This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.
There is significant opportunity to improve the nutritional quality of foods packed in children’s school lunchboxes. Interventions that are effective and scalable targeting the school and home environment are therefore warranted.
This study aimed to assess the effectiveness of a multicomponent, mobile health–based intervention, SWAP IT, in reducing the energy contribution of discretionary (ie, less healthy) foods and drinks packed for children to consume at school.
A type I effectiveness–implementation hybrid cluster randomized controlled trial was conducted in 32 primary schools located across 3 local health districts in New South Wales, Australia, to compare the effects of a 6-month intervention targeting foods packed in children’s lunchboxes with those of a usual care control. Primary schools were eligible if they were not participating in other nutrition studies and used the required school communication app. The Behaviour Change Wheel was used to co-design the multicomponent SWAP IT intervention, which consisted of the following: school lunchbox nutrition guidelines, curriculum lessons, information pushed to parents digitally via an existing school communication app, and additional parent resources to address common barriers to packing healthy lunchboxes. The primary outcome, mean energy (kilojoules) content of discretionary lunchbox foods and drinks packed in lunchboxes, was measured via observation using a validated school food checklist at baseline (May 2019) and at 6-month follow-up (October 2019). Additional secondary outcomes included mean lunchbox energy from discretionary foods consumed, mean total lunchbox energy packed and consumed, mean energy content of core lunchbox foods packed and consumed, and percentage of lunchbox energy from discretionary and core foods, all of which were also measured via observation using a validated school food checklist. Measures of school engagement, consumption of discretionary foods outside of school hours, and lunchbox cost were also collected at baseline and at 6-month follow-up. Data were analyzed via hierarchical linear regression models, with controlling for clustering, socioeconomic status, and remoteness.
A total of 3022 (3022/7212, 41.90%) students consented to participate in the evaluation (mean age 7.8 years; 1487/3022, 49.22% girls). There were significant reductions between the intervention and control groups in the primary trial outcome, mean energy (kilojoules) content of discretionary foods packed in lunchboxes (–117.26 kJ; 95% CI –195.59 to –39.83;
The SWAP IT intervention was effective in reducing the energy content of foods packed for and consumed by primary school–aged children at school. Dissemination of the SWAP IT program at a population level has the potential to influence a significant proportion of primary school–aged children, impacting weight status and associated health care costs.
Australian Clinical Trials Registry ACTRN12618001731280; https://www.anzctr.org.au/Trial/Registration/TrialReview.aspx?id=376191&isReview=true
RR2-10.1186/s12889-019-7725-x
Preventing the onset of overweight and obese status in children is a global public health priority [
Children consume up to two-thirds of their daily energy intake at school [
Current evidence regarding the effectiveness of school lunchbox interventions is equivocal. A recent systematic review of such interventions in the school and childcare setting identified just 10 trials and suggested they had little to no effect on the nutritional quality of foods packed or consumed by students [
Mobile text messaging– and mobile app–based interventions have been proven to be a scalable and effective approach for improving a variety of health behaviors—including those of parents—to provide a better child diet [
The research was conducted and reported in accordance with the requirements of the Consolidated Standards of Reporting Trials (CONSORT) statement [
A type I effectiveness–implementation hybrid cluster randomized controlled trial was conducted with 32 primary (students aged approximately 5-12 years) schools across 3 local health districts in NSW, Australia (
Consolidated Standards of Reporting Trials (CONSORT) flow diagram.
Schools were considered eligible if they met the following criteria: government primary schools catering for students from kindergarten to year 6 and located in one of the participating local health districts, greater than 120 student enrolments, current users of the preferred school mobile communication app (SkoolBag), and not participating in other nutrition-based research studies. Schools purchase the communication app for a nominal fee annually, which is then free for parents to download, enabling direct school–parent communication. The app is used by approximately 60% of schools in the region. Central schools (catering for students aged 5-18 years) and schools primarily catering for children with additional needs (such as intellectual disabilities) were excluded. According to a random number generator in Excel (Microsoft Corporation), eligible schools meeting the above criteria were sent a letter of invitation in random order. One week following the invitation, a member of the research team contacted the school principal via telephone to seek consent. A face-to-face meeting was offered to all schools to outline the requirements of the study. Recruitment and consent of schools occurred between February 2019 and May 2019. Recruitment continued until 32 schools provided active signed principal consent to participate.
Opt-in parental consent was required for children and parents to participate in the evaluation of the behavioral outcomes. Parents were also required to be active users of the school communication app, defined as downloading the school communication app on the parent consent form. A strategy to recruit parents and students was developed based on the pilot study and reviews of evidence for facilitating participation in school-based research [
Following baseline data collection, schools (cluster) were randomly allocated in a 1:1 ratio to the intervention or control group based on a random number function in Excel. Randomization was undertaken by a statistician not involved in contacting schools in the study intervention or assessment and stratified by the socioeconomic status of school locality using the Socio-Economic Indexes for Areas (SEIFA 2016), as socioeconomic status is associated with lunchbox contents and child diet [
The multicomponent intervention based on the previous pilot was codeveloped by a multidisciplinary team comprising academic and end-user stakeholders from government health agencies, educational systems, universities, and technology partners and included parent representatives with expertise in nutrition, school-based health interventions, behavior change, implementation science, and technology-based interventions.
The Behaviour Change Wheel [
Lunchbox nutrition guidelines: Using a template developed by the project team, school principals developed, endorsed, and disseminated nutrition guidelines to parents which were consistent with the WHO and the NSW Department of Education Nutrition in Schools policy [
Weekly pushed lunchbox messages: Through the SkoolBag app, 10 weekly electronic messages (push notifications) to support the packing of healthy lunchboxes were disseminated to parents or caregivers. Messages were codeveloped by the research team, public health nutritionists, health promotion practitioners, teachers, and parents and were optimized and refined via a study involving 511 parents [
Resources for parents: Links embedded in the app messages connected parents with electronic resources housed on the program website. These resources provided information regarding health consequences, simple healthy lunchbox swaps that addressed child preference, cost, convenience, and food safety. Physical resources, including a SWAP IT ideas booklet (lunchbox ideas), clear drink bottle for water, and an ice brick to support food safety, were also provided to parents and were distributed to students and parents via the schools’ usual methods of dissemination.
Curriculum resources for schools: Schools were provided with a short online teacher professional learning module (10 minutes) developed by the research team, which included public health nutritionists, health promotion practitioners, and teachers outlining the rationale for the study and providing the skills and resources required to deliver the classroom curriculum lessons. Schools were also provided stage-appropriate curriculum resources which were codeveloped by the research team with input from teachers, parents, and education partners to align with syllabus outcomes that were developed by dietitians and teachers in order to reinforce healthy food preferences. This required teachers to deliver 3 curriculum lessons 10 minutes in duration. Curriculum resources were designed to address the identified barrier to packing a healthy lunchbox of “child preference for discretionary foods.”
SWAP IT logic model. BCT: behaviour change technique; HPS: health promoting schools framework.
Schools allocated to the control group had access to the SkoolBag app but not the lunchbox intervention content. The SWAP IT website was freely accessible by the general public, including parents and schools; however, schools and parents were not notified or directed to this site. There was no information (nutrition or otherwise) provided to the control group, and they participated in data collection only and continued usual school business.
The primary outcome was the mean energy (kilojoules) content of discretionary foods packed in the school lunchboxes by parents who were users of the school mobile app, assessed at baseline and at 6-month follow-up. A detailed description of the study measures and data collection methods have been described in a published protocol [
The secondary outcomes associated with lunchbox energy were mean total energy (kilojoule) packed within the lunchbox; mean total energy (kilojoules) consumed from the lunchbox; mean energy (kilojoules) from discretionary foods and drinks consumed within the lunchbox; mean energy (kilojoules) from healthy foods packed and consumed from the lunchbox; and percentage of lunchbox energy from discretionary and healthy foods and drinks, both packed and consumed. Data were collected at baseline and immediately after the 6-month intervention with the SFC as outlined in the previous section. Following the analysis of the premeal lunchbox photo, dietitians analyzed the postmeal photo.
At baseline and at follow-up, parents were asked to report, via a short telephone survey, on their child’s intake of discretionary foods outside of school hours and on weekends to identify any compensatory nutrition behavior occurring outside of school hours. Measures were taken from the NSW Schools Physical Activity and Nutrition Survey [
To assess the foods packed in the lunchbox (premeal assessment) on a randomly selected school day prior to recess, at lunch, or during in-class vegetable and fruit breaks [
To assess the consumption of foods packed in the lunchbox (postmeal assessment), on the same day, students were asked to keep all unconsumed or partially consumed food items in their lunchboxes. Following all meal breaks, students were asked to place unconsumed or partially consumed items from their lunchbox onto the grid paper, and a second photograph of all remaining food was taken. Measures relating to consumption were based on the second photograph of the day being taken after all meal breaks had occurred and all uneaten food had been placed back into the lunchbox container. Consumption was calculated by subtracting the postmeal assessment from the premeal assessment.
Trained dietitians, blinded to group allocation, observed each school lunchbox photo in order to classify each food and drink item according to its SFC category and the serving size. All lunchbox photos were assessed by 2 dietitians working together to make a consensus decision on the analysis for each lunchbox. To further aid this process, decision rules were developed to ensure standardization of assessments. Differences in opinion between dietitians were resolved following consultation with a third dietitian assessor. Following the analysis of the premeal lunchbox photo, dietitians then analyzed the postmeal photo. Energy consumption was calculated by subtracting the energy content of foods and drinks remaining in students’ lunchboxes at the postmeal assessment from the energy content of foods and drinks in the lunchbox during premeal assessment (“foods consumed”).
We also assessed impact on engagement, as research suggests that improved nutrition correlates with greater school attendance, improved concentration, and higher academic achievement [
To ensure any reduction in energy intake occurring while at school did not result in compensatory intake outside of school hours (potential adverse event), parents were asked via a short telephone survey at baseline and at follow-up to report on their eldest eligible child’s intake of discretionary foods outside of school hours and on weekends. Measures were taken from the NSW Schools Physical Activity and Nutrition Survey [
It has been hypothesized that one potential adverse effect of encouraging healthier lunchbox swaps is increased family financial burden due to the potential higher cost of healthier products [
Analyses were conducted using SAS version 9.3 (SAS Institute) from January 2020 to June 2020. School and student characteristics were summarized for intervention and control schools. Summary statistics are used to describe all variables of interest. Students that resided in postcodes ranked in the top 50% of state postcodes based on the 2016 SEIFA [
The differences between groups in the primary and secondary outcomes were assessed using hierarchical linear (or logistic for binary outcomes) regression models. Models were adjusted for SEIFA, remoteness, and baseline values, and a random level intercept for schools was included to adjust for the clustered design of the study. Analysis followed intention-to-treat principles, where schools and students were analyzed according to their randomized treatment allocation. All statistical tests were 2-tailed with an α of .05. As specified in the study protocol [
According to our pilot results [
A sample of 94 schools was assessed for eligibility to participate in the study, and 91 were approached in order to obtain the quota of 32 consenting schools (35.2%). Consenting and nonconsenting schools were similar in geographic location, size, and school socioeconomic status, with the 32 consenting schools enrolling a total of 7212 students (or 5048 families). Of these, 3022 provided parental consent to participate in the lunchbox observation to evaluate the outcomes of the study (41.90%). From the 3022 consenting students, 2730 (1395 intervention and 1335 control) lunchboxes were observed at baseline and 2346 (1215 intervention, 1131 control) at follow-up, with the discrepancy being due to student absences and school events or excursions.
Sample characteristics of schools and students at baseline.
Characteristics | Intervention | Control | ||||
|
||||||
|
Allocation, n | 16 | 16 | |||
|
|
|||||
|
|
Urban | 8 (50.0) | 6 (37.5) | ||
|
|
Rural | 8 (50.0) | 10 (62.5) | ||
|
|
|
|
|||
|
|
Most disadvantaged | 13 (81.2) | 13 (81.2) | ||
|
|
Least disadvantaged | 3 (18.8) | 3 (18.8) | ||
|
Schools with greater than 10% Aboriginal or Torres Strait Islander student enrolments, n | 10 | 11 | |||
|
||||||
|
Allocation, n | 1216 | 1176 | |||
|
|
|||||
|
|
Female | 592 (50.04) | 550 (48.37) | ||
|
|
Male | 591 (49.96) | 587 (51.63) | ||
|
Mean age (years) | 7.88 | 7.68 | |||
|
|
|||||
|
|
Most disadvantaged | 938 (77.14) | 789 (67.09) | ||
|
|
Least disadvantaged | 278 (22.86) | 387 (32.91) |
aSocioeconomic status is based on SEIFA Index of relative socioeconomic disadvantage 2016: most disadvantaged = lowest quartiles of SEIFA; least disadvantaged = highest quartiles of SEIFA.
bInformation on sex missing for 72 students.
At 6-month follow-up, the difference between the intervention and control group in the mean energy (kilojoules) content of discretionary foods packed in school lunchboxes was –117.71 kJ (95% CI –195.59 to –39.83;
The mean total energy (kilojoules) packed in lunchboxes (–88.38 kJ; 95% CI –172.84 to –3.92;
Mean energy and percentage of energy from everyday and discretionary foods packed and consumed from student lunchboxes.
Outcome | Intervention | Control | Difference in energy between groups at follow-up, mean (95% CI) | |||||||
|
Baseline mean, (SD) (n=1216) | Follow-up, mean (SD) (n=946) | Baseline, mean (SD) (n=1176) | Follow-up, mean (SD) (n=886) |
|
|
||||
|
||||||||||
|
Primary outcome: lunchbox energy from discretionary foods packed in lunchboxes | 1214.86 (876.49) | 1156.77 (841.76) | 1067.38 (898.82) | 1105.06 (859.06) | –117.26 (–195.59 to 39.83) | .003 | |||
|
Lunchbox energy from everyday foods packed in lunchboxes | 1616.19 (628.34) | 1610.93 (624.41) | 1644.17 (621.73) | 1605.81 (610.02) | 32.85 (–31.61 to 97.31) | .31 | |||
|
Total lunchbox energy packed in lunchboxes | 2831.05 (927.81) | 2767.70 (873.52) | 2711.54 (962.33) | 2710.87 (878.44) | –88.38 (–172.84 to –3.92) | .04 | |||
|
||||||||||
|
Lunchbox energy from discretionary foods consumed from lunchboxes | 901.30 (745.60) | 876.70 (717.23) | 744.19 (717.20) | 802.75 (677.23) | –96.31 (–194.63 to 2.01) | .05 | |||
|
Lunchbox energy from everyday foods consumed from lunchboxes | 1270.85 (631.79) | 1282.56 (622.95) | 1304.69 (600.58) | 1341.72 (607.53) | –21.91 (–112.38 to 68.56) | .62 | |||
|
Total lunchbox energy consumed from lunchboxes | 2172.15 (895.82) | 2159.26 (810.78) | 2048.88 (853.84) | 2144.48 (743.22) | –117.17 (–233.72 to –0.62) | .05 | |||
|
||||||||||
|
Packed lunchbox energy from discretionary foods | 40.10 (23.31) | 39.04 (23.94) | 35.84 (23.69) | 37.90 (23.81) | –3.16 (–5.46 to –0.86) | .01 | |||
|
Packed lunchbox energy from everyday foods | 59.90 (23.31) | 60.96 (23.94) | 64.16 (23.69) | 62.10 (23.81) | 3.16 (0.86 to 5.46) | .01 | |||
Total cost (Aus $) of lunchbox items | 3.94 (1.35) | 3.91 (1.36) | 3.78 (1.38) | 3.78 (1.32) | –0.06 (–0.18 to 0.07) | .37 |
Mean school engagement measure by group at baseline and at follow-up.
Mean school engagement score | Intervention | Control | Difference in engagement between groups at follow-up, mean (95% CI) | ||||||
|
Baseline, mean (SD) (n=364) | Follow-up, mean (SD) (n=309) | Baseline, mean (SD) (n=299) | Follow-up, mean (SD) (n=241) |
|
|
|||
Behavior score | 4.12 (0.59) | 4.09 (0.62) | 4.11 (0.65) | 4.14 (0.66) | –0.05 (–0.15 to 0.04) | .24 | |||
Emotion score | 3.55 (0.91) | 3.33 (0.99) | 3.56 (0.92) | 3.40 (0.98) | –0.08 (–0.22 to 0.06) | .26 | |||
Cognitive score | 2.92 (0.87) | 2.80 (0.87) | 2.87 (0.83) | 2.83 (0.85) | –0.09 (–0.22 to 0.05) | .20 | |||
Total school engagement | 3.44 (0.66) | 3.31 (0.71) | 3.42 (0.68) | 3.35 (0.70) | –0.08 (–0.18 to 0.02) | .10 |
There were no differences between groups in the foods consumed outside of school hours, indicating no compensatory consumption of discretionary foods outside of care.
The total cost of lunchbox foods following the intervention did not differ between groups (–Aus $0.06; 95% CI –0.18 to 0.07;
This trial investigated the effectiveness of the SWAP IT intervention on the energy of students’ lunchbox foods, both packed and consumed, using an existing school communication app provided directly to parents. Relative to lunchboxes in the control group, the lunchboxes in the intervention group contained significantly less mean energy from discretionary foods corresponding to 117 kJ per day or a 600 kJ reduction over a school week. The SWAP IT intervention also resulted in a reduction in mean energy from discretionary foods that were consumed by students (96.31 kJ). The mean total lunchbox energy both packed and consumed was also significantly less in intervention lunchboxes, and the percentage of energy from discretionary foods decreased by 3.16%, while percentage energy from everyday foods correspondingly increased. The lunchbox energy coming from everyday foods that were consistent with dietary guidelines did not statistically differ between groups, indicating the change in total energy observed was primarily from a reduction in discretionary foods. These favorable nutrition outcomes occurred while the cost of packing a lunchbox remained stable across groups, indicating the changes made to lunchboxes did not result in additional costs. The intervention, however, did not result in changes to student school engagement at school.
Although it is challenging to make direct comparisons, the magnitude of reduction in energy from discretionary foods appears favorable compared to previous lunchbox interventions. Of the 10 included studies within a systematic review of lunchbox interventions conducted within the school and childcare environment [
To improve health at a population level, interventions shown to be effective under research conditions need to be scaled up to reach a large proportion of the population [
Although a reduction in energy from discretionary foods of 600 kJ per week may appear small at an individual level, at a population level, it has the potential to lower the risk of individuals being overweight or obese, result in a gain of health-adjusted life years, and make a significant contribution toward savings in health care costs [
The results of this trial should be interpreted within the context of its strengths and limitations. Study strengths include the experimental hybrid design, with randomized controlled trials being considered the gold standard for evaluating causal effects of interventions. The SWAP IT trial was also developed using behavior change theory and used direct observation and validated tools to assess lunchbox contents, which strengthened the ability of the study to accurately measure the true impact of the study outcomes. Although the effect size of the SWAP IT effectiveness trial was smaller than that of the previous pilot [
The SWAP IT intervention presents an effective digital behavior change solution to a large and long-standing public health problem of a high consumption of discretionary foods by children while at school. Given the significant impact on lunchbox food energy that has been demonstrated by the previous pilot trial and replicated in this effectiveness trial at a larger scale, the intervention provides an attractive option to policy makers to complement existing public health programs targeting the school nutrition environment. Following further evaluation to determine its implementation process, outcomes, and cost-effectiveness, models to further scale up and maximize the adoption of SWAP IT will ensure that a public health benefit can be realized.
Using the Behaviour Change Wheel process to map barriers to packing healthy lunchboxes with identified intervention functions and suitable behaviour change techniques (BCTs).
Food and drink items packed in lunchboxes.
CONSORT-eHEALTH checklist (V 1.6.1).
Consolidated Standards of Reporting Trials
New South Wales
Socio-Economic Index for Areas
School Food Checklist
World Health Organization
The authors wish to acknowledge the research assistants involved and the schools participating in this study, as well as the Department of Education, Hunter Region, for their contribution to the advisory group.
This project was funded by the NSW Ministry of Health, Translational Research Grant Scheme. The NSW Ministry of Health has not had any role in the design, data collection, analysis of data, interpretation of data, or dissemination of the study. The project also received infrastructure support from the Hunter Medical Research Institute. RS is supported by a National Health and Medical Research Council Translating Research into Practice Fellowship (#APP1150661) and a Hunter New England Clinical Research Fellowship; LW is supported by a National Health and Medical Research Council Career Development Fellowship (#APP1128348) and a Heart Foundation Future Leader Fellowship; SY is supported by a Discovery Early Career Researcher Award fellowship.
RS and AB led the development of this manuscript. RS, NN, SY, LJ, and LW conceived the intervention concept. RS and LW secured funding for the study. RS, NN, and LW guided the design and piloting of the intervention. RS, AB, LW, NN, LJ, and JW guided the evaluation design and data collection. CO developed the analysis plan. RS, AB, NN, LJ, JW, NK, NE, CO, AS, PR, CR, BS, MD, KR, BC, KG, and LW are all members of the advisory group that oversee the program and monitor data. RR, AW, NH, AB, LJ, and AC are all members of the project team that oversee the implementation and evaluation of the program. All authors contributed to developing the protocols and reviewing, editing, and approving the final version of the paper.
Authors RS, NN, LW, KG, NE, and JW receive salary support from their respective local health districts. Hunter New England Local Health District contributes funding to the project outlined in this protocol. None of these agencies were involved in the peer review of this grant. RS and NN are associate editors for BMC Public Health. All other authors declare that they have no competing interests.