Improving the school food environment and dietary intake of Dutch primary school students through the provision of a healthy school lunch: results of a pre-posttest effectiveness study | BMC Medicine

Design and recruitment

This study is part of the Healthy School Lunch project in the Netherlands [20]. The overall aim of this project is to encourage healthy eating behavior of children at primary schools by offering a healthy school lunch, based on the Dutch guidelines for a healthy diet [21]. Details on the development and design of the study and recruitment of the schools have been described elsewhere [20]. Briefly, to address the first research question, how providing a school lunch influenced students’ dietary intake at school, we conducted a quasi-experimental, pre-posttest effectiveness study. We recruited 3 intervention schools with children from grades 5–8 (aged 8–12 years) and these children received a healthy school lunch for a 6-month period (November 2018 to April 2019) and their dietary lunch intake data were measured. The focus of the first research question was the uptake of the intervention itself (e.g., the extent to which children who had parental consent actually consumed the foods offered as part of the school lunch), which was implemented at the group (classroom) level. Therefore, we did not include comparison schools to measure dietary intake for this research question [22, 23]. Instead, we included a follow-up measurement after 1 year to study the effect when the school lunch was no longer in place. To answer research question 2 “do children compensate for healthier school meals by eating less healthy outside school hours,” we used an intervention-comparison design and additionally recruited 3 matched comparison schools. Parents from both the intervention and the three matched comparison schools completed compensation questionnaires on behalf of their children.

In order to recruit three schools, the research team approached more than 20 primary schools located either in the capital of the Netherlands or a smaller municipality. In addition, a presentation was conducted for school personnel at a regional meeting for school directors in Ede. Schools with a traditional time schedule, where students have a midday break long enough to go home for lunch and return afterwards, were excluded. To be included, schools needed to have a continuous time schedule where children eat their packed lunch at school as part of the daily routine. A convenience sample of three primary schools agreed to participate in the intervention condition. The three intervention schools were situated in different parts of the Netherlands. School 1 was situated in the capital of the Netherlands with 872,000 inhabitants, and schools 2 and 3 were situated in two small towns with 14,000 inhabitants and 74,000 inhabitants. Schools 1 and 3 were situated in districts characterized by residents with a lower socioeconomic position (SEP) and school 2 was situated in a semi-rural town with children of parents with a relatively high SEP. The level of SEP was determined according to the Valuation of Immovable Property Act (VIPA), which is estimated annually in the Netherlands. This act establishes how municipalities assess the value of homes and businesses within specific neighborhoods, which is strongly associated with the average SEP of the people living in those neighborhoods. Neighborhoods with low or very low VIPA scores were classified as low SEP. The comparison schools were located in the same cities as the intervention schools and were matched based on the socioeconomic position of the neighborhood, determined by the VIPA of the neighborhood. Inclusion criteria for the comparison schools were the same as for the intervention schools, meaning that students followed a continuous schedule and ate their lunch at school rather than going home.

Active parental consent was obtained for all children before the study started. Children also gave verbal consent for their participation. Parents could request at any time to withdraw their children from the study. Across the three schools, 23 children did not have parental consent at any point during the study period and were excluded. Additionally, 35 children initially (at T0) lacked consent but obtained it later, while 24 children initially had consent but later lost it. Furthermore, 8 children (not included in the 24) withdrew from the study due to transferring to another school. The available data from these children was included in the analyses for each individual time point but not in the analyses of changes over time. The study has been approved by the Social Ethical Committee of Wageningen University, the Netherlands (nr. 09131098) and is registered at the Netherlands trial register (NTR): www.trialregister.nl.

All lunch products were funded by the project. Additionally, the costs of cups, lunchboxes, equipment (e.g., fridge, serving plates), and delivery were also covered by the project. Schools did not receive additional funding for implementing or organizing the school lunch. The research team supported the logistical and practical arrangements for the lunch, while the actual execution was the responsibility of the schools, often assisted by volunteers such as parents. A voluntary financial contribution was requested from parents to help cover the lunch costs, but no records were kept of whether or how much individual parents contributed. Each classroom had a small box where children could drop off their contributions. Schools were free to allocate the collected funds as they wished, but based on the schools’ experiences, little to no money was typically found in these boxes.

Data collection/procedure

Data were collected at four time points. Baseline data T0 (n = 249) were collected in September and October 2018 (lunch from home), T1 (n = 269) data were collected in January and February 2019 (after 3 months of implementation of the school lunch), T2 (n = 255) data were collected at the end of March and April 2019 (after 6 months of implementation of the school lunch). In the summer of 2019, it was decided to do a follow-up measurement to investigate if the effects of the intervention remained present when children again brought their lunch from home. The follow-up measurement (T3, n = 185) took place in October and November 2019 (6 months after the school lunch was finished). T3 data were not collected for the children in grade 8 during the intervention, as the data collection took place in the following school year and these children went to secondary school (n = 77).

Instead of using self-reporting questionnaires to determine the content of children’s lunches, we used photographs. Previous research showed that image-based dietary assessment, including food photography, makes dietary intake assessment much easier than using traditional methods [24]. It minimizes recall bias and reduces the respondent’s burden during the recording period, allowing rapid and easy collection of food intake data [25]. Besides, for young children, it is not possible to describe the content of their lunchboxes in detail. At each time point, researchers visited the children in their classrooms just before the lunch break. The researchers introduced themselves and explained the procedure. After the introduction, children placed their lunch on the table (brought from home or taken from the lunch buffet) and received a code card with an ID number. Each lunch was provided with a code card which contained an ID number of the child and where the number of drinks and type of bread could be specified. Children were asked to postpone eating until the researcher filled out the code cards and took a picture of the content of the lunches. After the lunchbreak, the leftovers or empty lunchboxes were again photographed. Observers recorded the type of drink present and provided an estimation of the portion size (e.g., about 200, 300, 500 ml) and number of bread slices. After the lunchbreak, children received a questionnaire on a tablet with the Eye Question app or a paper version, which took children on average 5–10 min to fill out. There were two or three researchers per class, one filled out the code cards and one took pictures of the lunches.

Potential compensation effects, defined as changes in children’s dietary intake outside school hours in response to the school lunch, were assessed using a parent questionnaire administered at T2 in both intervention and comparison schools. Teachers distributed the questionnaires to students, who delivered them to their parents. Parents were given 1 week to complete the questionnaire, after which the children returned it to their teachers.

Dietary intervention

The dietary intervention consisted of an ad libitum school lunch in the intervention schools during a 6-month period. The school lunch consisted of whole-grain bread with a selection of healthy sandwich fillings, such as 100% peanut butter (without added sugar or salt), hummus, 30 + cheese, and lean cold cuts like chicken. Sweet options included fruit spread and 100% apple syrup. Every day, 50 g of raw vegetables were provided per child as part of the lunch. These were so-called ready-to-eat vegetables, meaning they required no preparation and could be eaten by hand, such as mini cucumbers, baby carrots, and cherry tomatoes. We offered one additional item, such as soup, an egg, wraps, or vegetarian meatballs, once a week, in addition to the standard lunch based on sandwiches. These items were included to provide more variety, but they were not meant to replace the core components of the lunch. In schools 2 and 3, there was no lunch on Wednesday as children went home early in the afternoon.

All children were provided with a lunch set (box, bowl, cutlery, cup in a bag) that they had to use during the intervention period, and each child was responsible for bringing the set home to be cleaned daily. The lunch was self-served, allowing children to choose lunch items themselves by walking along a buffet at which all options were displayed. These buffets were located either in the classroom (school 1) or at one central place in the school (schools 2 and 3). All children prepared their own sandwiches at their table in their classrooms and ate their lunch in their own classroom. Children were still permitted to bring lunch items from home, but this was not encouraged. Two lunch menus were developed for 3 weeks, which were repeated in a 3-month cycle. The menus were inspired by our qualitative study among children, parents, and school staff [18]. Nutritional experts and chefs provided input during a workshop to make sure the lunch menus were healthy, varied, and tasty. The developed lunch menus were based on the Dutch nutrition guidelines [21] and in concordance with the recommendations of the Dutch Nutrition Centre. For each school, allergies and religious restrictions were asked in advance and were considered when developing the menus. Therefore, the menus were slightly different for each school. Schools were in close contact with the research team during the intervention. For instance, when a product was disliked by many children after repeated exposure, this product was removed from the menu to avoid food waste and replaced with an alternative product.

Outcome measures

The primary outcome measure of this study was the intake of the core food groups and the secondary outcome measures were “change in vegetable consumption,” “healthiness of the lunch,” “liking of the lunch,” and “compensation behavior.”

Intake of the core food groups

The intake of the core food groups was defined as the intake of grams of vegetables for lunch, number of slices of bread, and drinks in milliliters. For all the products, an estimation of consumption amounts was made based on the photographs and code cards [26]. Portion sizes of products were estimated based on standard portion sizes and weights derived from the Dutch Food Composition Database. For items not listed, such as ready to eat cucumbers, several samples were weighed, and the average weight was used as the standard. No software was used to estimate the portion size. The content of the lunch was classified into the following core food groups. The following food groups were created: bread (white, brown/whole grain, and other (including cornbread, croissant, currant bun, unclear)), drinks (water, (butter)milk, dairy drinks, sugar-sweetened beverages), fruit, and vegetables.

Magnitude of vegetable intake change

To investigate the magnitude of change in vegetable intake among children who showed any change, we calculated the difference in intake for this subgroup. To account for variation in the school lunches offered, vegetable intake was averaged across T1 and T2. The change in consumption was then calculated by subtracting the baseline value at T0 from this average. Only children who exhibited a change in vegetable consumption were included in this analysis (n = 157); those with no change were excluded (n = 74).

To provide more information regarding the change in vegetable consumption of the children who did not consume any vegetables at T0 (N = 231, 92.8%), we calculated the average amount of vegetables that these children consumed at T1 and T2.

Foods recommended within the Wheel of Five change

To evaluate the healthiness of the lunches, we assessed changes in the proportion of products included in the Wheel of Five (Fig. 1). The Wheel of Five is a practical tool developed by the Dutch Nutrition Centre, outlining foods that align with the Dutch dietary guidelines for a healthy diet [27]. This model is the national standard for dietary counseling in the Netherlands and is based on foods associated with a reduced risk of chronic diseases. We consider the food groups “vegetables,” “whole-grain bread,” and “milk products and tea without sugar” as healthy, given that they are included in the Wheel of Five. Food groups which are advised to be limited are considered unhealthy, such as sugar-sweetened beverages. Each lunch product was categorized according to the Wheel of Five (yes or no) using the very easy to use “Do I choose healthy?” app that integrates the criteria of the Wheel of Five. We calculated a percentage for the products that were included in the Wheel of Five for each child. To consider the variation of the offered school lunches, a mean percentage of T1 and T2 was calculated. To calculate the change in the percentage of products within the Wheel of Five, the value of T0 and T1/T2 were subtracted from each other. Only children with a change in percentage of Wheel of Five products were included in this part of the analysis (n = 207). Children with no change were excluded (n = 24).

The Wheel of Five is the Dutch national food-based dietary guideline, developed by the Netherlands Nutrition Centre. It promotes foods associated with a reduced risk of chronic diseases

Self-rated liking of the lunch

Self-rated liking of the lunch was measured by the question “the lunch was tasty.” Children answered this question with smileys on a 5-point Likert scale from a sad red smiley to a green happy smiley. This was measured at T1 and T2.

Potential compensation at home

The questionnaire focused on parents’ perception of changes in their child’s food consumption at home, outside school hours, during the final month of the school lunch period. This aimed to assess potential compensation effects, meaning shifts in home consumption as a response to the healthier school lunches. Parents were asked whether their child ate less, the same, or more of a range of products at home compared to 6 months earlier. The following product categories were included: “ready to eat vegetables, such as tomatoes, cucumber, radish, carrot”; “vegetables during the evening meal, warm or cold such as lettuce and salad”; “sweet snacks such as cookies, chocolate or candy”; “savory snacks, such as crisps, cheese or warm snacks”; “juice, lemonade or soft drinks with sugar (no diet)”; and “water, milk and sour milk.”

Demographic variables

Information on date of birth, gender, and migration background of the children was asked at T1. Migration background was obtained with three open-ended questions in which was asked in which country they, their mother, and their father were born. Children were categorized as having a migration background when at least one of their parents was not born in the Netherlands [28]. Migration background is further distinguished into Western or Non-western. Children were categorized as having a Western migration background if they were born in a country in Europe, North America, or Oceania. Non-western migration background includes countries of origin as Africa, Latin America, or Asia.

Statistical analyses

Descriptive analyses were used to summarize children’s demographic characteristics, dietary intake at each timepoint (T0 till T3), and the liking of the lunch at T1 and T2.

Numbers and percentages were presented for the demographic variables. Numeric data were generally skewed; therefore, the median of consumption per each food group was presented along with the 25th and 75th percentiles (Additional file 1: Table 1). To analyze whether there were any differences in the consumption of the core food groups compared to T0, we used a non-parametric related samples test (McNemar test for paired changes). We performed no multivariate analyses because adjusting for potential confounders would have only been necessary when variables were identified that were associated with both the exposure (healthier school lunches) and the outcome of our study. Since all children in the intervention were uniformly exposed, potential confounding variables could not be present.

To illustrate how the impact of the intervention varied depending on children’s starting point, we created a scatterplot that plotted the change in the percentage of consumed lunch products within the Wheel of Five during the school-provided lunch, against each child’s baseline percentage.

Difference in compensation behavior between the intervention schools and the comparison schools were investigated with a Pearson chi-square tests for independence. Significance level was set at α = 0.05. Data were analyzed with IBM SPSS Statistics 24 [29].