Dietary and physical activity (PA) habits are formed at a young age [1
], whereby unhealthy habits can already lead to overweight and obesity [2
]. The health behaviours of children are suboptimal in Western countries, including the Netherlands: 42% of children (aged 4–9 years) consume at least 150 g of fruit per day, this percentage drops to 20% for 9–12 year olds. The prevalence of vegetable intake shows similar percentages: 41% of 4–9 year olds and 25% of 9–12 year olds eats at least 150 g of vegetables per day [3
]. Regarding PA, only half (48%) of Dutch children (aged 4–12) meet the guidelines for PA of 60 min of moderate-to-vigorous physical activity (MVPA) per day [4
]. Consequently, 13–15% of Dutch children (aged 2–21 years) are overweight, and 1.8–2.2% are classified as obese, which is a two- to three-fold increase compared with 1980 [5
]. Childhood overweight often tracks into adulthood [6
] and is related to health and psychosocial problems, reduced quality of life, and higher health care costs [7
]. An association also exists between health and educational achievement: Health status affects the capacity to learn, while educational achievements affect health status [10
]. This link between health and education often results in persistent socioeconomic health inequity problems that continue to exist from generation to generation [11
Schools are increasingly recognized as significant in improving children’s health behaviours since a large proportion of a child’s day is spent there, and schools reach all children [13
]. However, school-based health interventions are often not integrated in the school system and are characterised by relatively low priority, a lack of coordination, and are often supply-driven, resulting in limited effects or effects that diminish in the long term [15
]. The Health-Promoting School (HPS) framework, initiated by the World Health Organization, aims for a whole-school approach, with a focus on reorienting school systems toward sustainable health promotion [17
]. HPS focuses not only on classroom-based health education, but also on changes in school policy and the schools’ physical and social environment, using bottom-up involvement of pupils, parents, teachers and staff. Several reviews have been published on the effectiveness of HPS in improving the health and well-being of school children [18
]. Even though the findings indicate small favourable effects in terms of PA and healthier food choices, the reviews also reveal that the findings were not uniform across the included studies. Many studies showed suboptimal results, often due to a short duration of the intervention, a lack of a whole-school approach and implementation challenges [22
]. Implementation challenges can be considered a result of the interaction between the intervention and the specific context [25
]. Therefore, various studies suggest revising the idea of interventions as something fixed or static, and considering them as ‘events’ occurring within the school system [28
The ‘Healthy Primary School of the Future’ (HPSF) is a Dutch initiative based on the HPS framework (including, e.g., whole school approach, participation, partnerships) and embraces the contextual systems approach [29
]. This initiative aims to sustainably integrate health and well-being within the school system.
The processes and effects of the initiative, implemented in four pilot schools, are being investigated in an overall study by a multi-disciplinary research group [29
]. The primary outcome of the overall study is children’s BMI z
-score, which significantly decreased after 2 years’ follow-up in the HPSF schools compared to control schools [31
]. The current study focuses on two key aspects of HPSF, i.e., healthy nutrition and PA. Recent research suggests that by addressing two clustered health behaviours, a spill-over or synergistic effect might occur, whereby the probability of enhancing one health behaviour increases when an individual has successfully changed the other health behaviour [32
]. This means that, for example, an increase in physical activity may lead to improved eating behaviours and vice versa. Therefore, simultaneously addressing healthy nutrition and PA might be more effective due to the facilitation of this potential synergistic effect.
The aim of the current study is to examine the effects of HPSF on children’s dietary and PA behaviours after 1 and 2 years’ follow-up compared with control schools, with two schools focussing on both nutrition and PA (full HPSF), and two schools focussing only on PA (partial HPSF). We hypothesized that in the full HPSF, effects will be noted on both dietary and PA behaviours, and in the partial HPSF mainly on PA behaviours. Additionally, we hypothesized that larger effects will be found in the full HPSF, due to the potential synergy between dietary and PA behaviours in children.
4. Statistical Analyses
Data were analyzed using IBM SPSS Statistics for Windows (version 23.0, IBM Corp, Armonk, NY, USA). Pearson’s chi-square tests and ANOVA tests were conducted to analyze the comparability of observed participant characteristics at baseline, i.e., gender, study year, SES status, ethnicity, BMI z-score, and PA and dietary behaviours, among the full HPSF, the partial HPSF, and control schools. The percentage of children who improved in a specific behaviour after 1 and 2 years, i.e., changed in a favourable direction compared with their baseline result, was studied by descriptive statistics. Linear mixed model analyses were used to assess the longitudinal intervention effects on children’s PA levels and behavioural outcomes; Generalized Estimating Equations were used for binary outcomes. Since measurements were repeated, within participants we used a two-level model with measurements as the first level and participants as the second level. The fixed part of the model consisted of group (full HPSF, partial HPSF, control), time (T0, T1, T2) and the interaction terms of group with time. We were not able to include class as a level in the model, because commonly more than one division of a class existed, e.g., 4a or 4b, and children often did not have fixed class divisions for all years. All analyses were adjusted for the covariates: gender, study year at baseline, SES, ethnicity, and children’s BMI z-score at baseline. The analyses regarding children’s PA levels were also adjusted for weather conditions (mean temperature, sun exposure in hours/day and precipitation in hours/day). Missing data, including missing data at baseline, were imputed using a multiple imputation method with fully conditional specification (FCS) and 10 iterations, generating 50 complete datasets. BMI z-score, gender, study year at baseline, school type, ethnicity, SES score, temperature, sun exposure, and precipitation were used to obtain a complete covariate set, with a likelihood-based approach being used for missing outcome variables. This latter was done for practical reasons as the number of outcome variables was too large. A two-sided p-value ≤ 0.05 was considered statistically significant. Standardized effect sizes (ES) were determined for numerical outcomes, which were computed as pooled estimated mean difference divided by the square root of the pooled residual variance at baseline. Binary outcomes resulted in odds ratios.
At baseline (T0), 2326 children and their parents were invited to participate in the overall study to investigate the effects of HPSF; 60.3% joined the study (n
= 1403). Because of the study’s dynamic population, a total of 1974 children and their parents participated in the study within the 2-year follow-up period (data collected at one time-point at least). Due to the selection used for the current study, i.e., only including the children who were in study years one to seven at baseline and excluding school switchers, we included 1676 children in the analyses. This selection and the study’s flow diagram are similar to the study that investigated the 1- and 2-year effects of HPSF on children’s BMI z
]. Of these children, 47.4% were boys, the mean age was 7.5 years old, and 94.1% had a Western ethnicity. In the full HPSF, 537 children were included, in the partial HPSF, 478 children, and in the control schools, 661 children. No covariates differed significantly at baseline between the three school groups, except for BMI z
= 0.034): children in the control schools (BMI-z
= 0.232) had a higher mean BMI z
-score compared with children of the full HPSF (BMI-z
= 0.051) and the partial HPSF (BMI-z
= 0.092). Regarding children’s dietary and PA behaviours, many significant differences existed at baseline, with unhealthier behaviours mostly found in the children in the control schools compared to the full and partial HPSF (Table 1
, Table 2
and Table 3
). Not all parents filled out the parent questionnaire: Parents of 1115 children (66.5%) completed the questionnaire at least once. The child questionnaire was filled out at least once by 96.1% of the children, the child lunch questionnaire by 98.3% of the children. Sufficient accelerometer data, i.e., enough wear time to be included in the analyses, in at least one measurement was reached in 81.5% of the children.
5.1. Children’s PA Behaviours
Significant favourable intervention effects were found in the accelerometry data in the full HPSF versus control schools (Table 1
). The percentage time spent sedentary had decreased more (ES = −0.23) and the percentage time spent in light PA had increased more (ES = 0.22) at T2 in children of the full HPSF compared with control schools. More than a quarter of all children (28.2%) improved, i.e., decreased their sedentary time at T2 in the full HPSF, which was more than the percentage of children in the control schools (21.6%). The percentage time spent in MVPA did not differ significantly in the full HPSF compared with control schools. However, the percentage of children who improved their time spent in MVPA was higher in the full HPSF (44.4%) than the control schools (35.8%). The parent-reported data regarding children’s PA behaviours showed mixed results: The total time per day spent on both PA behaviours (ES = −0.22) and sedentary behaviours (ES = −0.29) had decreased more at T2 in the full HPSF compared with control schools. In the partial HPSF, no significant intervention effects were found in the accelerometry data or parent-reported data compared with control schools (ES between −0.07 and 0.08).
5.2. Children’s Dietary Behaviours
Significant favourable intervention effects were found for parent-reported children’s dietary behaviours in the full HPSF. Children’s healthy dietary behaviours (total score for breakfast, fruit, vegetables, and water) improved significantly more in the full HPSF compared with control schools at T1 (ES = 0.20) and T2 (ES = 0.19) (Table 2
). Effect sizes per item of this total score were largest for water consumption (Supplementary materials Table S1
). Children’s unhealthy dietary behaviours decreased significantly more for the full HPSF versus control schools at T1 (ES = −0.23). A significant favourable intervention effect was also found for child-reported water consumption at school: at T1 and T2, a significantly higher increase was found in children of the full HPSF compared with control schools (T1: ES = 1.03; T2: ES = 1.14). More than three-quarters of all children improved, i.e., increased their water consumption at school at T1 and T2 in the full HPSF, which was almost double the percentage of children compared with the control schools. In the partial HPSF, no significant intervention effects were found for parent-reported children’s dietary behaviours compared with control schools (ES between −0.14 and 0.07). Results on child-reported unhealthy dietary behaviours showed a significant favourable intervention effect at T2: a significantly larger decrease in the partial HPSF compared with control schools (ES = −0.25).
5.3. Children’s Lunch Intake
Significant intervention effects were found for children’s lunch intake (child-reported) in the full HPSF: A significantly higher increase was found at T1 for the consumption of fruit (OR = 2.63), vegetables (OR = 3.17) and dairy products (OR = 4.43) compared with control schools (Table 3
). These higher increases remained significant at T2 for the consumption of vegetables (OR = 4.39) and dairy products (OR = 4.52). The consumption of grains and butter during lunch decreased significantly more at T1 (grains: OR = 0.43; butter: OR = 0.22) and T2 (grains: OR = 0.45; butter: OR = 0.19) in the full HPSF compared with control schools. The consumption of at least two food types during lunch increased significantly more in the full HPSF compared with control schools (OR = 3.51 (T1) and 2.98 (T2)). The consumption of five common food type combinations improved by approximately 30–40% at T1 and T2 in the full HPSF. In contrast, this percentage was much less in the control schools (8–20%) (Supplementary materials Table S2
). In the partial HPSF, the consumption of vegetables (OR = 0.58), dairy products (OR = 0.45) and butter (OR = 0.64) during lunch significantly decreased more at T2 compared with control schools.
HPSF is a health-promoting school initiative that uses a contextual systems approach [17
]. The initiative aims to create health-promoting changes in different aspects of the school system, i.e., school’s physical and social environment, school’s health policy, education, and school routines. On top of the HPS framework, the aim was to create some form of positive disruption in the school, which should lead to momentum for bottom-up processes to institutionalise health-promoting routines in the school. The aim of the current study was to examine the effects of HPSF on children’s dietary and PA behaviours after 1 and 2 years’ follow-up compared with control schools. Favourable intervention effects on children’s dietary and PA behaviours were found for the full HPSF. In contrast, almost no significant favourable results were found for the partial HPSF, where we expected favourable effects on children’s PA behaviours. These effects are in line with the findings of the review of Langford et al., who investigated comparable school-based initiatives [18
]. This review stated as well that PA behaviours significantly improved only in the initiatives with a focus on both healthy nutrition and PA behaviours, and not in initiatives that focused solely on PA behaviours. In contrast to our study, this review did not find any significant results on behaviours related to healthy nutrition for the schools with a focus on both healthy nutrition and PA. However, comparison is limited, since we used total scores and this review used fruit and vegetable intake and fat intake as outcomes.
The results found in the full HPSF regarding the accelerometry data can be seen as small intervention effects according to Lipsey’s guidelines for effect sizes (small (0–0.32), medium (0.33–0.55) and large (>0.55)) [53
]. The significant effects on children’s sedentary time and light PA, but not on MVPA, are in line with other studies of school-based initiatives [54
]. Contrary to the accelerometry-data, parent-reported data showed in the full HPSF not only a favourable effect (decrease in sedentary behaviours), but also an adverse effect (decrease in PA behaviours). These differences in effects found by using the two methods might be explained because assessing PA by subjective parent-reported questionnaires has a lower validity than objective accelerometry [56
]. However, the differences might also be due to the focus of the PA-related questions for parents being outside of school hours, while the accelerometers assessed PA over the whole day. Children of the full HPSF have less time for sedentary and PA behaviours outside of school because of the extended school day. Since both behaviours decreased, it does not necessarily mean that the extra PA at school resulted in children compensating for PA outside of school hours, which has been found in other studies [57
]. More in-depth research is needed to investigate the difference in effect during and outside of school hours on children’s PA behaviours.
The large favourable intervention effect on school water consumption in the full HPSF is probably a result of implementing additional health-promoting changes related to water, e.g., handed out water bottles to all children and improved their school water policy, which created a more health-promoting environment and policy in the school. Both are important aspects of the HPS framework. Both schools referred to the momentum in the school created by the lunch to implement these water-related changes [40
Furthermore, the increase in the consumption of fruits, vegetables and dairy products, and the decrease in grains and butter in the full HPSF seem to indicate that children eat more different food types during lunch. Their lunch intake seems to have changed from a typical Dutch bread-based lunch to a more diverse lunch. The large favourable intervention effect on the intake of at least two food types during lunch seems to validate this conclusion.
The intervention effects in the full HPSF were quite similar at both time points, and the T2 intervention effects were even higher for children’s PA behaviours than the T1 intervention effects. This seems to indicate that the effects are not only due to the children’s enthusiasm for and cooperation with the new changes in school, but that new habits and routines may have developed in the children’s health behaviours. However, longer follow-up periods are needed to investigate the long-term effects.
The main difference between the full and partial HPSF was the implementation of the lunch, the duration of the lunch break time, and the implementation of additional health-promoting changes. However, the two versions of HPSF also had many similarities: both implemented the structured PA sessions in a comparable way, and they were quite similar in the coordination of HPSF and the support of external partners. Nonetheless, the full HPSF was more effective than the partial HPSF, also regarding children’s PA behaviours. Three possible explanations can be given. First, as hypothesized, simultaneously addressing nutrition and PA seemed to create a synergistic effect that led to greater effectiveness. Various studies have indeed suggested that dietary and PA behaviours are associated and that the probability of enhancing a second behaviour, e.g., PA, increases when an individual has successfully changed a first behaviour, e.g., healthy nutrition [32
]. Second, both the full and partial HPSF used a contextual systems approach and included top-down and bottom-up processes to create health-promoting changes in the school [29
]. Since the two top-down changes were also contextualized bottom-up, this resulted in some differences between schools in the form of the changes, e.g., assigning external pedagogical employees to a specific activity or to a specific class. The content of the changes remained comparable, however. Moreover, the results of the process evaluation of Bartelink et al. indicated that the lunch turned out to be a positive disruption in the full HPSF that created momentum for more bottom-up processes, including more involvement and support of teachers and parents, and it has led to additional health-promoting changes (e.g., health-promoting policy) [40
]. The partial HPSF did not implement the lunch, which resulted in limited bottom-up processes and no additional health-promoting changes in these schools. Due to this lack of additional changes, the whole school approach as suggested by the HPS framework is limited, which might explain the differences in effect between the full and partial HPSF. Third, the partial HPSF did not extend the lunch break time, whereas the full HPSF created a longer break by extending it by approximately 60 min. Consequently, the time for the structured PA sessions was longer in the full HPSF compared with the partial HPSF.
Although we hypothesized that differences in effect would exist between the full and partial HPSF, we did not expect that in the partial HPSF no effects on children’s PA behaviours would be found at all. An explanation for this absence might be that children’s PA behaviours in the specific weeks of measurements were not representative of the children’s PA behaviours in general. Moreover, the effects on children’s PA behaviours might also be too small to detect. The results of the study regarding the effects of HPSF on children’s BMI z
-score found a small but significant decrease in BMI z
-score after 2 years in both the full (ES = −0.08) and partial (ES = −0.07) HPSF [31
], which suggests that also in the partial HPSF, some changes have occurred in the children’s health behaviours [58
]. Many small improvements on several different health behaviours can lead to a decrease in BMI z
-score, since it is the co-existence and interaction of specific nutrition and PA behaviours that results in a positive (or negative) energy balance and weight gain (or loss) [59
7. Limitations and Strengths
The longitudinal quasi-experimental design can be seen as a limitation of this study, since we were unable to (cluster-) randomize schools. However, due to this design, we were able to test the effectiveness in terms of differences in children’s health behaviours between the three school groups over time, and were also able to enroll schools on the basis of motivation, which reflects the real-life situation of school health promotion. However, due to no randomization, it has probably resulted in significant baseline differences between the three groups. The baseline differences in BMI z-scores and health behaviours seem to indicate that children in the control schools have developed stronger habits in unhealthy behaviours, which have already led to more overweight or obesity. These stronger habits can be more difficult to change, but also show more room for improvement for the children in the control schools compared with the full and partial HPSF, which can result in an underestimation of the effects. To deal with the limitation of no randomization, we controlled in all analyses for BMI z-score at T0, gender, study year at T0, SES score, and ethnicity. Moreover, a methodological strength of the study is the objectively measured PA levels, all collected in the same season, and the matching of all measurements in the same week.
In addition to the abovementioned methodological limitation regarding assessing PA behaviours among parents, the use of questionnaires in general had its limitations as these are subjective measurements, which may lead to socially desirable answers [61
]. Therefore, we used different data sources to obtain information about the children’s health behaviours. The advantage of using questionnaires for children was that a high response rate (child questionnaire: 96.1%, lunch questionnaire: 98.3%) could be achieved by classical inquiry; children are often more honest in their answers and children’s behaviours during the whole day can be assessed [62
]. By pretesting, we made adjustments for age and improved the clarity of the questionnaires. However, these child-appropriate questions lead to less detailed data. The advantage of using a parental questionnaire was that more detailed questions could be asked. However, only children’s behaviours outside of school hours can be assessed by them, and the response rate was much lower (66.5%).