Are Perceived Neighborhood Hazards a Barrier to Physical Activity in Children? FREE

LOWER socioeconomic status (SES) is associated with lower levels of physical activity, more television viewing, and an increased amount of body fat in children and adolescents.^{1 - 4} In part, this association may be due to the comparatively hazardous physical environment in which many lower-income families live.^{2 ,5 - 7} For example, hazards that might serve as barriers to physical activity and general health include lack of recreational facilities or athletic programs and the presence of crime, gangs, traffic, drugs, noise, or prejudice.^{8 - 11} In our study we explore the subjective perception of neighborhood hazards that might pose barriers to physical activity in a sample of children.

Higher levels of neighborhood hazards have been linked to reports of less physical activity in 2 studies of middle-class adults of primarily European American descent.^{10 - 11} However, few studies have explored the relationship between physical environmental factors and physical fitness in children. One study of chronically ill children reported that activity patterns were directly related to child and parent perceptions of neighborhood safety.⁹ The generalizability of this study is unclear. Parents of chronically ill children may be more attuned to environmental dangers and may be more likely to limit their children's opportunities for physical activity.

Despite the potential for increased hazards in communities of low SES, the association between neighborhood hazards and physical activity among healthy children has not been examined. We looked for evidence of this relationship in a sample of children with diverse ethnic and economic backgrounds. We hypothesized that children who perceived more neighborhood hazards would be less physically active, less physically fit, and have a higher body mass index (BMI).

All fourth-grade students (N = 845) enrolled in 8 northern California elementary schools were eligible to participate in the study. Of the total eligible children, 796 (94%) participated. A passive-consent procedure was used; 12 parents refused to allow their children to participate, and other eligible children not included were absent or unavailable on assessment days. Parent or guardian telephone interviews were attempted for all eligible children, and 518 parents or guardians completed interviews. The study was approved by the Stanford University Committee for the Protection of Human Subjects in Research, Stanford, Calif.

Assessments were carried out by trained project staff. To ensure confidentiality, students were assigned a special identification number that was used for tracking. Each survey contained 2 cover sheets, the first including a printed label with the student's name and identification number and the second including a label containing only the identification number. Data collectors removed and destroyed the first cover sheet when they collected the survey. Children completed questionnaires during a regular class period. The survey was read aloud to the entire class and took approximately 45 minutes to complete. Surveys were prepared with facing pages written in English and Spanish or English and Vietnamese. Children were given the option to have the survey read aloud to them in English, Spanish, or Vietnamese; 88% of the children completed the survey in English, 9% in Spanish, and 3% in Vietnamese. All physical measures of participating children were obtained at stations set up in the classroom or at a nearby outdoor area.

All parent interviews were conducted on the telephone by trained project staff. The primary target adult in the household was the mother. Most interviews required 10 to 15 minutes for completion. Parent interviews were conducted in English (33.6%), Spanish (43.9%), and Vietnamese (22.5%). Active consent was obtained by telephone and recorded. All parent and child assessments were completed within the same 2-month period.

Children self-reported their sex and date of birth. School district data provided pan-ethnic labels for all children, such as Latino, that combined several specific ethnic groups within 1 larger grouping. Parents reported their own sex and specific ethnic label.

Parents self-reported education levels and occupations for themselves and another parent or guardian living in the house. In this sample, parent occupation was chosen as an indicator of SES instead of parent education level or a combination of the two. Parent education level was not used as an indicator of SES because of the high percentage of immigrants in the sample, who may have been educated in another country but whose income in the United States may not reflect their education level. The higher occupation level was chosen if more than 1 parent or guardian occupation was reported. Parent occupation was coded into the 7 Hollingshead¹² categories and then dichotomized into lower (1-4) and higher (5-9) SES levels based on the midpoint of the scale.

Child acculturation was determined based on language preference for English or another language when at home, with friends, and when watching television. These items were adapted from the language use subscale of the Bidimensional Acculturation Scale for Hispanics.¹³ We acknowledge that the acculturation process is more complex than what language use can represent; however, measurement of acculturation based on language has become a common shorthand method that typically accounts for more than half the variance of acculturation measures. Language use has been found to be the primary determinant of acculturation. English usage and other language usage items were averaged separately, and midpoint cutoff scores were used to dichotomize these measures into high and low categories. Four acculturation categories, as described by Cuéllar et al,¹⁴ were determined based on the following combinations: traditional (high use of another language/low English use), marginalized (low use of another language/low English use), assimilated (high English use/low use of another language), and bicultural (high use of another language/high English use).

Children's perceived neighborhood hazards were assessed with 8 self-reported items based on a 3-point Likert-type scale: 1 indicated "not a problem," 2 indicated "a little problem," and 3 indicated "a big problem." This scale was adapted from the Hazards Scale developed by Aneshensel and Sucoff.⁸ All of the items used the beginning phrase "How much of a problem is . . . ?" The 8 items included the following neighborhood hazards: traffic, trash and litter, crime, too much noise, gangs, lack of access to parks, prejudice, and drugs. All 8 items were summed for a total score with a possible range of 8 to 24, with a higher score indicating the perception of more neighborhood hazards. Questions were chosen to reflect dangers that would be most relevant to a younger population and to outdoor physical activity. The original scale has previously demonstrated significant associations with lower SES and a variety of measures of mental well-being.⁸ In our sample, the internal consistency of this scale was found to be α = .76.

Self-reported physical activity was assessed with a modified version of the Self-administered Physical Activity Checklist (SAPAC) developed by Sallis et al¹⁵ for the National Heart, Lung, and Blood Institute's Child and Adolescent Trial for Cardiovascular Health project. The SAPAC has been demonstrated to be a reliable and comprehensive measure of physical activity in children. In the original SAPAC, children reported the number of minutes they participated in 21 common physical activities before, during, and after school on the previous day. Because of pretesting, we modified the SAPAC to include only after-school activities, added 2 more common activities (for a total of 23), and simplified the response to be a forced choice of "none," "less than 10 minutes," or "more than 10 minutes." Children reported their previous day's physical activity on 2 different days, and these 2 reports were averaged. These simplified response options have previously resulted in a high percentage of agreement (86%) with direct observation ratings of children's physical activity.¹⁶

The maximal multistage 20-m shuttle run test of physical fitness was used to assess cardiorespiratory fitness.^{17 - 18} In varied samples of children and adolescents, this test has been found to be a valid measure of fitness compared with maximum oxygen consumption as measured by treadmill testing (r = 0.69-0.87), and is sensitive to change.^{19 - 23}

Standing height was measured to the nearest millimeter using a portable direct-reading stadiometer. Students were measured with their shoes removed and with the body positioned so that the head, heels, and buttocks were against the vertical support of the stadiometer. The head was aligned so that the auditory canal and the lower rim of the orbit were in a horizontal plane. Two measures of height were obtained, and the average was used in the data analyses.

Body weight was determined to the nearest 0.1 kg using digital scales, with the subjects wearing light indoor clothing without shoes or coats. Two measures of weight were obtained, and the average was used in the data analyses.

Body mass index was computed from the formula kg/m², which is generally considered to be the preferred index of relative body weight as a reflection of adiposity.^{24 - 25}

Sample characteristics are shown in Table 1 and Table 2 for children and parents, respectively. The mean ± SD age of participating fourth-grade children was 9.0 ± 0.37 years. Differences between the sexes were found for the measure of physical fitness (t₂₃₄ = − 4.18; P<.001); boys ran more laps than girls (mean ± SD, 17.61 ± 11.20 laps and 14.66 ± 7.58 laps, respectively). There were no significant differences between boys and girls for the perception of neighborhood hazards, self-reported physical activity, height, weight, or BMI.

One-way analysis of variance was conducted to examine differences by child acculturation, SES level, and ethnicity. There were no significant differences in neighborhood hazards, reported physical activity, or physical fitness by child acculturation level. There were no significant differences in physical fitness or reported physical activity by SES level, but there were significant differences in neighborhood hazards by this measure (F₁₃₉₈ = 4.26; P = .04). Children of lower SES reported more neighborhood hazards (mean ± SD, 13.51 ± 3.83) than children of higher SES (mean ± SD, 12.73 ± 3.48).

There were no significant ethnic differences in perceptions of neighborhood hazards. There were significant ethnic differences for reported physical activity levels (F₄₆₉₇ = 5.04; P<.001) and physical fitness (F₄₇₁₉ = 3.26; P = .01). Latinos reported significantly higher rates of physical activity than Asians (mean ± SD, 8.67 ± 4.61 and 7.19 ± 4.39, respectively), and Latinos ran significantly more total laps than Asians (mean ± SD, 17.21 ± 10.65 laps and 14.45 ± 7.36 laps, respectively).

Spearman correlations among study variables are presented in Table 3. As expected, children of lower SES reported more neighborhood hazards. Contrary to our hypothesis, the perception of more neighborhood hazards was positively correlated with reported physical activity. Although there was a slight positive association between self-reported physical activity and BMI, amount of body fat as measured by BMI was significantly negatively associated with physical fitness.

Children in the same school are more likely to live in similar neighborhoods. Therefore, we conducted analyses to examine whether school effects masked an association between neighborhood hazards and physical activity. First, 1-way analysis of variance and χ² tests were conducted to examine differences in means for variables of interest by school. There were no significant differences by school for perception of neighborhood hazards, self-reported physical activity, or physical fitness. School differences were found for ethnicity (χ²₁₄ = 85.84; P<.001), SES level (χ²₇ = 46.35; P<.001), and BMI (F₇₇₃₇ = 2.58; R² = 0.02; P = .01). Additionally, we investigated intraclass correlations with physical activity using mixed models with compound symmetry as the covariance structure. All intraclass correlations were close to 0; thus, the original Spearman correlations are reported.

Second, a test was conducted to determine homogeneity between schools and the correlations between neighborhood hazards and physical activity, physical fitness, and BMI. Results indicated no significant heterogeneity between schools (P>.05 for all tests). This test was conducted for Spearman correlations between neighborhood hazards and physical activity (pooled r = 0.14), neighborhood hazards and physical fitness (pooled r = − 0.01), and neighborhood hazards and BMI (pooled r = − 0.01), which are nearly identical with those in Table 3.

Neighborhood hazards and SES were negatively correlated (r = − 0.13; P = .01) (Table 3). Therefore, all further correlations were conducted separately by lower and higher SES levels (Table 4). For both SES levels, physical fitness and BMI were inversely correlated, as expected. There was no significant association between neighborhood hazards and reported physical activity or neighborhood hazards and physical fitness for the lower SES group. A significant but low negative correlation was found for neighborhood hazards and BMI for children of lower SES; a higher BMI was associated with the perception of fewer neighborhood hazards. For children of higher SES, the perception of more neighborhood hazards was associated with more reported physical activity. Neighborhood hazards were not significantly associated with physical fitness or BMI for children of higher SES.

Although children from families of lower SES reported more neighborhood hazards than their peers of higher SES, the hypothesis that the perception of more hazards would be associated with lower levels of physical activity and fitness was not supported. Indeed, several of our findings contradicted study hypotheses. Among children of lower SES, those with the perception of more neighborhood hazards had significantly lower levels of body fat. Among children of higher SES, those with the perception of more neighborhood hazards reported higher levels of physical activity.

Several caveats must be considered when interpreting the results of this study. First, the measure of neighborhood hazards used in this study may have lacked sensitivity because it did not directly assess the degree to which children perceived that the hazard was a barrier to their engaging in physical activity. Additionally, the measure did not assess perceived fear of the hazard; rather, participants rated the degree to which each potential hazard was a problem. Thus, whereas children may perceive that drug dealers in their neighborhood are problematic, they may not feel that such a problem poses a direct barrier to playing outdoors. This perception may vary by SES.

Second, the neighborhood hazards scale consisted of 8 potential barriers to physical activity, ranging from crime to noise. Other potential barriers to physical activity may exist in neighborhoods of lower SES that are not included in the 8-item scale. For example, researchers have discussed the effects of larger sociocultural issues on the health habits of lower-income families.²⁶ It may be fruitful to assess competing time demands faced by lower-income children, such as baby-sitting or other family responsibilities. Moreover, the instrument did not include questions about the cost and quality of available locations for physical activity or organized sports. Future studies exploring the relationship between environmental barriers and physical activity in neighborhoods of lower SES should include assessments of cost and quality of accessible facilities for physical activity or organized teams.

Third, parents' perceptions of neighborhood hazards may affect the regulation of their children's opportunities for physical activity. This may explain the finding that children with higher BMI scores were more likely to perceive fewer neighborhood hazards; these children may not be as aware of neighborhood problems as their more active counterparts because of parental regulation of activity. Parents' perceptions of neighborhood hazards should be assessed when the relationship between perceived neighborhood hazards and a healthful lifestyle is studied in young children.

Fourth, the assessment of physical activity in children is difficult at best. The fact that the self-reporting of physical activity was not significantly correlated with physical fitness, whereas physical fitness and BMI were highly correlated, suggests that the self-report measure may contain significant measurement error. Problems with self-reported physical activity measures have been amply documented. Previous research has found poor to moderate correlation between self-reports of physical activity and accelerometer data in children.²⁷ In reviews of the literature, the validity of self-report measures for children varies widely.^{28 - 29} It is possible that a demand bias exists in these measures; children are aware that they should be exercising more, and heavier children may be even more aware of this fact. The 1-day recall format, used here in the modified SAPAC, has typically demonstrated the highest correlations with direct observations. However, our results suggest that the instrument may be problematic for many assessment situations.

Our findings are not consistent with other research reports in this area,^{10 - 11} but our sample demographics were also different from those of previous studies. Whereas we did not find any ethnic differences in perception of hazards, children of lower SES reported more hazards than those of higher SES. Moreover, the significant relationships between hazards and physical fitness variables were different by SES level of children. Although there is not enough information in our study to interpret the meaning behind these differences, it appears that the relationship between SES and hazards is more complex than initially proposed.

In summary, this study provides no evidence to support the common assertion that children from neighborhoods of lower SES do not engage in physical activity or are any less physically fit because of neighborhood dangers such as crime or gangs. Although our findings show that children in neighborhoods of lower SES are more likely to perceive environmental dangers, these hazards do not appear to present direct barriers to their engaging in physical activity. Children from lower-income homes have been found to have higher rates of sedentary behavior and television watching, which may provide more insight into the link between lower income and low physical activity. We suggest that future studies in this area examine other potential barriers to physical activity for lower-income children by including assessments of sedentary behavior, parental fear of violence, parental regulation of children's leisure activities, and the cost and quality of available locations for physical activity and organized sports.

Accepted for publication April 26, 2001.

Funding for this research was provided by grant RO1 CA68082 from the National Cancer Institute, Bethesda, Md.

Research has reported that less physical activity is associated with higher levels of neighborhood hazards in populations of middle-class adults of European descent. Despite the potential for increased hazards in communities of low SES, the association between neighborhood hazards and physical activity in children has not been examined.

Children of lower SES report higher levels of neighborhood hazards. However, contrary to previous research, more hazards were associated with more physical activity in our results. This finding was not further explained by school or SES heterogeneity. The implication is that the relationship between neighborhood context and physical activity is complex, and future studies should include assessments of sedentary behavior, parental fear of violence, parental regulation of children's leisure activities, cost and quality of play areas, and availability of organized sports.

Corresponding author and reprints: Andrea Romero, Mexican American Studies and Research Center, University of Arizona, Economics Bldg, Rm 208, Tucson, AZ 85721 (e-mail: romeroa@u.arizona.edu).