SEARCH

SEARCH BY CITATION

Keywords:

  • safety;
  • residence characteristics;
  • environment;
  • mothers;
  • crime

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

Objective: To test the hypothesis that mothers of young children would have a higher prevalence of obesity if they lived in neighborhoods that they perceived as unsafe or as having a low level of collective efficacy.

Research Methods and Procedures: Using data from the Fragile Families and Child Wellbeing Study, a cross-sectional analysis was conducted of 2445 women living in 20 large (population ≥ 200, 000) U.S. cities. BMI was measured on 72% and self-reported on 28%. Perception of neighborhood safety was assessed with the Neighborhood Environment for Children Rating Scales. The collective efficacy measure was adapted from the Project on Human Development in Chicago Neighborhoods.

Results: Thirty percent of the women were married, 38% lived below the U.S. poverty threshold, and 66% reported no education beyond high school. Approximately one-half of the women were non-Hispanic black, and one-fourth were Hispanic (any race). After adjustment for sociodemographic factors (household income, education, race/ethnicity, age, and marital status), smoking, depression, and television time, the prevalence of obesity (BMI ≥ 30 kg/m2) increased across tertiles of neighborhood safety from safest to least safe (37% vs. 41% vs. 46%, p = 0.004) but did not differ across tertiles of collective efficacy from highest to lowest (41% vs. 40% vs. 42%, p = 0.67).

Discussion: In a national sample of women with young children, obesity was more prevalent among those who perceived their neighborhoods to be unsafe.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

The characteristics of neighborhoods can influence how and where people spend their time (1). For example, certain neighborhood characteristics may lead people to spend less time outdoors. More indoor time may contribute to obesity by decreasing physical activity and by increasing sedentary behaviors and food intake.

People's perception of social disorder and of collective efficacy in their neighborhood may influence how much time they spend outdoors. Social disorder is a measure of neighborhood safety that describes potentially threatening neighborhood conditions and behaviors such as the presence of drug dealing, loitering, and public intoxication (2, 3). Such neighborhood activities may lead people to remain in their homes, thereby constraining their outdoor activity and reducing daily energy expenditure.

Collective efficacy is a measure of people's perception of closeness or connection with their neighbors (social cohesion and trust) and the capacity of neighbors to intervene on behalf of their community to reach common goals (informal social control) (4). Distrust and lack of collective accomplishment are associated with violence and victimization, even after taking into account the economic and demographic characteristics of residents (4). Therefore, low levels of perceived collective efficacy, like low levels of perceived neighborhood safety, might be associated with a tendency for people to remain indoors.

If people do not feel safe in their neighborhoods or connected to their neighbors, this could also increase levels of perceived psychological stress. For some, chronic stress could lead to functional impairment through anxiety or depressed mood. These emotions could increase social isolation and reduce energy expenditure in routine or leisure-time activity, especially outdoors. Perceived stress or symptoms of depression or anxiety can also increase food intake in some individuals (5, 6).

Despite the hypothesized link between the perceived level of neighborhood safety and obesity, we are aware of no studies in adults that have evaluated this association. Using data collected in 20 U.S. cities, we tested the hypotheses that women with young children who lived in neighborhoods that they perceived as unsafe and as having lower levels of collective efficacy would have a higher mean BMI and a higher prevalence of obesity than those who lived in neighborhoods where these characteristics were perceived to be more favorable.

Research Methods and Procedures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

Study Design and Sample

The Fragile Families and Child Wellbeing Study is a birth cohort study following 4898 children and their parents. This report focuses on the mothers. The sample was drawn from births between 1998 and 2000 in 20 large (population ≥ 200, 000) U.S. cities in 15 states at 75 birth hospitals. Because unmarried mothers were oversampled (3:1) relative to married mothers (7), 3712 of the women in the study were unmarried (so-called fragile families) when the baseline survey was conducted at delivery.

Women were ineligible (<5% of sampled births) if their child was placed for adoption, if they did not speak either English or Spanish adequately to understand the survey, or if they were too ill after delivery. Most hospitals did not allow women <18 years of age to participate. Among eligible women, 82% of those married and 87% of those unmarried agreed to participate. The institutional review boards at all birth hospitals, as well as those at Princeton and Columbia Universities, approved the data collection procedures. All participants gave informed written consent.

Approximately 3 years after delivery, 2620 women (53%) participated in an in-home survey during which height and weight measurements were obtained and mothers responded to questions about their neighborhood. The racial/ethnic composition of the mothers was different among those who were followed-up at 3 years and those who were not (non-Hispanic white, 20% vs. 23%; non-Hispanic black, 52% vs. 43%; Hispanic, 25% vs. 29%; other race/ethnicity, 3% vs. 5%; p < 0.001 by overall χ2 test). However, there were no significant differences in baseline income, marital status, age, or education.

Measures

BMI

BMI (kilograms per meter squared) was calculated from the women's measured or self-reported heights and weights. In two of the cities, all heights were reported. In the other 18 cities, heights were measured, and those who refused were asked to report their height. All subjects were weighed unless they were pregnant, exceeded the scale limit of 140 kg, or refused. In each of these situations, the subject was asked to report her current weight (or her prepregnant weight if she was pregnant).

The analytic sample included the 2445 (93%) women with BMI data. Of these, a total of 675 (28%) had reported, rather than measured, height or weight; 180 were pregnant, 385 reported height, 63 reported weight, and 47 reported both.

Neighborhood Safety/Social Disorder

Perception of neighborhood safety was assessed using the Neighborhood Environment for Children Rating Scales (8). This eight-item scale, designed to measure social disorder, asked the women how often they saw events in their neighborhoods such as loitering adults, gang activity, drunks or drug dealers hanging around, and disorderly or misbehaving groups of youths or adults.

The four response options (never, rarely, sometimes, and frequently) were assigned the values 1 through 4, respectively. Internal reliability of the eight items, as measured by Cronbach α, was 0.91. The scale score was computed as the mean value of the responses for all women who completed at least six of the eight items. The scale scores ranged from 1 to 4, with lower scores indicating a higher level of perceived safety. Because the scale scores were not normally distributed, they were divided into tertiles. The safest tertile (referred to as high safety) included all women with a score equal to 1.0 (i.e., answering never to all items). The medium safety tertile scores ranged from 1.1 to 2.0, and the least safe tertile (referred to as low safety) included all women with a score > 2.0.

Collective Efficacy

The measure of collective efficacy was adapted from the 1995 Community Survey of the Project on Human Development in Chicago Neighborhoods (4). This construct consisted of two components: social cohesion and trust, the level of trust and attachment among neighbors; and informal social control, a belief in the capacity of neighborhood residents to intervene to help the community reach collective goals. Each component consisted of five items. For the social cohesion and trust component, the women recorded their level of agreement to statements about their neighbors, such as, “People in this neighborhood can be trusted.” The five response options were strongly agree, somewhat agree, neither agree nor disagree, somewhat disagree, and strongly disagree. The informal social control construct consisted of questions about how likely neighbors were to intervene in certain situations, such as if a fight broke out in front of their house or if children were skipping school. The five response options were very likely, somewhat likely, neither likely nor unlikely, somewhat unlikely, and very unlikely.

The response options for each of the 10 items, ranging from strongly agree to strongly disagree (social cohesion and trust) and very likely to very unlikely (social control), were assigned the values 1 to 5. Internal reliability of the 10 items, as measured by Cronbach α, was 0.85. A collective efficacy score was computed for all mothers who completed at least four of the five items for each of the two components (i.e., at least eight of the 10 total items) by averaging the responses to the completed items. The scale scores ranged from 1 to 5, with a higher number indicating a lower level of collective efficacy. These scores were divided into tertiles to facilitate the interpretation of results. Women in the low-efficacy tertile were those who rated their neighborhoods as having the least collective efficacy (highest scale score), and women in the high efficacy tertile were those who rated their neighborhood as having the most collective efficacy (lowest scale scores).

Covariates

Income was calculated as a ratio of household income divided by the income at the federal poverty threshold for a given household size in the year of reporting (see http:aspe.hhs.govpovertypoverty.shtml). Other sociodemographic covariates included education, race/ethnicity, and current marital status.

Smoking status (yes/no variable) and 12-month history of a major depressive episode [yes/no variable computed from the Composite International Diagnostic Interview-Short Form, version 1.0 (World Health Organization) (9, 10, 11)] were obtained at 3 years. These two conditions are potentially related to both the exposure (neighborhood context) (12, 13) and the outcome (obesity) (14, 15). Women also reported the number of hours (0 to 24) the television (TV)1 was on in the house each day. This variable served as a proxy for TV viewing.

Data Analysis

Obesity was defined as having a BMI ≥ 30 kg/m2. Across the three tertiles of neighborhood safety and collective efficacy, we compared the mean BMI using one-way ANOVA, and we compared the prevalence of obesity using χ2 tests. We then used General Linear Models to compare mean BMI and percentage obese across these tertiles after adjusting for covariates.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

The mean age of the women was 27.6 (±6.0) years. Thirty percent were married, 38% lived in households reporting income below the federal poverty threshold (income-to-poverty ratio < 1.00), and 66% reported no education beyond high school (Table 1). Approximately one-half the women were non-Hispanic black, and one-fourth were Hispanic (any race).

Table 1. . Number* (percentage) of subjects, prevalence of obesity, and mean BMI by level of each covariate
 N (%)Obese (%)BMI ± standard deviation
  • *

    N < 2445 for any covariate if data on indicator were missing.

  • BMI ≥ 30 kg/m2.

  • p for χ2 test for percentage obese and for ANOVA or Student's t test for BMI.

Household income-to-poverty ratio   
 <0.50474 (20)4730.3 ± 7.9
 0.5 to 0.9436 (18)4429.8 ± 7.6
 1.0 to 1.9668 (27)4530.1 ± 7.7
 2.0 to 2.9350 (14)4129.1 ± 7.0
 ≥3.0517 (21)2727.5 ± 6.5
 p <0.001<0.001
Education   
 Less than high school855 (35)4329.7 ± 7.5
 High school degree or equivalent755 (31)4530.1 ± 7.5
 Some college603 (25)4129.4 ± 7.7
 College graduate or more230 (9)2026.2 ± 5.1
 p <0.001<0.001
Race/ethnicity   
 White, non-Hispanic480 (20)2827.4 ± 7.0
 Black, non-Hispanic1279 (52)4730.4 ± 7.8
 Hispanic (any race)598 (25)4129.2 ± 6.8
 Other race, non-Hispanic81 (3)3127.2 ± 6.8
 p <0.001<0.001
Age (years)   
 <2073 (3)2926.6 ± 6.3
 20 to 24872 (35)3829.1 ± 7.5
 25 to 29701 (29)4630.2 ± 7.9
 ≥30799 (33)4129.3 ± 7.1
 p 0.002<0.001
Relationship status   
 Married746 (30)3728.6 ± 6.7
 Cohabiting748 (31)4229.5 ± 7.5
 Single949 (39)4429.9 ± 7.9
 p 0.010.003
Depression   
 Yes474 (19)4730.5 ± 8.5
 No1967 (81)4029.1 ± 7.2
 p 0.0040.001
Smoking   
 Yes632 (26)3929.1 ± 7.8
 No1810 (74)4229.5 ± 7.3
 p 0.290.18

The mean BMI was 29.4 ± 7.5 kg/m2. Forty-one percent had a BMI ≥30 kg/m2, and 9% had a BMI ≥ 40 kg/m2. The women were more likely to be obese and to have higher BMI if they were less educated, unmarried, had lower income, or were Hispanic or non-Hispanic black (Table 1). Women who reported symptoms in the prior 12 months consistent with a major depressive episode had higher mean BMI and were more likely to be obese.

Women who were more educated, married, non-Hispanic white, older, and had higher income were more likely to perceive their neighborhoods as being safer and as having more collective efficacy (Table 2). Women with a depressive episode were more likely to perceive their neighborhoods as less safe and as having lower collective efficacy. The average amount of time (hours per day) the TV was on in the house was higher in obese women (8.4 vs. 9.1; p = 0.009) and was higher when neighborhoods were perceived as being less safe (low-safety to high-safety tertile, 10.3, 8.1, 7.1; p < 0.001) and as having lower collective efficacy (low-efficacy to high-efficacy tertile, 9.5, 8.8, 7.7; p < 0.001).

Table 2. . Percentage of women in each neighborhood safety and collective efficacy tertile by level of covariate*
 Neighborhood safety tertileCollective efficacy tertile
 LowMediumHighLowMediumHigh
  • *

    p for overall χ2 test (sociodemographic indicator by tertile) < 0.001 for all comparisons, except for smoking by collective efficacy tertile, where p = 0.01.

Household income-to-poverty ratio      
 <0.50502822403822
 0.5 to 0.9443026363727
 1.0 to 1.9373726353530
 2.0 to 2.9303733363034
 ≥3.0163846242749
Education      
 Less than high school442828393526
 High school degree or equivalent403723363529
 Some college283834303238
 College graduate or more73954192655
Race/ethnicity      
 White, non-Hispanic174241232948
 Black, non-Hispanic463123383527
 Hispanic (any race)293635333532
 Other race, non-Hispanic283141363034
Age (years)      
 <20482824384022
 20 to 24413623373726
 25 to 29363331343432
 ≥30273439312841
Relationship status      
 Married203842273142
 Cohabiting403327353332
 Single443224393625
Depression      
 Yes453223403525
 No333532333334
Smoking      
 Yes463321353728
 No313534343234

Neighborhood Safety

In bivariate analysis, women had higher mean BMI and a higher prevalence of obesity as the level of perceived neighborhood safety decreased (Table 3). After adjusting for sociodemographic factors (household income, education, race/ethnicity, age, and relationship status as the categorical variables listed in Table 1), obesity prevalence and mean BMI were still significantly associated with neighborhood safety. Across the three tertiles, from the high-safety to the low-safety tertile, the mean BMI went from 29 ± 0.28 to 29 ± 0.26 to 30 ± 0.27 kg/m2 (p = 0.001), and the prevalence of obesity went from 37% to 41% to 46% (p = 0.007). In these multivariate models, the further addition of three variables (smoking, depression, and number of hours per day the TV was on) had minimal impact on the relationship between BMI or obesity and safety (adjusted model in Table 3). After adjustment for all covariates, the prevalence of obesity was 9 percentage points higher among women in the least safe tertile compared with those in the safest tertile. Using an obesity cut-off point of BMI ≥ 40 kg/m2, obesity prevalence still increased significantly from the high-safety to the low-safety tertile (7%, 8%, 11%; p = 0.04) after adjustment for all covariates.

Table 3. . Unadjusted and adjusted* mean BMI and obesity prevalence by neighborhood safety and collective efficacy tertile
 BMI (mean ± standard error)BMI ≥30 kg/m2
 UnadjustedAdjustedUnadjustedAdjusted
  • *

    Adjusted for household income-to-poverty ratio, education, race/ethnicity, age, relationship status, depression, smoking, and number of hours per day TV was on in the house.

  • Overall p for one-way ANOVA.

  • p for overall χ2 test.

Neighborhood safety tertile    
 Low31 ± 0.2830 ± 0.274846
 Medium29 ± 0.2629 ± 0.263941
 High28 ± 0.2529 ± 0.293537
 p<0.0010.002<0.0010.004
Collective efficacy tertile    
 Low30 ± 0.2730 ± 0.264442
 Medium29 ± 0.2729 ± 0.264140
 High29 ± 0.2529 ± 0.273941
 p0.030.590.080.67

Collective Efficacy

In bivariate analysis, women who perceived lower levels of collective efficacy had higher mean BMI and also tended to have a higher prevalence of obesity (Table 3). However, after adjusting for all covariates, there was no clear relationship between collective efficacy and either mean BMI or obesity prevalence. Using an obesity cut-off point of BMI ≥ 40 kg/m2, obesity prevalence was not significantly different from lowest to highest level of collective efficacy (10%, 8%, 8%; p = 0.34) after adjusting for covariates.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

In this national study of over 2400 women with preschool children, those who perceived their neighborhoods as safer had a lower mean BMI and were less likely to be obese. This was true after controlling for multiple measures of socioeconomic status that might potentially confound the relationship between neighborhood safety and obesity. There was no significant association between collective efficacy and either BMI or obesity after adjustment for covariates.

To our knowledge, this is the first study in adults to examine the relationship between obesity and perceived neighborhood safety or collective efficacy. Several studies have explored the relationship between neighborhood safety and physical activity. In a national telephone survey of 1600 adults, neighborhood safety was shown to be a barrier to physical activity, particularly in low-income households (16). In an analysis of data from the 1996 Behavioral Risk Factor Surveillance System in five states, higher levels of perceived neighborhood safety were associated with a lower prevalence of inactivity (17). In qualitative studies, safety has been reported by minority women to be a barrier to physical activity (18).

In contrast to these studies, others have shown that safety and crime are not associated with physical activity levels. In a cross-sectional study of 1800 U.S. adults, subject report of high crime in the neighborhood was not significantly related to physical activity after adjusting for age, sex, race, income, and education (19). In a study of 2900 women 40 years of age and older, neither the perception of high levels of crime nor lack of safe places to exercise was associated with leisure-time physical activity (20).

Although obesity was related to our measure of neighborhood safety, it was not related to our measure of collective efficacy. It is possible that outdoor physical activity is more influenced by the fear associated with social disorder than by a sense of low social cohesion and trust in the neighborhood or a low level of informal social control. In addition, women who perceived their neighborhoods as having low collective efficacy may still have had meaningful social connections with individuals outside of their neighborhood. These connections may have resulted in psychological benefits that offset any psychological distress associated with the perception of low collective efficacy in their neighborhood.

From our cross-sectional study, we cannot infer that low levels of neighborhood safety cause obesity. It is possible, for example, that obese women may be more inclined to perceive their neighborhood as unsafe because they are subjected to higher levels of discrimination (21). We have no measurement of physical activity to evaluate whether reduced physical activity mediates the relationship between safety and obesity. The hours per day that the TV was on in the household is only a proxy measure for TV viewing time, which is, in turn, not necessarily correlated with physical activity levels.

We also had no information on diet. Neighborhoods with low levels of safety may also have a lower availability of foods that could reduce the risk of obesity (22, 23). However, we did control for individual-level indicators of race/ethnicity and social class that may be related to both diet and neighborhood safety. It is also possible that neighborhood safety affects obesity by other factors, such as psychological stress, that influence both energy intake and expenditure. Although we had no measure of perceived stress, chronic stress may contribute to depression, which was included in our analysis but which did not seem to mediate the BMI-safety relationship.

Although multiple mechanisms are likely to be involved in the association between neighborhood safety and obesity, the one most often proposed involves the effect of safety on physical activity. Therefore, the lack of data on physical activity in this study is an important limitation that should be addressed in future research.

Theoretically, an experiment in which neighborhoods were randomly assigned to an intervention that improves safety and/or the perception of safety would be the strongest design to test causality. In the Moving to Opportunity study, low-income families were randomly assigned to receive housing vouchers that could be used to move to neighborhoods that varied in the percentage of residents living in poverty (24). Women who received vouchers to obtain housing in higher income neighborhoods had a significantly lower prevalence of obesity at follow-up. Although the neighborhood safety levels were not measured, women in the higher income neighborhoods did have lower levels of anxiety and depression, suggesting the possibility that psychological stress may be part of a causal pathway between neighborhood safety and obesity. Adding measures of safety perception, psychological stress, physical activity, sedentary time, and dietary intake to experiments or quasi-experiments of neighborhood mobility would help increase our understanding of the effect of the neighborhood environment on obesity.

Although the neighborhood context is often mentioned as an important factor influencing adult physical activity levels and obesity (25, 26), this is the first study, to our knowledge, to find an association between perceived safety and BMI. This finding has particular implications for the obesity epidemic because our study focused on women of childbearing age who have young preschool children. Many women were living in low-income households. Not only do low-income minority women experience higher rates of obesity (27, 28), but many are single mothers who are important role models and facilitators of their children's activity levels. Thus, future studies should not only assess how perceived safety influences adult physical activity and obesity, but they should also attempt to evaluate what aspects of perceived neighborhood safety influence children's outdoor play.

Acknowledgement

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

This work was supported by NIH Grants R01-HD41141 (R.C.W.) and K24-DK065018 (T.A.W.).

Footnotes
  • 1

    Nonstandard abbreviation: TV, television.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References