity.

Neighborhoods are important environments for shaping diet, physical activity, and obesity status early in life and worthwhile settings for public health interventions. Theories have suggested that neighborhoods with concentrated poverty and crime could contribute to a wide range of problematic outcomes due to weakened social ties, limited collective socialization, and reduced access to institutional resources (e.g., high-quality schools, parks, libraries).^6,7 Growing research shows that neighborhood structural (e.g., SES, unemployment, crowded housing), physical (e.g., physical activity resources), and social (e.g., violence, low social cohesion, social norms favoring unhealthy behaviors) disadvantages are positively associated with youth weight status, independent of personal demographic characteristics and family SES.^8–15 However, few studies have evaluated the influence of neighborhood structural, physical, and social characteristics simultaneously and longitudinally.

The current study used latent profile analysis to identify neighborhood profiles (i.e., homogenous subgroups) and then investigated how distinct profiles predicted trajectories of clinician-measured body mass index (BMI) from age 14 to 22 years among female adolescents and young adults in New York City (NYC). Stark disparities in SES, physical condition, social norms, and health outcomes exist across NYC neighborhoods,¹⁶ which allowed us to gain a deeper understanding of the effect of neighborhood context and social determinants on obesity among urban adolescent populations. We hypothesized that living in a disadvantaged neighborhood—defined by deficits across structural, physical, and social domains—would be associated with the highest BMI with the steepest increase during adolescence. We also expected to find mixed neighborhood profiles (i.e., advantaged in certain domains but disadvantaged in others) and we explored how they compared to the most disadvantaged neighborhood profile with respect to youth BMI trajectories.

Methods

Data

This study consisted of sexually active female participants enrolled in an ongoing human papillomavirus (HPV) surveillance study at a large urban adolescent health clinic in NYC. Female participants aged 13 to 21 years who were sexually active and were planning to receive or had already received the HPV vaccine (GARDASIL®) were recruited on a rolling basis and followed every six months up to age 25. Written informed consent was collected from all participants before enrollment, with a waiver of parental consent for adolescents 13 to 17 years. During each study visit, participants received a comprehensive gynecological examination and completed self-administered questionnaires assessing demographic characteristics, maltreatment history, and psychosocial functioning in a private office space. This study was approved by the institutional review board at the Icahn School of Medicine at Mount Sinai. A full description of the study design has been published previously.¹⁷

The analytic sample included 1,328 participants who completed a baseline visit from October 2007 to April 2018 and provided their home address (see eFigure1 for a flow chart of participant selection). While participants had varying numbers of study visits, the current analysis was restricted to a total of 4,465 observations from the first five visits (i.e., baseline, 6 months, 12 months, 18 months, and 24 months), as the original aim of the study was to follow the participants for two years. The overall completion rate for follow-up visits was 81.02%. Compared to participants who had at least one follow-up visit, those who did not were less likely to be born in the US (χ² (1) = 9.70, p = 0.002) and attend school at baseline (χ² (1) = 4.90, p = 0.027); the two groups were not statistically different in neighborhood residence or other demographic characteristics.

Measures

Body Mass Index.

Height and weight were collected at each visit using a standardized clinic protocol by trained medical staff and recorded in electronic medical records. BMI was calculated by the ratio of weight in kilograms over height in meters squared to allow for comparison across all age groups included.¹⁸

Neighborhood characteristics.

Home addresses at enrollment were used to determine individuals’ census tracts, which were then linked to neighborhood-level data from the American Community Survey, Crime Open Database, Primary Land Use Tax Lot Output Data Files, and the NYC Department of Health and Mental Hygiene Community Health Survey. As the enrollment year of the cohort spanned from 2007 to 2018, we selected neighborhood variables close to the year of 2012 (i.e., the median) and prioritized multiple-year aggregate measures over single-year measures, to approximate the neighborhood exposure for the cohort. Based on prior literature, 12 neighborhood variables were selected to represent the structural, physical and social environment (see eTable1 for data sources and operationalization for neighborhood variables).^9,14 Among the 2,168 designated census tracts in NYC, 87 tracts with zero or small population (under 100 people) were excluded to avoid missing or unstable parameter estimates.

Covariates.

Demographic characteristics were assessed at baseline via self-reported age, race/ethnicity (mutually exclusive categories of Hispanic, Mixed Hispanic and Black, non-Hispanic Black, or non-Hispanic Others including White, Asian, Native American and other), nativity status (1 = born in the US; 0 = not), sex with a female partner (1 = ever had sex with a girl or female partner; 0 = not), parent receipt of public assistance such as welfare checks or Medicaid, food insecurity (1 = sometimes or often do not have enough to eat in the household, and 0 = enough to eat), parent(s) living status (mutually exclusive categories of lived together, separated, or deceased), and whether the participant currently attended school. Self-reported age at menarche was included as a covariate due to its correlation with BMI.^19,20 To account for the potential effect of neighborhood change on BMI, we controlled for whether a participant changed address (1 = yes; 0 = no) during a follow-up visit.

Statistical Analysis

First, latent profile analyses (LPA) were conducted to empirically identify neighborhood profiles among NYC census tracts based on all 12 structural, physical, and social characteristics.²¹ We sequentially estimated two to nine profiles.²² The best-fitting model was chosen based on the interpretability as well as multiple fit statistics, including Bayesian Information Criterion (BIC), Akaike Information Criterion (AIC), Lo-Mendel-Rubin (LMR), and adjusted log-likelihood test; entropy was also examined to assess profile classification accuracy.²² Specifically, smaller absolute values of BIC, AIC, and LMR log-likelihood indicate better model fit; a significant LMR test p-value suggests that the k profile model fits better than the k-1 profile model; and an entropy value closer to 1 (range from 0 to 1) indicates clearer classification between latent profiles.²² After identifying the optimal model, neighborhoods were assigned to their most likely latent profile. LPA was conducted in Mplus Version 8.3.²³

Next, multilevel growth curve models were estimated to examine associations between neighborhood profiles and BMI trajectories across adolescence, with multiple visits over time (Level 1) nested within persons (Level 2) nested within neighborhoods (Level 3). Age, in half-year, served as the time variable to estimate age-related changes in BMI. The age of participants during the study period ranged from 14 through 23 years; five participants were top-coded at 22 years to avoid potential bias introduced by very small frequencies of specific age groups. Age was centered at 22 years, given that variability in BMI increased with age. A series of models were estimated: 1) an unconditional growth model (Model 1) to describe the average trajectory of BMI including the fixed effects of intercept (the average BMI at age 22) and slope (the rate of change in BMI per year); 2) a conditional growth model (Model 2) adding the latent profile membership, coded as dummy variables with the most disadvantaged profile as the reference group, as predictors of the random intercept (to test whether a certain profile had a significantly different average BMI at age 22 compared to the reference profile) and random slope (to test whether a certain profile had a different rate of change compared to the reference profile); and 3) a fully adjusted growth model (Model 3), where a comprehensive set of personal and family characteristics were included to test whether neighborhood profile effects remained. Alternative unconditional growth models (e.g., linear, curvilinear, cubic) were compared using log-likelihood ratio tests to determine the best functional form of change.

Multiple imputations were performed using chained equations to address missing data (ranging from 0.2% on race/ethnicity to 8.2% on food insecurity).²⁴ All independent variables and non-missing covariates were used as predictors in the imputation process. Ten complete datasets were imputed to help ensure stable parameters and standard error estimates. The growth curve analyses were conducted using the mi estimate command in Stata Version 15.1.²⁵

Results

As shown in Table 1, participating girls and young women were between 14 and 21 years at baseline (M = 18.16, SD = 1.37 years); 44% were Hispanic, 14% Mixed Hispanic and Black, 37% non-Hispanic Black, and 5% another race/ethnicity. Over half of the participants (56%) had parents or guardians on public assistance. Descriptive statistics for neighborhood-level variables are presented in Table 2.

Table 1.

Descriptive Statistics of Baseline Study Variables

Study Variables	Mean (SD)	Frequency (%)	Min	Max
Age in years	18.16 (1.37)	-	14	21
Race/ethnicity
Hispanic	-	589 (44.45%)	0	1
Mixed Hispanic and Black	-	180 (13.58%)	0	1
Non-Hispanic Black	-	491 (37.06%)	0	1
Non-Hispanic Other	-	65 (4.91%)	0	1
Born in the United States	-	1189 (89.53%)	0	1
Parent receipt of public aids	-	737 (55.88%)	0	1
Food insecurity
Enough to eat	-	1224 (92.45%)	0	1
Sometimes not enough to eat	-	83 (6.27%)	0	1
Often not enough to eat	-	17 (1.28%)	0	1
Family structure
Parents not separated	-	286 (21.83%)	0	1
Parents separated	-	934 (71.30%)	0	1
Parent(s) deceased	-	90 (6.87%)	0	1
School status
Currently attend school	-	1,122 (84.49%)	0	1
Currently not in school	-	203 (15.29%)	0	1
Age of menarche	11.82 (1.55)	-	8	17
Ever had sex with female partner	-	283 (21.54%)	0	1
Changed address during follow-up	-	288 (21.69%)	0	1
Body mass index	25.81 (6.37)	-	13	64.62

Notes. Mean and standard deviation are shown for continuous variables; frequency and percentages are shown for categorical variables.

Table 2.

Descriptive Statistics of Neighborhood-Level Variables (N = 2,081)

Neighborhood Variables	Mean	SD	Median	Min	Max	25th	75th
Neighborhood Structural Characteristics
Percent poverty	19.22	12.67	16.25	0	68.53	9.23	26.37
Median household income (in thousands)	60.93	29.60	56.61	9.74	250	40.59	75.42
Percent completed college	20.16	10.14	18.30	0	59.22	12.83	25.74
Percent Hispanic	26.96	22.61	18.47	0	96.38	9.32	40.67
Percent Black	23.77	29.19	7.96	0	97.78	1.59	39.13
Percent foreign-born	37.76	15.08	37.26	3.49	84.73	25.48	48.31
Percent female-headed household	10.86	8.63	8.52	0	43.33	3.96	15.95
Percent crowded housing	9.59	7.39	7.93	0	57.72	4.04	13.28
Percent unemployed	5.52	2.90	5.03	0	31.82	3.41	7.14
Neighborhood Physical Characteristics
Outdoor recreational resources	13.90	12.11	9.10	0.70	50.20	5.40	19.50
Neighborhood Social Characteristics
Percent disconnected youth	8.16	10.78	4.62	0	100	0	12.91
Neighborhood crime	4.96	7.01	3.39	0	139.72	1.52	6.39
Percent sugary drinks	28.01	7.83	26.30	12	45.60	22	34.10
Percent no fruits and vegetables	12.75	4.58	12.10	4	23	9	15.60
Percent no physical activity	23.42	4.76	22.80	9.7	32.30	21.20	26.60

Notes. SD = standard deviation.

BIC, AIC and LMR test values decreased from the 2-profile to 5-profile solution, and continued to decline, but at a slower rate, from the 5-profile to 9-profile solution, indicating improved model fit until the 5-profile solution (see eTable2). LMR tests showed improved model fit until the 7-profile solution. The entropy values of the 2-profile to 9-profile solutions were all well above .80, which is considered desirable.²¹ In terms of the interpretability, the 5-profile solution produced profiles that were distinct from each other. However, while the 6-profile solution created an additional small profile (6% of census tracts), it did not provide a new neighborhood pattern (i.e., the additional profile was similar to an existing profile in that both were characterized by socioeconomic and housing disadvantage, but overall favorable social environment). Based on the fit statistics, profile sizes, and interpretability, the 5-profile solution was chosen as the best-fitting model. The five classes identified were labeled as: High Structural/High Social Advantage (14% of census tracts), which was characterized by the most favorable structural and social characteristics, despite moderate crime rates and few outdoor recreational resources; Moderate Advantage/Low Crime (38%), which had favorable structural characteristics, the lowest crime rates, and the best access to parks; Low SES/High Activity (27%), which had unfavorable structural (socioeconomic, housing, unemployment) and social (crime and eating norms) conditions, but favorable outdoor recreational resources and physical activity norms; Low SES/High Social Advantage (11%), which had unfavorable socioeconomic conditions, crowded housing, but overall favorable social environment; and High Disadvantage (14%), which had the least favorable structural, physical, and social environment. Neighborhood characteristics across the 12 variables for each profile are summarized in Figure 1. The geographical distribution of the five neighborhood profiles across NYC is shown in Figure 2.

Figure 1.

Descriptive Means of Neighborhood Characteristics by Latent Neighborhood Profiles. Y-axis shows z-score transformed means, with higher scores indicating better environments. (r) represents reverse-coded variables.

Figure 2.

Distribution of Identified Neighborhood Profiles across New York City. Gray areas represent census tracts with no residents or less than 100 residents (e.g., bodies of water, cemeteries, industrial areas, or parkland).

Model 1 showed that the average BMI at age 22 was 28.15 kg/m² (95% CI = [27.62, 28.67], p < 0.001), with an average increase of 0.58 kg/m² per year from age 14 (95% CI = [0.47, 0.69], p < 0.001; Table 3). The quadratic and cubic functional form growth models did not improve the model fit; thus, the linear growth function was used in subsequent models. Model 2 showed that living in Low SES/High Activity neighborhoods was associated with a lower BMI compared to living in High Disadvantage neighborhoods at age 22 (b = −1.21, 95% CI = [−2.35, −0.06], p < 0.05). There was a marginal interaction effect between Low SES/High Activity and age, suggesting that participants living in Low SES/High Activity neighborhoods had a slower rate of growth in BMI by 0.21 kg/m² per year (95% CI = [−0.45, 0.02], p = 0.07), compared to those living in High Disadvantage neighborhoods. These effects remained after adjusting for personal and family characteristics (Model 3). None of the other neighborhood profiles were significantly different from the reference profile. Among the covariates, girls who were born in the US had a higher BMI (b = 2.27, 95% CI = [1.00, 3.55], p < 0.001) than girls who were born outside the US. An earlier age at menarche was associated with a higher BMI (b = −0.68, 95% CI = [−1.03, −0.33], p < 0.001). Girls who experienced food insecurity had a higher BMI than girls who did not (b = 1.53, 95% CI = [0.19, 2.87], p < 0.05). Girls whose parents separated had a lower BMI compared to girls whose parents were not (b = −0.92, 95% CI = [−1.82, −0.02], p < 0.05).

Table 3.

Multilevel Latent Growth Curve Model Predicting Body Mass Index from Neighborhood Profiles

	Model 1	Model 2	Model 3
Predictors	b (95% CI)	b (95% CI)	b (95% CI)
Intercept at 22 years	28.15 (27.62, 28.67) ***	28.85 (28.02, 29.67) ***	26.95 (25.19, 28.71) ***
Linear age	0.58 (0.47, 0.69) ***	0.70 (0.54, 0.87) ***	0.72 (0.55, 0.89) ***
Neighborhood profiles (ref = High Disadvantage)
High Structural/High Social Advantage		−1.55 (−3.91, 0.81)	−1.17 (−3.61, 1.26)
Moderate Advantage/Low Crime		−1.43 (−3.35, 0.48)	−1.63 (−3.59, 0.33)
Low SES/High Activity		−1.21 (−2.35, −0.06) *	−1.32 (−2.49, −0.16) *
Low SES/High Social Advantage		2.70 (−2.15, 7.55)	2.42 (−2.45, 7.28)
Neighborhood profiles X Age
High Structural/High Social Advantage X Age		−0.29 (−0.76, 0.19)	−0.22 (−0.72, 0.29)
Moderate Advantage/Low Crime X Age		−0.25 (−0.65, 0.16)	−0.26 (−0.67, 0.16)
Low SES/High Activity X Age		−0.21 (−0.45, 0.02) †	−0.22 (−0.46, 0.02) †
Low SES/High Social Advantage X Age		0.57 (−0.58, 1.71)	0.56 (−0.59, 1.70)
Race/ethnicity (ref. = Hispanic)
Mixed Hispanic Black			−0.07 (−1.20, 1.05)
Non-Hispanic Black			0.11 (−0.71, 0.92)
Non-Hispanic Other			−1.30 (−3.07, 0.48)
Age of menarche			−0.68 (−1.03, −0.33) ***
Born in US			2.27 (1.00, 3.55) ***
Parent receipt of public aids			0.15 (−0.58, 0.89)
Family structure (ref. = parents not separated)
Parents separated			−0.92 (−1.82, −0.02) *
Parent(s) deceased			−0.61 (−2.08, 0.86)
Food insecurity			1.53 (0.19, 2.87) *
Attend school			0.72 (−0.29, 1.74)
Sex with a female partner			0.20 (−0.66, 1.07)
Changed address during follow-up			−0.49 (−1.37, 0.38)

Notes. SE = standard error; US = United States

^***p < .001

^**p < .01

^*p <.05

^†p < .10

Discussion

Growing evidence suggests that the neighborhood characteristics play an important role in shaping young people’s attitudes, behaviors, and exposure to risks, and are therefore key to understanding early determinants of life-long health.²⁶ Neighborhoods vary greatly in their SES, physical condition, and social norms;¹⁶ we captured this heterogeneity by studying empirically-driven neighborhood profiles in NYC that better represent the intersections of structural, physical, and social factors that naturally occur in real-world settings. We identified five distinct neighborhood profiles in NYC and found evidence suggesting that these neighborhood profiles contribute to BMI trajectories from adolescence to young adulthood beyond personal and family characteristics. This study of young people’s neighborhoods and BMI trajectories not only captures the complexities of neighborhoods in NYC, but also helps to inform understanding of neighborhood context and health disparities in urban areas more broadly.

We found five distinct neighborhood profiles, most of them mixed (e.g., characterized by advantages in some characteristics but not in others), pointing to the complex ways in which neighborhood structural, social, and physical characteristics intersect in the patterning of neighborhoods.²⁷ These neighborhood profiles provide a level of detail not offered by prior variable-centered studies^9,14 and begin to elucidate how multidimensional neighborhood systems can contribute to health conditions during adolescence.

We found that living in Low SES/High Activity neighborhoods was associated with a lower BMI at 22 years, compared to living in High Disadvantage neighborhoods. These findings align with previous literature that has documented associations between neighborhood physical activity resources, including parks, recreation centers, pools and other indoor and outdoor facilities, and healthier weight in adolescents.^9,11,28–30 According to the social disorganization theory, neighborhoods where outdoor recreational facilities are easily accessible may encourage adolescents to engage more actively in exercise, games, and sports, and parents also are more likely to model physical activities themselves and encourage their children to play.^5,31 Our results are consistent with a study that has used LPA to study neighborhood recreation environments, which found that living in neighborhoods characterized by open space had higher physical activity levels and were less likely to be obese than youth living in neighborhoods characterized by less open space.²⁹ Another study of low-income youth found that children living in neighborhoods characterized by density and proximity to food outlets were more likely to be overweight or obese.³² Studies that considered physical as well as structural and social domains were less prevalent, and could provide additional information on how neighborhoods influence adolescent weight status.

We also found a slower rate of change in BMI among participants living in Low SES/High Activity neighborhoods, suggesting an incremental effect of neighborhood environment on adolescent weight status. It is plausible that as adolescents gain more autonomy and increasingly explore physical activity resources around their homes, neighborhood physical activity facilities become critical in fostering healthy physical activity habits that persist into later life.

In addition to the availability of neighborhood outdoor recreational resources, healthy activity norms provided by Low SES/High Activity neighborhoods may also contribute to its supportive and protective effects on BMI. Prior research suggests that positive physical activity norms in the community could provide youth with positive role models and allow faster transmission of healthy messages, conducive to maintaining a healthy weight for adolescent girls.^14,33 It is important to note this study assessed activity norms based on the behaviors of adults within the neighborhood. Given that adolescent peers influence adolescent activity behaviors and weight status,³⁴ future studies may give attention to norms among adolescent members of a neighborhood.

Strengths and Limitations

Strengths of the current study include the large sample size, evaluation of neighborhood characteristics across structural, social, and physical domains, clinician-measured BMI, bi-annual follow-up periods, and consideration of a host of sociodemographic and health covariates associated with adolescent BMI.

Our study has several limitations that need to be considered. First, while residential instability is common in NYC and is more frequently experienced by low-income families,³⁵ the current study used participants home address data at enrollment and did not account for the amount of time that an adolescent had resided in her neighborhood. Although we controlled for whether participant changed address during a follow-up visit to adjust for potential effects of moving, caution is needed when drawing conclusions about neighborhood effects based on exposure at one time point. Second, neighborhood physical environment was measured using outdoor recreational resources (e.g., parks, playgrounds, outdoor pool, etc.) in a census tract; however, it did not account for the availability of other types of physical activity resources near one’s home (e.g., gyms, bicycle trails), and it did not capture the actual utilization of these outdoor recreational resources by participants. Future research could use new procedures such as GPS tracking and community asset mapping to capture the actual experience of the individual within the neighborhood.^36,37 Third, although the participants lived in all five boroughs of the NYC, it is possible that the sample does not fully represent some types of neighborhoods. Despite the limits on generalizability, this sample represents an understudied population of low-income urban racial/ethnic minority female youth at high risk for obesity, and therefore, understanding neighborhoods effects on health within this particular group has substantial public health implications. Additionally, we provide the neighborhood profile data in the supplemental materials, such that future researchers can use these profiles to measure neighborhood effects in other, diverse NYC samples.

Conclusion

This study contributes to the literature on the neighborhood effects on obesity by demonstrating the importance of nuanced configurations of the urban neighborhood environment in understanding adolescent female weight trajectories. Our findings can help inform prevention efforts in promoting healthy weight trajectories in adolescent and young adult female populations. State and local-level policy makers need to evaluate not only structural, but also social and physical conditions of a neighborhood and develop public health policies and programs that support health equity and provide flexibility to better respond to the specific needs and contexts of different communities. When working with adolescents and their families, pediatricians and other clinicians should consider relevant neighborhood resources and incorporate these characteristics in developing actionable health plans. For instance, youth who live in neighborhoods characterized by corner stores and little to no outdoor recreational spaces may need additional guidance on healthy food and exercise options as an alternative to convenient unhealthy options that may be more accessible in their communities. They should also be aware of youth facilities that provide healthy meals, socioemotional support and physical activities where they can refer young people to. Although the availability of physical activity resources has the potential to promote healthy weight, it alone is not sufficient to reduce obesity.³⁸ State and local-level policy makers need to work with the greater community to make sure parks are safe, in good condition and are inviting to youth in order to maximize utilization. Last but not least, strong and effective school-based health, nutrition, and physical activity education programs should be implemented in disadvantaged urban school districts that disproportionately serve racial/ethnic minority and low-income students to address structural inequalities and, ultimately, reduce health disparities.

Implications and Contributions.

In a longitudinal cohort of 1,328 predominantly Hispanic and Black female youth, neighborhoods characterized by availability of more physical activity resources was associated with a healthier weight trajectory during adolescence. Findings suggest that neighborhoods may be important targets for interventions to promote healthy weight and reduce health disparities during adolescence.

List of Abbreviations:

HPV

human papillomavirus

BMI

body mass index

SES

socioeconomic status

NYC

New York City

U.S.

United States