Long-term impact of a school-based nutrition intervention on home nutrition environment and family fruit and vegetable intake: A two-year follow-up study

Long-term data on maintenance of intervention effects of health promotion programs targeting fruit and vegetable (F&V) intake in children is lacking. We conducted a two-year follow-up of Brighter Bites, a school-based nutrition education and food co-op intervention found to be effective in increasing child intake of F&V. A one-group, pre-post evaluation design was used to assess the two-year post intervention impact of the program on child and parent dietary intake and home nutrition environment. In 2016–2017 school year, we conducted a follow up of 262 parent-child dyads who had previously participated in Brighter Bites in a 2013–2015 evaluation study in six low-income Texas elementary schools. Child dietary intake was measured using a parent-reported food frequency questionnaire, and surveys measured parent F&V intake, and home nutrition environment. Results of a multi-level regression analysis showed that, two years post-intervention, as compared to baseline, there was a significant increase in child intake of fruit, vegetable, and fiber, and significant decreases in total fat intake and percent daily calories from sugary beverages (p < 0.05). Parent dietary data showed significant increases in fruit intake, and intake of F&V combined (p < 0.05). Changes in home nutrition environment included: increased frequency of cooking behaviors, increased usage of nutrition facts labels in making grocery purchasing decisions, and increased food availability of F&V (p < 0.05). This study demonstrates potential long-term sustained impact of a comprehensive school-based intervention among low-income children and their families.


Background
Adequate fruit and vegetable (F&V) intake is critical for proper child growth and development (US Department of Health and Human Services and U.S. Department of Agriculture, 2015; Ogden et al., 2010;World Health Organization, 2013). Even with health promotion efforts, F&V intake remains below recommendations across age groups in nationwide surveillance conducted as part of the National Health and Nutrition Examination Survey (NHANES) (US Department of Health and Human Services and U.S. Department of Agriculture, 2015). Moreover, children from lower socioeconomic households are likely to consume fewer F&V than higher socioeconomic households (Lee-Kwan et al., 2017).
While health promotion programs have successfully demonstrated short-term impact on F&V intake among children, few studies report on long-term maintenance of effects on behaviors (Zarnowiecki et al., 2014;Jones et al., 2011). There is a need for long-term follow-up of dietary interventions among low-income populations to determine sustainability of intervention effects, and to further refine intervention strategies (Zarnowiecki et al., 2014;Jones et al., 2011;Appleton et al., 2016); supported by multiple systematic reviews (Jones et al., 2011;Appleton et al., 2016).
Brighter Bites is a 16-week school-based nutrition intervention targeting increased F&V intake among low-income children and families. A two-year quasi-experimental non-randomized controlled study conducted in 2013-2015 to assess the impact of Brighter Bites among 1st grade children across one school year demonstrated significant improvements in child dietary intake and home nutrition environment among participating families compared to those in wait-list comparison schools (Wang and Stewart, 2013). Subsequently, we present results of a two-year follow-up study in 2016-2017 to determine program maintenance effects.

Design
We conducted a two-year follow-up using a one-group pre-post evaluation design across the six intervention schools that participated in Brighter Bites in the 2013-2015 evaluation study. Control schools from the parent study could not be included in the two-year follow-up; waitlist controls received the intervention at the end of 2015 (see Fig. 1).

Description of brighter bites
Brighter Bites is a theory-grounded school-based health promotion program. Detailed description of Brighter Bites is provided elsewhere (Wang and Stewart, 2013). Briefly, Brighter Bites intends to increase F&V demand and consumption by children and parents through a 16week school-based food co-op during one school year offering weekly fresh produce distribution (~50 servings/family); nutrition education in schools and for parents; and weekly recipe tastings during produce pick up time. Brighter Bites has a 3-on 3-off formula whereby they are in a school three years in a row (assuming the school chooses to continue). This allows families sustained access to the program for up to three years if they choose to continue. Brighter Bites is implemented in early childhood centers and elementary schools, typically serving children ages 3-12 years old and their families.

Participants
In the parent study in 2013-2015 (Wang and Stewart, 2013); six intervention schools received Brighter Bites for one school year (n = 407 parent-child dyads); six comparison schools implemented an evidencebased coordinated school health program (n = 310 parent-child dyads) in Houston, Texas (see Fig. 1). A convenience sample of public and charter schools that enrolled 1st grade children with >75% of children enrolled in the free/reduced lunch program were eligible to participate in the parent study. Within each school, two to three 1st grade classrooms were targeted. Consent packets were sent home to all parents in selected classrooms; all students were eligible to participate in Brighter Bites, but only those who consented were measured. Pre-/post evaluation was conducted across intervention and comparison schools at baseline and end of one school year (Wang and Stewart, 2013). In the 2016-2017 school year, we conducted a two-year follow-up of families in intervention schools only. At this time, all schools in the parent study were receiving the Brighter Bites program. All participants completing baseline and post-intervention measurements in the parent study were eligible for inclusion in this follow-up study (completers). Once school consent was obtained for the follow-up study, child participant school

Evaluation
All schools from wait-list control group received the intervention following termination of the original intervention evaluation study; these schools were not included in the twoyear longitudinal follow-up. and grade information from the parent study was used to send home consent packets and study surveys with children to parents. Of 407 families participating in Brighter Bites who completed baseline data in the parent study, there were 283 completers (69.5% of families completing baseline) at the end of 2013-2015 parent study, of which 262 parent-dyads (92.6% of completers) agreed to participate in this two-year follow-up study. Respondents to the follow-up study were more likely to be older (p = 0.03), mother in relationship to child (p = 0.001), Hispanic (p = 0.001), born in a country other than the U.S. (p < 0.001), and bilingual (p = 0.001) (data not presented in tables). Study flow is presented in Fig. 1. Written informed consent was obtained from all parents. The study was approved by the University of Texas Health Science Center, Committee for Protection of Human Subjects.

Measures
Follow-up measurements were obtained in Spring 2017 using the same measures as the parent study (Wang and Stewart, 2013).

Child dietary intake
Child dietary intake was measured using the parent-reported, previously validated Block Kids Food Screener (BKFS) which includes 41 items to assess both frequency and quantity of foods consumed (Sharma et al., 2016). The BKFS was sent home with children, self-completed by parents, and returned to project staff through children. Completed measures were sent to Nutrition Quest for analysis.
Responses were used to assess number of servings of various foods consumed, including fruit and fruit juices, vegetables, potatoes (including French fries), whole grains, meat/poultry/fish, dairy, legumes, saturated fat (calculated from intake of each FFQ line item aside from sodas), and "added sugars" (calculated from intake of sweetened cereals, soft drinks, and sweets) (Sharma et al., 2016). All portion sizes are age and sex-specific. Dietary intake data were calculated in nutrient densities and divided by total caloric intake to standardize all dietary intake data to 1000 kcal to adjust for increase in caloric intake over time, which is expected as children grow, and to allow for comparability within and between subjects. Intake of sugary beverages was determined (both kcal and frequency).

Parental F&V intake
Parental F&V intake was measured using a previously validated 10item self-report Fruits and Vegetable Screener National Institutes of Health (NIH) block screener (Hunsberger et al., 2015). Frequency of F&V consumption was assessed over the past month with 9 response options (never to 5 or more times per day). This was scored using National Cancer Institute calculation protocols (linked to MyPyramid guidelines); intake is presented in cups (Thompson et al., 2002).

Home nutrition environment
Parental rules for limiting portion sizes, screen time, fried foods, fast food, and sugary beverages, and eating family dinners and rules to finish all foods on plates were measured using a self-report questionnaire of items previously validated with similar populations (National Institutes of Health, 2020; Ding et al., 2012;Edmundson et al., 1996;Baranowski et al., 2000). Home mealtime environment was assessed using previously validated items for frequency of cooking from scratch at home, eating out, using nutrition facts labels, F&V served at mealtimes and snacks, sugary cereals, and sugar-sweetened drinks at meals (Ding et al., 2012;Edmundson et al., 1996;Baranowski et al., 2000). These items used a Likert-type scale.

Statistical analyses
All analyses were performed using STATA software, version 14.2. Means, standard deviations (SD) and frequencies were computed for all demographic data and other variables of interest. Repeated measures mixed effects linear regression models were applied to account for clustered data with time (level 1) nested in subjects (level 2), and schoollevel clustering. All macro and micronutrients were standardized to 1000 kcal/day. Standardizing intake to 1000 kcal/day adjusts for the increase in intake over time and allows for comparability of nutrients across time periods. Changes in child F&V intake, parent F&V intake, and home nutrition environment from baseline (2013-2014) to twoyear post-intervention follow-up (2016-2017) were estimated. Sociodemographic variables were included in models only if coefficients were changed by >10%. Significance was at p < 0.05. Additionally, we adjusted for Brighter Bites attendance among families in all models. Missing data are likely not random, Maximum Likelihood Estimate (MLE) was used in analyses but we did not use any imputations because <10% of data was missing (Penkilo et al., 2008).

Discussion
Our study adds to current literature assessing maintenance of intervention effects among low-income children and families. We saw maintenance of many post-intervention behavior and home environmental changes at two-year follow up, including child F&V intake, decreased child consumption of added sugars, and parental understanding and usage of nutrition labels to make food purchasing decisions (Wang and Stewart, 2013). Moreover, we saw additional positive behavior changes in parent intake not evident immediately postintervention, including increased intake of vegetables, and decreased frequency of eating out (Wang and Stewart, 2013). Prior studies and theoretical models have demonstrated that parent behavior is an important mediator for change in child F&V intake (Rasmussen et al., 2006). Those that incorporate strategies of F&V provision coupled with education could have long-term impacts on both parent and child diet.
Our study adds to current literature on longer-term intervention effects for children and parents. In a 2016 systematic review, Appleton et al. found that long-term effects are often not assessed, making it challenging to determine maintenance of intervention effects (Jones et al., 2011). Prior studies have identified barriers to purchasing produce among low-income families including cost, transportation, and lack of quality and variety available (Haynes-Maslow et al., 2013). One likely reason for maintenance effects seen in our study is because programs such as Brighter Bites use access plus education strategies whereby lowincome families get a substantial amount of produce to take home regularly while learning how to use it, thus creating health habits. Studies show both access and education are important to achieve behavior change (Verghese et al., 2019). Families on a limited budget may not purchase or consume F&V due to lack of access, affordability, or knowledge on how to use it, or fear that their family may not consume it (Drewnowski and Darmon, 2005;Hilmers et al., 2012), thus effectively reducing demand for F&V over time in these families. With Brighter Bites, families get a free trial of 20-25 lbs. of 8-12 different kinds of produce weekly for 16 weeks in a school year while learning how to use it, allowing low-income families with children to consume F&V without financial risk while they are in the program. Brighter Bites has a three on three off formula whereby they are in a school three years in a row (assuming the school chooses to continue). This allows families sustained access to the program for up to three years if they choose to stay in the program. Furthermore, evidence suggests that children need to try new foods 12-14 times before liking it (Birch and Fisher, 1998), which is reinforced in Brighter Bites. Our study demonstrates that these strategies are potentially creating a longer-term demand for F&V such that families may continue these habits by obtaining F&V even after the program ends.

Strengths
Strengths of the study include follow-up time to evaluate sustained intervention effects. Additionally, this study measures both child and parent diet; includes measures of home nutrition environment; and is focused on a low-income sample. Finally, the response rate of our sample was relatively high (92.5%).

Limitations
This study relies on self-report of dietary intake and home nutrition environment. Without a comparison group, it is not possible to determine if observed changes are solely attributable to the Brighter Bites intervention; estimates may be inflated due to selection bias due to attrition. Missing data are likely not random, MLE was used in analyses; we did not use any imputations because <10% of data was missing. We did not have a comparison group for follow-up; it received Brighter Bites after the parent study period. However, changes seen in the current follow-up study suggest maintenance of intervention effects among those who participated in the long-term follow-up. Given self-report data, there is social desirability bias; we expect this would be consistent across time-points and should not affect the point estimate. These limitations notwithstanding, our study demonstrates potential sustained positive longer-term impacts of a comprehensive school-based intervention among low-income children and families.  α Coefficients were calculated using Multilevel Mixed Effects Models.
λ CI stands for confidence interval. γ Total daily number of MyPyramid servings for fruits which includes consumption of 100% juice, fresh, canned, and frozen fruits and excludes fruit drinks like Kool-Aid, lemonade, Hi-C, Tang, and Twister. My Pyramid defines servings in cup equivalents with 1 cup of fruit, 100% fruit juice, or ½ cup of dried fruit as 1 cup equivalent. (Graham, 2009). δ Total daily number of MyPyramid servings for vegetables which includes consumption of lettuce salad, tomato sauce, vegetable soups and other vegetables which excludes white potatoes, cooked dried beans, and vegetables in mixtures. MyPyramid defines servings in cup equivalents with 1 cup of raw, cooked, or canned vegetables; 2 cups of raw leafy green vegetables; and ½ cup dried vegetables as 1 cup equivalent (Graham, 2009).
ε Sum of total daily number of MyPyramid servings for fruits and vegetables.
ϕ Odds ratios were calculated using Mixed Effects Logistic Regression.
ω Odds ratios were calculated using Multilevel Mixed Effects Ordered Logistic Regression. * Findings statistically significant at P ≤ 0.05. a Added sugar is based on consumption of sweetened cereals, soft drinks, and sweets based on the Block Kids Food Screener.

Declaration of Competing Interest
Dr. Sharma is on the Board of Directors of Brighter Bites nonprofit organization, the goal of which is to improve access to fresh fruits and vegetables and nutrition education among underserved communities. This is an unpaid, advisory board position. The other authors have no conflicts of interest relevant to this article to disclose.