Abstract
The Healthy, Hunger-Free Kids Act of 2010 raised standards for foods sold in US schools. The aims of this paper are to estimate policy effects on dietary quality of US school-aged children, and determine if effects differed by free/reduced-price meal eligibility or age group. We estimated within-child associations between school food consumption and dietary quality (using Health Eating Index - 2010), by comparing two 24-hour dietary recalls per child from the 2007-2016 National Health and Nutrition Examination Survey. We used inverse probability weighting to account for population changes during this period. We scaled changes of association (implementation period versus pre-passage) by proportion of energy from school foods. School foods were 10% of annual calories, 33% on days when children ate school foods. Each percentage point of energy from school foods was associated with 0.08 more Healthy Eating Index points during implementation than pre-passage. The scaled estimated effect of policy implementation on annual dietary quality is 0.84 Healthy Eating Index points, 2.8 points on days when children eat school foods. Estimated effects were slightly greater for children eligible for free/reduced-price meals and high school-aged children, though confidence intervals include zero. The results indicate that targeted policies can improve children’s dietary quality.
Poor dietary quality is the leading risk factor related to death and disability in the US (1). Although dietary quality of US children improved from 1999 to 2012, average dietary quality remains poor (2,3). In addition to immediate impacts of eating a suboptimal diet, eating habits formed during childhood and adolescence persist into adulthood (4–6), underlining the relevance of childhood dietary quality to health across the life course.
The National School Lunch Program (NSLP) and School Breakfast Program (SBP) were established to provide federal support to children’s nutrition (7). The NSLP serves 30 million lunches and the SBP serves 15 million breakfasts per school day (8). NSLP and SBP provide free or reduced-price (FRP) meals to children from households with incomes up to 1.85 times the poverty level (7). Children who receive FRP meals are more likely to eat school meals than children from higher income households (9). Addditionally, elementary and middle school students are more likely to eat school meals than high school students (9).
Following a period of federal, state and local policy reforms (10), the Healthy, Hunger-Free Kids Act of 2010 (HHFKA) (11) was signed into law on December 13, 2010. HHFKA reformed national standards for all foods served in schools (10), better aligning school foods to new dietary guidelines (12). HHFKA rules affecting NSLP/SBP meals (10), competitive foods (e.g. a la carte items), and snacks (13), went into effect over a two-year period from 2012 to 2014.
Among other changes, the HHFKA required schools to offer more servings of fruits and vegetables, greater variety of vegetables, half of grain servings as at least 50% whole grain, limit fat content in milk, and set calorie and sodium limits (12). New meal standards (14) began July 1, 2012 for school lunches and July 1, 2013 for school breakfasts. Minimum nutritional standards for all other foods sold in schools went into effect on July 1, 2014 (13). Although some rollback occurred in the new administration (15), each requirement remains more stringent than rules prior to July 1, 2012. See Web Appendix for a summary table of HHFKA changes to NSLP/SBP meals.
Evaluations of the dietary effects of HHFKA have examined changes in food selection, consumption, and plate waste in school cafeterias (16–20). Students’ cafeteria food selection and consumption shifted to more nutrient-dense, less energy-dense foods (20), including more vegetables (16,18,19) and whole fruits (16,18,19), and less fruit juice (18). One study found increased plate waste following HHFKA implementation actually reduced fruit and vegetable consumption (17).
A post-HHFKA report compared 24-hour dietary recalls from children who ate a school meal with matched nonparticipants (8). School meal participants had better dietary quality than nonparticipants (8), but the study did not compare dietary recalls before and after HHFKA.
The purpose of this study was to estimate the effect of HHFKA implementation on the total dietary quality of all US school-aged children. This was the first study to examine pre-post HHFKA differences in the effect of the NSLP/SBP on population-level, average annual dietary quality. The primary comparison in this study was between the four years before passage of the HHFKA (2007-2010) and the four years after implementation began (2013-2016). We also examined if the effect of HHFKA implementation differed for children eligible for FRP meals versus ineligible, and high school-aged versus younger children.
METHODS
Data source
We used data from five waves of the National Health and Nutrition Examination Survey (NHANES), a repeated cross-sectional survey of the civilian, non-institutionalized population of the US (21). NHANES files used in this analysis are Demographic Variables & Sample Weights, Dietary Interview - Individual Foods, First Day, Dietary Interview - Total Nutrient Intakes, First Day, Dietary Interview - Individual Foods, Second Day, Dietary Interview - Total Nutrient Intakes, Second Day, and Diet Behavior & Nutrition.
Participants in this analysis were 4- to 19-year old children or adolescents who completed two, 24-hour dietary recalls and attended a K-12 school during the school year, in the 2007-2008 through 2015-2016 cycles (N = 9,532). Participants missing one or more variables used in standardization (N = 1,007) were excluded, for a final analytic sample of 8,525.
Dietary assessment
NHANES participants completed questionnaires at home, then visited a mobile examination center for a standardized physical examination and in-person 24-hour dietary recall (22), in which the participant or their proxy reported everything they ate during the previous 24-hour day (23). NHANES dietary interviews were conducted by proxy (typically a parent) through age 5, and with proxy assistance through age 11 (24). A second dietary recall was completed by telephone on a nonconsecutive day after the MEC visit (22). Elementary school-aged children’s reports in person or by telephone do not differ in accuracy (25).
Misreported food is a significant problem of memory-based dietary records, with both items eaten but not reported, and items reported but not eaten, appearing in the records of unassisted children (25) (NB, young children in NHANES were assisted). Although individual 24-hour dietary recalls for children may have high levels of these errors, validation studies have found they provide a valid population-level average of children’s energy intake (26).
Outcome: Healthy Eating Index-2010
Healthy Eating Index-2010 (HEI-2010) was developed to measure adherence to US dietary guidelines (27) for individuals and populations. HEI-2010 is a valid and reliable measure of dietary alignment with those guidelines (28). HEI-2010 includes 12 components, of which nine assess the adequacy of dietary intake of fruits, vegetables, whole grains, etc.; three assess dietary components to be consumed in moderation (refined grains, sodium, and empty calories) (28).
We used the recommended per-day scoring algorithm (29) to create two HEI-2010 scores for each person, for each 24-hour dietary recall. These scores were calculated using the Food Pattern Equivalents Database versions of the NHANES dietary intake files, available from US Department of Agriculture Food Surveys Research Group (30).
Exposure: Healthy, Hunger-Free Kids Act of 2010
This analysis includes three exposure time periods: 1) Pre-passage of the HHFKA (2007-2008 and 2009-2010), 2) Post-passage, but before implementation of new school meal rules (2011-2012), and 3) Implementation period (2013-2016). Because schools began to meet HHFKA goals after the policy was passed, but before legal implementation (31), it was necessary to define an intermediate “post-passage” period. The main comparison for all analyses will be between the pre-passage and implementation period NHANES cycles.
Exposure: energy from school cafeteria foods
Each food item reported in the 24-hour dietary recall has a caloric value and reported source (e.g., “Store - grocery/supermarket,” “Cafeteria in a K-12 school”). We followed a peer-reviewed method for calculating the contribution of school foods to total energy intake (32), summing calories only from food items reported from “Cafeteria in a K-12 school” and dividing by total caloric intake from all food items.
Free and reduced-price meal eligibility
We defined FRP meal eligibility as binary household eligibility by poverty-income ratio. Eligible children are in households at or below 1.85 times the poverty-income ratio (7). If a participant does not have reported income (approximately 9% of participants), we substituted self-reported receipt of FRP meals. We prioritized the income indicator over self-reported FRP meal receipt because participants were only asked if school meals were FRP if they reported eating school meals. Therefore relying on self-report would overestimate school food intake among eligible students.
High school-age
We identified children who reported having completed 8th grade or higher, or age 15 or older at the time of exam if grade level is missing, as high school-aged (versus younger).
Standardization
Several demographic and economic characteristics: child’s sex, age, and race/ethnicity, household size and poverty-income ratio, and the reference person’s (typically a parent) education level and marital status, may modify the relationship between HHFKA implementation and dietary quality. These factors are predictive of participation in NSLP (9) and have been reported to relate to differences in child (2) and adult (33) dietary quality. Demographic and economic changes in the US population during the 10-year period included in this sample could therefore bias estimates of the total effect of the HHFKA.
To eliminate the association between HHFKA passage/implementation and the measured socio-demographic composition of the NHANES sample, we standardized the five samples to the socio-demographic distribution of the 2015-2016 NHANES cycle using inverse probability weighting (34–36).
We performed standardization by fitting an unsaturated multinomial logistic regression, using the sampling weights provided by NHANES (36), to predict the probability that an individual would be observed in each of the five cycles. Regression predictors were measured identically in each NHANES cycle: participant’s continuous age and household size, categorical racial/ethnic group (Mexican American, other Hispanic, non-Hispanic white, non-Hispanic black, other race - including multiracial), gender (male/female), eligibility by income for FRP meals, household reference person’s marital status (currently married/not married), and reference person’s educational attainment (college graduate/lesser attainment). Because this analysis pools five cycles, we divided each participant’s dietary sample weight by five (21).
We created stabilized weights by dividing the predicted probability of being observed in the 2015-2016 cycle by the probability of being observed in the cycle the person was actually observed in: (34). The final analysis weights are the product of the stabilized weights and the NHANES dietary two-day sample weight (36), which accounts for demographic/geographic sampling probabilities, non-response to each 24-hour recall, day of the week of the first 24-hour recall, and Day 1 and Day 2 weekday-weekend categories (37).
Analysis
We estimated the total effect of HHFKA implementation on dietary quality using a two-step process. First, we estimated the proportion of total calories from school foods (including non-school days) in the specified population or sub-population. Second, we estimated the population-specific effect of each percentage point of calories from school foods in each NHANES cycle (32). Thus the scaled estimated effect of food from school on diet quality is the proportion of energy they reported getting from school foods times the effect per percentage point. We address my main research question of how the HHFKA impacts diet quality by estimating the difference in the scaled effect during the HHFKA implementation period compared to the pre-passage period.
The two regression steps were conducted in R using the ‘survey’ package (38) to account for the sampling design (21). We weighted the analyses using the stabilization weights described above and applied appropriate stratification (21,39). Subgroup-specific proportion of energy from school foods was calculated using the R marginal effects estimation package ‘ggeffects’ (40).
To estimate the population-specific effect of each percentage point of calories from school foods, we calculated change scores for the outcome, difference in HEI-2010 score from the first 24-hour recall to the second 24-hour recall, ΔHEI-2010, and the exposure, difference in proportion of energy from school cafeteria foods from the first to the second 24-hour recall, Δ%Energy from schools. Both the outcome and the exposure have a theoretical range of −100 to 100. Change scores were necessary to maintain the appropriate survey design-based analysis as widely available survey procedures do not accommodate repeated measures.
We adjusted analyses for day of the week of the dietary recall, which confounds the association between school food exposure and dietary quality. We assigned a change score of −1 if the first recall occurred on a weekend and the second on a weekday, 1 if the reverse, and 0 if both occurred on weekdays or weekends. To distinguish participants with two weekday recalls from those with two weekend recalls, a second variable was given a value of 1 if both recalls occurred on weekend days and 0 otherwise. Weekend covariates were interacted with time period.
Because the process that governs change in energy from school foods between Day 1 and Day 2 is random with respect to other independent predictors of change in HEI between recalls, we assume the conditional association to be an unconfounded estimate of the effect of energy from school foods on dietary quality. This analysis assumes the within-person effect of school foods on HEI occurs only on days when school foods are eaten.
The final step of the analysis was to translate model estimates into population-level effects. The total estimated effect of the HHFKA on the diets of all US children and specified subpopulations is the product of two random variables estimating first, the proportion of dietary intake from school foods, and second, the implementation period change in the effect on HEI score per percentage point of energy from school foods. Means and variances for the scaled effect, subgroup-specific scaled effects, and differences in estimated effects between subgroups were calculated using standard mathematical formulas for the product and difference of random variables.
Feasibility and sensitivity analyses
We conducted a feasibility analysis to ascertain if we could detect a difference in nutritional density of school foods in children’s 24-hour dietary recalls. We used a peer-reviewed method (41,42) to calculated the nutrient density of school-aged children’s weekday lunches in NHANES 2009-2016 dietary recalls. We used survey regression to compare changes in nutrient density of lunches from school cafeterias versus any other source.
We conducted two sensitivity analyses. We tested the linearity of association between school foods and dietary quality among all school-aged children by adding a quadratic interaction term (Δ%Energy from schools × |Δ%Energy from schools|) to the model. We also tested the fit of three versus five time periods (individual cycles). We compared the fit of those models to the models included in this report using nested ANOVA. See Web Appendix for feasibility and sensitivity analysis results.
RESULTS
Participant characteristics
The demographic characteristics of the total sample, the 2015-2016 target sample, and standardized total sample are shown in Table 1. The standardized total sample closely matches the demographic characteristics of the 2015-2016 cycle. The probability weights ranged from 0.4 to 4 (mean = 1.1).
Exposure to school foods
Ten percent of annual energy intake for school-aged children, including non-school days, came from school cafeterias (95% confidence interval [CI]: 9, 11; see Table 2). This is similar to a previous estimate of energy intake from school food using earlier NHANES cycles (32). Limited to children who reported school foods in their Day 1 recall, 33% of Day 1 energy was from school foods (95% CI: 32, 35).
Children eligible for FRP meals consumed more energy from school foods than did children who were not eligible (13 vs. 7 percentage points, P < 0.001, 2-sided). High school-aged children consumed less energy from school foods than younger children (7 vs. 11 percentage points, P < 0.001, 2-sided). The proportion of energy from school foods did not change significantly across the three time periods.
Association between school foods and dietary quality
The within-person conditional association between change in school food consumption and change in dietary quality is shown in Table 3. The unconditional association for responses with no weekend days is shown in Figure 1. In the total population in the pre-passage period, each percentage point of energy from school foods was associated with a 0.04-point increase in the participant’s HEI score (95% CI: 0.01, 0.08). That association increased by 0.08 HEI points in the implementation period, growing to a 0.13 point improvement in HEI for each percentage point of energy from school foods (95% CI: 0.06, 0.19). Full model results are included in the Web Appendix.
Among students eligible for FRP meals, each percentage point of energy from school foods was associated with greater HEI in the pre-passage period (95% CI: 0.02, 0.10). That association increased by 0.09 HEI points in the implementation period to 0.15 HEI points per percentage point of energy (95% CI: 0.08, 0.22). Among students not eligible for FRP meals, energy from school cafeterias was not strongly associated with greater HEI scores in the pre-passage period (95% CI: −0.03, 0.08). However, the effect grew by 0.08 HEI points in the implementation period to 0.10 HEI points (95% CI: 0.00, 0.20).
For elementary/middle school students, each percentage point of energy from school in the pre-passage period was associated with a 0.06-point increase in HEI scores (95% CI: 0.02, 0.10). That association grew by 0.06 HEI points in the implementation period to 0.12 HEI points per percentage point of energy (95% CI: 0.04, 0.19). For high school students, school foods were neither beneficial nor harmful in the pre-passage period (95% CI: −0.07, 0.09). School meals became substantialy more beneficial in the implementation period, growing to 0.14 HEI points per percentage point of energy (95% CI: 0.00, 0.28).
The feasibility analysis showed that the nutrient density of children’s weekday lunches from school cafeterias increased significantly in the implementation period compared to the pre-passage period, while children’s weekday lunches from other sources were virtually unchanged. The sensitivity analyses did not find improvement of fit using either a nonlinear association between school food intake and dietary quality or a five-level time variable.
Scaled estimated policy effect
Scaled to the relative contribution of school foods to children’s overall energy intake (10%), the estimated effect of the HHFKA on US school-aged children’s average dietary quality was 0.84 HEI points (95% CI: 0.28, 1.4). On days when children reported school cafeteria foods, averaging 33% of energy intake, the scaled estimated effect of the HHFKA on children’s dietary quality was 2.8 HEI points (95% CI: 0.95, 4.7).
The scaled estimated effect was approximately twice as great for students eligible for FRP meals (1.2 HEI points, 95% CI: 0.44, 1.9) than not eligible (0.58 HEI points, 95% CI: −0.08, 1.2), although the 95% confidence interval of the difference includes zero (−0.40, 1.6). The scaled estimated effect of the HHFKA on dietary quality was slightly greater for high school students (0.95 HEI points, 95% CI: 0.13, 1.8) than younger children (0.67 HEI points, 95% CI: −0.06, 1.4), although the 95% confidence interval for the difference again includes zero (−0.82, 1.4).
DISCUSSION
This study examined the dietary effects of the HHFKA in the context of children’s total diets. We found that the first four years of HHFKA implementation was associated with improved dietary quality on days when children ate school cafeteria foods. The findings of our study, scaled to children’s average yearly school food consumption, suggest that that HHFKA implementation between 2013 and 2016 improved the dietary quality of all US children by 0.84 points per day, and by 2.8 points on days when children ate school meals. Although the daily effect of the HHFKA appears small, the NSLP and SBP provide US children with 45 million meals per school day (8), leading to a potentially large aggregate effect.
The HHFKA appears to have been particularly effective for students eligible for FRP meals, who have greater reliance on school foods, and for high school students, whose school foods did not boost dietary quality prior to the HHFKA. Contrary to some predictions, and consistent with other evidence (43), we found no evidence that school food consumption declined during HHFKA implementation.
Prior assessments of the dietary effects of the HHFKA analyzed children’s cafeteria food selection and consumption (16–20). While this approach demonstrated the healthfulness and desirability of new school foods, it did not contextualize policy effects within children’s total diet. More importantly, pre-post studies risk unmeasured confounding due to secular trends in children’s dietary intakes. That would be particularly difficult to rule out in this case because children’s dietary quality was improving in the pre-HHFKA period (2). This study protects against that threat because, at each time point, children serve as their own comparison group.
One limitation of this study is that it does not distinguish between changes in the quality of food from school cafeterias and changes in the quality of food from other sources. However, the results of the feasibility analysis showed improvements in school lunches, but no changes in the nutrient density of weekday lunches from other sources. This suggests that the cause of the observed improvement was changes in school foods, and not foods outside of school.
A more serious limitation is the inability to distinguish between federal, state, and local changes in school meal standards. This limitation is inherent in conducting a national observational study without detailed geographic information; however, the alignment of HHFKA passage and implementation with NHANES cycles gives us confidence that improvements in school foods observed in this study are from the policy of interest. The positive association between school foods and dietary quality in the pre-passage period may be at least partially due to many state- and district-level policies implemented in the decade before HHFKA passage, and earlier advocacy, which together made the HHFKA politically possible (10,44).
Further improving the quality of school foods could be an effective component of a broader set of school practices and policies for reducing the long-term risk of chronic disease (45). All subgroups we examined would benefit from increased NSLP/SBP participation, as long as school foods are as nutritious as those offered now. High schools in particular could do more to promote healthy school foods and restrict access to off campus lunches. Federal regulators must resist pressure to return competitive foods of low nutritional value to schools. Federal policy can improve the overall diet of US children, when policy is made in their best interest.
Data Availability
All data used in preparation of this manuscript are publicly available.
Acknowledgments
The authors would like to thank Dr. Travis A. Smith for his insight on a methodological issue. Additionally, the authors gratefully acknowledge support from the Minnesota Population Center (P2C HD041023) funded through a grant from NICHD.
Abbreviations
- FRP
- Free or reduced-price
- HEI-2010
- Healthy Eating Index – 2010
- HHFKA
- Healthy, Hunger-Free Kids Act of 2010
- NHANES
- National Health and Nutrition Examination Survey
- NSLP
- National School Lunch Program
- SBP
- School Breakfast Program