Data Collection

The Trial of Activity in Adolescent Girls (TAAG) study is a multicenter group-randomized trial designed to test an intervention to reduce the usual decline in moderate to vigorous physical activity in middle-school girls [18]. The present study was conducted to provide psychometric properties for the instruments to be used in the TAAG study, which was sponsored by the National Heart, Lung, and Blood Institute. Adolescent girls in the sixth and eighth grades were recruited from schools in or near the six TAAG field centers: Baltimore, MD; Columbia, SC; New Orleans, LA; Minneapolis, MN; San Diego, CA; and Tucson, AZ. Data collectors participated in a centralized training to ensure that standardized procedures, scripts, and protocols were used. Girls were recruited while attending required classes at their respective public schools. All participating schools were public schools, one per site, with a mean total enrollment of 872 students (standard deviation 448). The proportion of students receiving free or reduced price lunch varied across schools, from 31.8% to 89.7% (weighted mean 43.5%, standard deviation 12.8%).

Students completed the self-administered questionnaire at school, supervised by the data collectors during February-April 2002. A standardized introduction to the survey was read, and data collectors were available for questions. Readministration of the survey was conducted approximately 2 weeks later (median 12 days, range 6–24 days), using the same procedures to examine reliability of the instrument. Parents or guardians provided written informed consent prior to survey administration. The girls also provided written assent at the time of survey administration. This study was approved by the Institutional Review Board at each field center and at the coordinating center (University of North Carolina – Chapel Hill).

Survey Instrument

Due to the lack of a developed written questionnaire appropriate for use with adolescents to assess physical environmental factors and transportation in detail, a new questionnaire was developed. Some of the questions were modified from existing questionnaires designed for adults, such as the Amherst Study parent questionnaire [19], the Neighborhood Environment Walkability Scale [20], and a recreational environmental scale [21]. We chose a priori to analyze each of the items separately. Although we were able to group the items into broad domains, we thought it important to describe the individual items (e.g., physical environmental factors and transportation) and their associations with the outcomes separately. In contrast to attitudinal variables, there was not an expectation that the environmental items grouped together would be interrelated, even though they helped define the same construct.

Statistical Analysis

To examine test-retest reliability of the questions, percent agreement (calculated as the number of response pairs agreeing exactly divided by the total number of response pairs) as well as unweighted (2-level) and weighted (3–5-level) kappa coefficients for categorical variables were calculated overall and separately by grade. For continuous variables, intraclass correlations (ICCs) were calculated, using PROC MIXED treating the individual as the repeated measure [25]. The ICCs were recalculated treating site as a random effect, which did not meaningfully change the results and therefore they are not presented. As a rough guide, we followed the ratings suggested by Landis and Koch for agreement [26]: 0–0.2 poor, 0.2–0.4 fair, 0.4–0.6 moderate, 0.6–0.8 substantial, and 0.8–1.0 almost perfect. For the composite measures, Cronbach alpha coefficients were calculated to indicate internal consistency of the items comprising the scales (e.g., how well the items related to one another).

Associations with physical activity and active transport to school were examined using mixed model logistic regression to account for dependence in the data among girls attending the same school. The logistic models were calculated with two outcomes: (1) physical activity score split at the median (equal to or above the median versus below) and (2) biking, walking, or skating to school (1–5 days versus no days). All statistical models were adjusted for grade (sixth versus eighth grade), race/ethnicity (white versus nonwhite), and site. Site, which also represented school (because there was only one school per site), was treated as a random effect in the model using five dummy variables (six sites). In order to explore the relative importance of these variables, we selected the statistically significant (p < 0.05) physical environmental and transportation variables (shown in Tables 4 and 5) and checked for collinearity among these variables. The only variable we did not explore was the continuous measure of the number of facilities near their home, since this contributed to the measure that assessed the number of facilities near their home plus the number of facilities they would go to. None of the variables were collinear, using the guide of proportion of variation being greater than 50% between any two covariates. These variables were then added to a full model, also including grade, race/ethnicity, and site, and the physical environmental and transportation variables were dropped one by one until only significant variables (p < 0.05) remained.

Analyses were conducted using SAS version 8.2 (Cary, NC), and all models were fit using PROC MIXED [25]. To estimate odds ratios the GLIMMIX macro was used, which implements a generalized linear mixed model specifying the error distribution as binomial and the link as logit [27]. It should be noted that because of the common occurrence of our outcomes due to the way they were categorized, the odds ratios we present are likely an overestimation of the relative risk [28].

Test-Retest Reliability

The median physical activity score was 3.0 (interquartile range 2.4–3.6, range 1–5). The test-retest reliability of the physical activity score was 0.72 (Table 2), with no meaningful differences by grade. The alpha was 0.61 overall, 0.58 for sixth-graders, and 0.59 for eighth-graders. Overall, 42.3% of girls reported any active transport to school, ranging from 13.3% in Minneapolis to 54.3% in Baltimore. Overall, 4.9% reported 1 day of active transport to school in the past week, 7.2% reported 2–3 days, 2.5% reported 4 days, and 15.1% reported everyday. The test-retest reliability of the number of days of active transport to school was 0.60 with no differences by grade.

Associations with Physical Activity and Active Transport to School

Safety of Environment

Neighborhoods with signs of physical disorder (e.g., graffiti, vandalism) and social disorder (e.g., loitering, drug use) may discourage individuals from engaging in outdoor activities [31, 32]. Concerns for neighborhood safety may be reported more frequently for girls than boys [11]. Our study found that agreeing it was safe to walk or jog in the neighborhood was associated with higher levels of physical activity. Findings in the literature of the effect of safety on physical activity of youth have been mixed. Although perception of safety was associated with increased physical activity in some studies [33], a lack of association was reported in other works [19, 29, 34–40] and with decreased physical activity in at least one study [41]. However, studies using objective measures of crime identify unsafe neighborhoods as being associated with less physical activity [42–44]. For example, in a smaller study conducted in Texas, the density of violent crimes within 0.5 mile of the homes of mostly Mexican American seventh-graders was associated with less outdoor physical activity among girls but not boys [42]. In this referenced study, girl's perception of safety was associated with higher levels of outdoor physical activity.

Our study also found a lack of association of most safety variables with active transport to school. We did find that agreeing that "walkers and bikers on the streets in my neighborhood can easily be seen by people in their homes" was associated with less report of active transport to school, rather than more. A 2004 U.S. survey of households with children 5–18 years old indicated that parents reported traffic and crime as barriers to walking and biking to school [45]. Safety has also been reported as an important barrier to children walking and bicycling to school in other U.S. studies [46] and in other countries, such as New Zealand [47], England [48], Canada [49], and Australia [50–54]. In particular, three quantitative studies examined the relationship between environmental factors and active transport to school. First, a study in England [48] found that children of parents who worried about their children becoming lost or being abducted were less likely to walk to school. Second, a study in Australia found that parental perceptions of the neighborhood were associated with walking and cycling to school among 10–12-year-olds [51, 54]. Another Australian study of adolescent youth found that walking to school was negatively associated with parental concern about traffic but positively associated with having friends living in the neighborhood for girls but not for boys [52]. The discrepancy between our findings and these reports could be due to the contrast between parental perception, which we did not measure, compared to the perception of youth, which we measured. It may be that parents' perceptions of safety are more important than those of youth in determining active transport to school. In fact, the Australian study [51] found that there were some discrepancies between parental and child report of perceptions of traffic and safety and that indeed the parental perceptions might be more important.

Aesthetics of Environment

Aesthetic features, such as trees, things to look at (e.g., better scenery), and garbage were not associated with increased walking/cycling to school but were associated with physical activity. This is supported in a study of Portugese youth in 7th-12th grades, where having many interesting things to look at (this question was similarly worded to our question) was a significant positive predictor of physical activity [29]. Furthermore, we found that unpleasant smells in the neighborhood was not associated with overall physical activity but was associated with higher reports of active transport to school. Girls who actively travel to school on foot or by bicycle may be more able to notice odors in the environment than they would if they were traveling by automobile or bus.

Facilities Near Home

According to a review by Sallis et al. [6], having equipment and supplies available for activity was positively associated with physical activity in only one of five studies of adolescents. We found that girls were more than twice as likely to be more physically active when they reported enough sports equipment at home. Differences between studies may be due to the differing measures of access to fitness/sports equipment, as well as differing measures of physical activity. Some studies may have also lacked sufficient statistical power to detect an association.

An environment rich in physical activity resources could remain underutilized unless consideration is given to how accessible the facilities are to the population being served. Again, according to the review by Sallis et al. [6], having access to facilities and programs for activity was positively associated with physical activity in three of four studies of children 4–12 years of age. Since then, several other studies have examined associations with community facilities. A study of youth in grades 1–12 found no association of a single item measure assessing access to parks, playgrounds, or gyms with objectively measured physical activity [19], whereas in a study of Portugese youth in 7th-12th grades, a single item measure of access to physical activity facilities was associated with self-reported physical activity [29]. In a study of 87 adolescent girls, the number of self-reported community facilities was correlated positively with cardiorespiratory fitness but not with physical activity [55]. In a study of mostly Mexican American seventh-graders, euclidean distance from their home to the nearest public play area (i.e., playgrounds, swimming pools, fields) was associated with outdoor physical activity among boys but not girls [42]. A national U.S. study of 7th-12th graders found that presence of a physical activity facility within a given census block group was associated with self-reported physical activity [56]. Furthermore, a study from Australia found that youth 10–12 years old who perceived poor access to parks in their neighborhood were less likely to walk or cycle to school [51]. However, a prospective study of high school girls did not find an association between a 3-item lack of physical activity resource measure and changes in self-reported physical activity [57].

In the present study, girls who reported more physical activity facilities was associated with higher levels of physical activity and active transport to school. This could be the case since girls who are active and walk and bicycle through their environment may be more aware of their surroundings and thus, the physical activity facilities that exist. This was also similar for girls reporting biking or walking trails in their neighborhood. This finding has important implications for further study. It is not clear whether the girls' perceptions of facilities matched what actually existed in their neighborhoods. Studies comparing objective measures of facilities to perceptions can help further address this.

Transportation

Having parents who provide transportation, in general or specifically to physical activity destinations, was positively associated with physical activity behaviors in some studies but not all [35, 37, 58, 59]. Several of these studies showed gender differences in this finding [35, 58]. Hoefer et al. [13] found that parental transportation was associated with physical activity and that parental transportation may be more important for girls than boys. This was also supported in a national parental survey of youth 9–13 years of age, where transportation problems were reported more often as a barrier for girls than boys [11]. We found that having parents who allowed girls to walk in the neighborhood on their own, use public transportation on their own, or could walk/bike to transit from their home was associated with higher levels of physical activity.

Limitations

The work reported here provides support for continuing to explore the role of the physical environment and transportation on girls' physical activity and active transport to school. This study is limited by several factors. First, there may have been other important confounders that we did not account for in the adjusted models for which we did not have measures, such as socioeconomic status, number of children in the family, and parents' and siblings' physical activity. Second, our self-report measures are subject to social desirability bias, which may have affected both the independent and dependent variables [60]. Future studies exploring this should consider using objective measures in addition to self-report measures. Third, despite the short time between administrations, true changes in physical activity could have occurred between the two surveys, which would affect our reliability estimates and could affect our ability to assess associations with physical activity and active transport to school [61, 62]. If this did occur, the true agreement could be higher than we reported but would also be attenuated due to measurement error. It is unlikely that the neighborhood measures changed in this short time period. Fourth, this study is cross-sectional in design, so the direction of the relationships is not known. Fifth, in this study we have tested many associations, but chose not to adjust for multiple testing because we considered this study exploratory. Therefore, significance should be interpreted with caution and replication of results is needed. Lastly, this study explored factors related to active transport to school, but not active transport from school. Results from qualitative research suggest that in some cases barriers to active travel from school may differ from those going to school [46].