BALANCE was a group-randomized, nested-cohort study developed and designed in partnership with Parents As Teachers (PAT), a national parenting and child development program (20). We recruited postpartum adolescents who retained their pregnancy weight to participate in the BALANCE weight-reduction protocol. We used data from BALANCE baseline assessments that participants completed between January 2007 and April 2008. As part of our BALANCE study during 2008-2009, we collected school wellness policies from the websites of schools or school districts attended by our participants. If the policy was unavailable on the website, we contacted the school and requested a copy. We also verified that collected policies were in effect in 2006-2007.

We recruited 1,330 ethnically diverse participants into BALANCE who were enrolled in PAT Teen Parent Programs from 27 states (Appendix A). In addition to enrollment in the PAT Teen Parent Program (for ages 13-19 y), eligibility criteria included 1) a willingness to participate throughout the study period, 2) being less than 1 year postpartum, and 3) not being pregnant or planning to become pregnant during the study period. For our analysis, we further excluded participants who had either graduated or withdrawn from school (n = 275), were currently breastfeeding (n = 109), or were missing residential zip code and school information (n = 299). In total, 647 postpartum adolescents located in 251 schools from 203 school districts in 27 states, contributed to our findings. The institutional review board of Washington University in St. Louis approved this study.

Our sample had a mean age of 17.2 (standard deviation 1.1 y). Forty-eight percent were white, 30% were black, and 22% were other; most were receiving some form of aid from either WIC (91%) or the federally sponsored free or reduced-lunch program (40%), and they were approximately 6 months postpartum (182 days). Approximately half of participants were at a normal weight and half were overweight or obese.

Participants' height and weight were collected by trained PAT staff to determine BMI z score classification according to criteria specified for adolescents by National Health and Nutrition Examination Survey procedures (21).

Adolescents then completed the online Snack and Beverage Food Frequency Questionnaire (SBFFQ), which was used to measure specific high-calorie snack and beverage consumption patterns of participants. Following a similar format to that of the Diet History Questionnaire (22), the SBFFQ examined each participant's intake of 31 items during the previous 7 days by asking on how many days, how many times per day, and how much of the item the participant consumed. Food items were assessed by subgroups: sweetened beverages (eg, soda and fruit juice), salty snacks (eg, potato chips), sweet snacks (eg, hard candy), meal-type snacks (eg, french fries), fruits and vegetables, and water consumption. Intake was converted into the total calories consumed for each individual food item and summed to obtain the daily calorie total. The test–retest reliability for the separate measures ranged from moderate to substantial with the following intraclass correlation coefficients: water (.71), sweetened beverages (.68), salty snacks (.43), meal-type snacks (.64), and fruits and vegetables (.46) (23). The test–retest reliability for the composite measure of total calories was acceptable (.63).

Physical activity was measured with 3 items asking participants how many minutes they spent walking at a slow, brisk, or very brisk pace on the 2 weekdays preceding completion of the measure, and on 1 weekend day (24). Participants reported their age, race/ethnicity, education level, breastfeeding status, and postpartum status. They also reported their participation in aid programs (WIC and the National School Lunch Program), which we used as indicators of socioeconomic status.

We used the 96-item School Wellness Policy Coding Tool developed by Schwartz and colleagues (18) to assess the strength and comprehensiveness of the school wellness policies in each school district (Appendix B). Each of the 96 content items was coded with a score of 0, if the item was not mentioned; 1, if the item was a "weak" statement making it hard to enforce because of vague, unclear, or confusing language; or 2, meaning the item "meets or exceeds expectations" since it was mentioned in a specific and directive manner suggesting commitment to enforcement (Table 1).

Data analyses were conducted in 2 stages. First, we sought to determine the comprehensiveness and strength scores of school wellness policy language for school districts attended by our participants. Second, we sought to relate the overall comprehensiveness and strength of school wellness policy language to the measured energy-balance behaviors and BMI z scores of BALANCE participants. All analyses were conducted by using SPSS version 17.0 (SPSS, Inc, Chicago, Illinois).

We evaluated the language quality for each policy item of the coding tool by the percentage of school districts with a rating of "meets or exceeds expectations." For assessing the language quality of the 7 policy sections and the overall district policy score, we computed the sample mean and standard deviation for both comprehensiveness and strength with methods suggested by Schwartz and colleagues (18).

The school wellness policy language scores for both comprehensiveness and strength were split into low, middle, and upper tertiles. We compared demographic characteristics of BALANCE participants among school wellness policy language tertiles with χ², Kruskal-Wallis, or 1-way analysis of variance tests, as indicated by measurement level. Univariate, general linear models were constructed to assess the relationship between school wellness policy comprehensiveness and strength tertiles and measured energy-balance behaviors. We explored the possibility that relationships between policy language quality and energy-balance behaviors may vary by either race/ethnicity or BMI, by testing the race/ethnicity × policy score tertile and BMI × policy score tertile cross-product terms. Final models were adjusted for race/ethnicity, as both the scoring of policy quality and energy consumption of snacks appeared to vary by race/ethnicity in our sample. The statistical assumptions underlying each test were checked for violations (eg, homogeneity of variances and outlying and influential cases). Given that our sample had little variation regarding school wellness policy comprehensiveness or strength scores, we selected the 40 highest and 40 lowest scoring districts for further analysis.

Appendix B displays the 96 items measured by the policy coding tool and the percentage of districts with policies that received a rating of 2 (meets or exceeds expectations). In general, federally mandated statements accounted for a high percentage of items that met or exceeded expectations in each section. Five school districts had policies that did not address any of the 7 sections of the policy coding tool. The section that received the highest number of zero ratings was nutrition standards for competitive and other foods and beverages (n = 101 school districts); the least number of zero ratings was for standards for US Department of Agriculture child nutrition programs and school meals (N = 16 school districts).

When assessed for group differences across tertiles of school wellness policy comprehensiveness and strength scores, race/ethnicity and BMI z score were unbalanced (Table 2). Specifically, white mothers were more commonly found in districts with the highest policy rating, while black mothers were more commonly found in districts with the lowest policy rating. Additionally, the lower tertiles of both comprehensiveness and strength scores included adolescents with higher BMI z scores, though the group comparison was not significant for policy strength. We found no evidence of effect modification for either race/ethnicity or BMI when considering the relationship between school wellness policy quality and energy-balance behavior outcomes. In our initial adjusted models assessing snack and physical activity behaviors of participants, we found no significant relationships between policy comprehensiveness or strength tertile and energy-balance behaviors.

In the 40 school districts that had the highest scores for policy comprehensiveness, adolescent mothers reported consuming fewer daily calories from sweetened beverages and more water (Table 3). There was an inverse relationship between policy comprehensiveness and strength and salty, sweet, and meal-type snacks and total snack calories. Policy strength was significantly associated with a lower BMI z score and was also inversely related to sweetened beverage consumption.

Our study had several limitations. First, this is a cross-sectional study that does not allow for assessment of temporal relationships. We did not assess policy effect or implementation which may vary by school district. We had limited information on the school districts in our sample, so were unable to address heterogeneity and generalizability issues. Many of the policies under observation were not required until the start of the 2006-2007 academic year, which may not provide enough time to see the full effect of the policies on measured behaviors. We also present information on a group that may not be generalizable to broader school-district populations. Finally, interpretation of our findings should be considered within the limitations of self-report measures.

School wellness policies associated with PAT programs for adolescent mothers in multiple states may be limited in their ability to promote a healthy school environment. Improvements in the quality of school wellness policies may help to enhance the school environment and, in turn, energy-balance behaviors of adolescents. Future studies, reflecting naturalistic or prospective designs, are needed to evaluate the effect of the strength and comprehensiveness of policy language on energy balance in high-risk postpartum teenagers. Evidence from this work can provide additional guidance to federal or state government in mandating not only policy content, but systematic evaluation. To be active advocates for their adolescent, parents need to be accurately informed about the quality of the wellness policies in their adolescent's school. Quality assurances are needed so that school wellness policies are not missed opportunities for encouraging energy-balance behaviors and preventing obesity among adolescent mothers.