Poor diet characterized by excessive intake of foods high in sugar, salt, and sodium and low intake of core foods such as fruits and vegetables is associated with a higher risk of chronic disease [1]. The consumption of foods high in saturated fat, sugar, or salt exceeds optimal levels worldwide [2], and in countries such as the United States, the United Kingdom, and Australia, excessive consumption of such nutrients has been reported across all age groups [3-5]. In Australia, foods high in saturated fat, sugar, or sodium, and low in nutrients such as fiber (from here on referred to as discretionary foods) [6], contribute up to 38% of the total daily energy intake of a child aged 4 to 8 years [5]. Given this, international and national dietary guidelines recommend limiting the consumption of discretionary foods [7-10].

Early childhood education and care (ECEC) services constitute an ideal setting for reaching and engaging families to influence lifelong eating habits [11,12]. In the United States, the United Kingdom, and Australia, up to 80% of children attend some kind of formal ECEC service in the year before school [13-15], with an average attendance of 2 to 3 days per week [15]. A significant number of ECEC services in these countries require children to bring food from home in a lunch box (approximately 30%-50%) [13,14,16]. Australian studies report that between 50% and 60% of lunch boxes include discretionary foods [17,18], with one reporting an average of 2 serves included per lunch box [18]. As such, there is potential to improve children’s diet by targeting the packing of discretionary foods in lunch boxes.

A systematic review of interventions to improve the contents of lunch boxes identified just 3 randomized controlled trials that examined interventions to improve the packing of discretionary foods within the ECEC setting [17,19,20]. Of the 3 trials, 2 reported significant results for reducing the packing of discretionary foods [17,19]. Both were intensive multicomponent interventions, including teacher education, child curriculum, and face-to-face education sessions with parents [17,19]. Despite the success of these interventions, evidence suggests that the sustainability and scalability of such intensive designs may be limited [21,22]. In particular, face-to-face parent-related components have been identified as difficult for parents to attend because of time and travel burden [23] and challenging for ECEC services to implement [21].

The use of digital health–delivered interventions has been proposed as a way of overcoming barriers to engaging parents in nutrition interventions, as well as enhancing scalability [24-26]. Digital health interventions (DHIs), in particular the use of mobile phone apps or mobile health (mHealth) interventions, have been identified as an effective way of providing health-related information to parents [27] and influencing child health outcomes [25,28-30]. In addition, mobile apps designed to facilitate parent ECEC service communications are increasingly available for use by ECEC services in Australia, offering an alternative to face-to-face delivery of nutrition interventions by drawing on existing digital platforms [26]. Using existing apps to deliver nutrition interventions has also been recommended to capitalize on technologies developed by experts in user-centered design and with established commercial appeal [26]. Therefore, integrating a lunch box intervention into an existing ECEC parent communication app shows promise as a novel way of delivering digital content and engaging with parents.

To the authors’ knowledge, no trials have been conducted to assess whether mHealth interventions can support the packing of healthy lunch boxes for children attending ECEC services. To address this gap in research evidence, the aim of this study is to determine the efficacy of using an mHealth intervention versus no intervention in ECEC services to reduce the packing of discretionary foods in children’s lunch boxes. We hypothesize that as a result of the intervention, lunch boxes of children with parents in the intervention group would achieve a mean reduction of 123 kJ from packed discretionary foods relative to the control. Additional outcomes include the intervention’s impact on children’s consumption of food, parental use of the app, intervention fidelity, and other process evaluation measures.

Ethical approval was obtained from the Hunter New England (HNE) human research ethics committee (06/07/26/4.04) and ratified by the University of Newcastle ethics committee (H-2008-0343). The trial is reported in line with the CONSORT-EHEALTH (Consolidated Standards of Reporting Trials of Electronic and Mobile Health Applications and online Telehealth) reporting guidelines (Multimedia Appendix 1) [31], and a protocol has been previously published [32].

The study used a cluster randomized controlled trial design and targeted the parents of children attending center-based ECEC services (ie, long day care and preschools). The ECEC services were located in the HNE Local Health District of New South Wales (NSW), Australia. ECEC services were randomized to a 10-week intervention group or no intervention control group. Within NSW, long daycare services can provide center-based education and care for children aged from 6 weeks to 6 years for ≥8 hours per day. Preschools typically enroll children aged between 3 and 6 years and provide care for 6 to 8 hours per day [33]. In 2016, approximately 920,370 people were reported to live in the HNE area, with 51,900 of these being children aged 0 to 4 years [34]. The area includes major metropolitan centers and inner regional communities, with a small proportion (14%) of people in remote communities [35].

The sampling frame comprised ECEC services from the region that required parents to provide food for consumption at the service (ie, lunch box services). The list of ECEC services that required parents to provide food was obtained from a government database and represented approximately 53% of all ECEC services in the region [16]. Of these services, existing users of the app required for the intervention and those not using any app (ie, services able to commence using the intervention app) were identified using previous, unpublished data on app use by services in the region.

Recruitment for the trial occurred in 2 phases. Initially, ECEC services were eligible to participate in the trial if they enrolled children aged 3 to 6 years and were existing users of the designated parent communication app. As this eligibility criterion did not result in adequate ECEC services being recruited, phase 2 recruitment extended the eligibility criteria to include ECEC services not yet using the app but willing to commence using the app for the trial. For both phases, recruitment involved research assistants posting and emailing information statements and consent forms to ECEC services outlining the study, data collection procedures, and requesting participation. Written consent was obtained from service managers. Parents or carers (hereafter referred to as parents) of children aged 3 to 6 years were eligible to participate if their child attended during the designated days of data collection and if they used or indicated a willingness to download the intervention app on the consent form. To obtain parental consent for participation in the study, ECEC service staff distributed hard copies of parent information statements and consent forms. Parents could also consent on the day of the data collection.

Before baseline data collection, ECEC services were randomly allocated to the intervention group or no intervention control group in a 1:1 ratio by a statistician independent of the trial using a computerized random number generator. Before randomization, ECEC services were stratified by rural location and socioeconomic status (SES) of the service, as evidence indicates that these factors are associated with family dietary intake [36,37]. To ensure equity of access to the intervention, ECEC services were also stratified by those with high numbers of Aboriginal child enrollments (defined as those with >10% Aboriginal children enrolled). Stratification by this number of factors was deemed appropriate for the sample size [38]. Owing to the nature of the intervention, ECEC services and parents were not blinded to the intervention; however, outcome assessors were blinded to the service allocation.

Details of the intervention and its development, including the application of theory, cultural, and other stakeholder consultation processes, have been published elsewhere [32]. Briefly, the intervention was based on an mHealth lunch box intervention originally designed for primary (elementary) schools by the research team, which reported promising pilot data on the packing of discretionary items in school [39,40]. Similar to the original intervention, the behavior change wheel (BCW) and COM-B (Capability, Opportunity, Motivation, and Behavior) model were used to inform the current intervention [41]. The BCW framework is based on 19 theories of health behavior change and facilitates the systematic development of behavior change interventions. The COM-B model is the behavior system behind the BCW framework, which supports the identification of essential conditions for changes, including capability, opportunity, and motivation [41]. The resulting mHealth intervention, SWAP IT for Childcare, comprised 7 behavior change techniques targeting 8 barriers to packing healthy lunch boxes. (Multimedia Appendix 2). Barriers were identified from the literature and formative interviews with a convenience sample of parents from 3 local ECEC services. The adaptation process was then guided by the Framework for Reporting Adaptations and Modifications–Enhanced (Multimedia Appendix 3) [42].

The app used for the intervention (Skoolbag) was an existing app designed to be available for use by schools or ECEC services to communicate with parents. Commonly, ECEC services use the app to share parent newsletters, reminders, and other service information. The app had not been previously used to deliver nutrition interventions in the ECEC setting.

The SWAP IT for Childcare intervention comprised 3 components.

The ECEC services allocated to the control group received no intervention. ECEC services had access to the app to use for their own communications but did not receive any of the intervention strategies or content.

This study aimed to recruit 390 children from 18 ECEC services. Allowing a 15% attrition rate, this would enable the detection of a mean difference of 123 (SD 200) kJ in the primary outcome (kilojoules from discretionary foods), with an α of .01 (adjusting for multiple outcomes using Bonferroni adjustments) and an estimated intraclass correlation coefficient of 0.1 with 80% power. Such an energy reduction would represent approximately a quarter of a serve reduction in packed discretionary foods and could be expected to result in the detection of approximately 2.2 g less sugar, 0.6 g less saturated fat, and 44 mg less sodium [32].

A statistician independent of the study performed all analyses using SAS (version 9.4) statistical software. Descriptive statistics were generated for service, parent, and child characteristics. The Australian Statistical Geography Standard Remoteness Structure [51] was used to classify parent residences as a major city, inner regional city, or outer city location. This classification was included as health disparities, including nutrition-related risk factors, exist between major city and inner regional cities (characterized by shorter distances by road to access services) and outer city locations (characterized by longer distances by road to services) [52].

A measure of between-group differences in primary and secondary outcomes was assessed for paired data using 2-level hierarchical linear regression models. Models were adjusted for potential ECEC service level clustering through a service random effect and controlled for baseline service EPAO score and if ECEC services were existing app users before the trial. An intention-to-treat analysis was undertaken [53], with multiple imputations undertaken for primary and secondary outcomes using the SAS MI and MIANALYZE procedure. Complete case analysis and prespecified subgroup analyses by child gender and SES were undertaken by adding a group by subgroup interaction fixed effect. SES was determined by parent postcodes and classified as being in the top or bottom 50% of NSW according to the Socioeconomic Indexes for Areas [54]. A sensitivity outcome analysis, modeled similarly to the complete case analysis, was undertaken whereby the effects of the intervention were assessed in parents known to have only downloaded the app (181/400, 51%). In addition to the analysis prespecified as part of the trial protocol [32], an exploratory analysis was performed to explore any association between how often the service routinely used the app (mean hours per week) and the parent viewing rate of within-app messages using Spearman rank order correlation.

The combined consent rate of ECEC services was 45% (18/40; 40 services approached to reach the required 18 services). A service in the intervention arm ceased using the required app and withdrew from the study (before baseline data but after random allocation). Within the 17 consenting ECEC services, 400 parent-child dyads consented to participate in the baseline data collection. The mean parent consent rate was 51%. Data on packed lunch box contents were collected for 88.8% (355/400) of the children at baseline and for 84.3% (337/400) after the intervention. Consumption data were collected for 88% (352/400) of children at baseline and for 84.3% (337/400) after the intervention (Figure 1). The number of parents from the intervention group completing the web-based survey after the intervention was 21.6% (41/190). Parents who completed the survey were less likely to have a Technical and Further Education certificate or diploma (P=.04) or be from an inner regional city. (P=.04). Postintervention service pen and paper surveys were completed by all intervention ECEC services.

Baseline characteristics of ECEC services, parents, and children with lunch box data at baseline are reported in Table 1. Characteristics were similar across groups; however, the intervention group had a higher proportion of parents located in the outer regional areas than the control group. As the intervention was delivered at the cluster level rather than at the individual level, geographical remoteness was accounted for by stratifying services by this factor during randomization and not controlled for as part of the analysis. Before the intervention, 12% (2/17) of ECEC services in the intervention group were existing users of the app, and 35% (6/17) of ECEC services had not been using any app.

The results for the primary outcomes after multiple imputation and for complete case analysis are presented in Table 2. There was no significant difference between groups for the primary outcomes (kilojoules from discretionary foods and kilojoules from saturated fat, free sugars, and grams of sodium from all foods) packed or consumed. In addition, there was no difference in secondary outcomes, including packed or consumed serves of any core food groups or discretionary food serves between groups (Table 3). Data regarding the nutrients and food groups consumed are presented in Multimedia Appendix 5. The sensitivity analysis, which included only those parents known to have downloaded the app (181/355, 51%), also did not detect any statistically significant differences (Multimedia Appendix 6). The subgroup analysis by gender and SES demonstrated no significant differences between groups (Multimedia Appendix 7).

To our knowledge, the SWAP IT Childcare trial is the first trial to evaluate the efficacy of an mHealth intervention to improve lunch box contents in ECEC settings internationally. However, contrary to our hypothesis, the intervention had no impact on the primary outcome of the number of kilojoule from discretionary foods packed in children’s lunch boxes. In addition, process evaluation found low parent app downloads and message viewing rates.

When looking to compare the findings of the trial with other published studies, we were unable to identify any other healthy eating–focused mHealth interventions or DHIs targeting parents and children through ECEC settings. Therefore, we have primarily compared the findings with those of the original SWAP IT trial publications conducted in the primary school setting [39].

Although the results of our parent survey data indicated that delivery of a lunch box program via the app was considered appropriate, the mean viewing rate of the within-app message was only 26% (SD 14.9). Furthermore, only 61.8% (102/165) of parents in the intervention group were known to have downloaded the app, leaving 38.2% (63/165) of parents unexposed to program content. Such process data are in contrast to findings reported in a SWAP IT pilot study undertaken in a primary (elementary) school setting [39]. The intervention in the SWAP IT pilot study was similar to this study, including push notifications and within-app message content; however, it also included the distribution of resources to parents, policy, and classroom resources [39]. The pilot study reported higher parent app downloads (89%) and message views (between 35% and 120% over the 10-week intervention) and was able to demonstrate a reduction in the packing of kilojoules from discretionary foods (−221 kJ; P=.08, nonsignificant) [39].

Contextual differences between the ECEC and school settings in the study region may potentially explain the differences in the findings between these trials. In the school trial, all schools were existing users of the app used to deliver the SWAP IT program compared with just 2 ECEC services in the intervention arm of this trial. In addition, most parents were existing users of the app (355/400, 88.8%) [39]. Greater integration and preprogram use of the app may have meant that parents in the school trial were more likely to be routinely accessing the app to receive information and, in doing so, were more likely to access the SWAP IT program content. This premise is supported by our exploratory analysis, which found a positive nonsignificant association between how much time the ECEC services used the app for general communication with parents and the SWAP IT parent message viewing rate. In addition, it is possible that the inclusion of supportive ECEC service–based strategies targeting nutrition curriculum and policy, similar to that undertaken to the trial in schools, may have been useful in increasing the impact of the intervention.

Our ability to detect an effect on discretionary food packing may have also been limited by the lower than anticipated baseline packing of discretionary foods in the participating ECEC services [48]. The trial baseline data found ECEC child lunch boxes contained only a mean of 1.33 (SD 1.16) serves of discretionary foods [48]. This was considerably lower than the mean of 2.5 serves per lunch box reported in schools in the same region [55]. Although the serves of discretionary foods packed in the ECEC services exceeded recommended serves per day for our age group (ie, 0-0.5 serves) [6], the lower number of serves packed at baseline possibly limited scope for further improvement without a more intensive intervention approach. In contrast, a large scope for improving vegetable packing and consumption in children attending ECEC services was evident from the trial baseline data (only 0.3 mean vegetable serves packed and 0.1 serves consumed) [48]. As such, future iterations of the intervention would benefit from the inclusion of specific strategies to improve vegetable packing and intake, along with continued efforts to minimize discretionary foods.

The intervention’s lack of impact on parent packing of discretionary foods may have also been because of the decline in the rate of program message opening over the course of the trial. The pattern of the continued drop-off in use observed in our study is similar to those reported in many mHealth and DHI interventions [39,56-59]. The drop-off in engagement occurred in the trial despite the inclusion of several features explicitly designed to enhance engagement (user consultation, use of behavior change techniques, reminders, and information being provided from a credible source) [60,61] and the addition of a specific intervention strategy to prompt ongoing use (service manager delivered support messages). Additional factors found to be associated with sustained DHI use in the literature include perceived ongoing relevance, novelty value, tailoring, and self-monitoring [60,62]. Further qualitative exploration of parents’ reasons for low engagement and consideration of additional evidence-based features that could be added to the program to enhance engagement would be beneficial.

The strengths of this study include explicit mapping of behavior change theory to design the intervention, use of objective gold standard measures to assess lunch box contents, high follow-up rate, comprehensive process measures, and randomized trial design. A key limitation of the trial is that the app analytics could only capture use data at the service level, meaning that the true proportion of parents who accessed the messages is unknown (as message views may have been from parents from the service but not participating in the evaluation component of the trial). This limited our ability to conclusively determine whether the null findings were generally more likely to be a result of poor access, the program content itself, or a combination. This limitation extends to our additional analysis of only those parents who were known to have downloaded the app. Finally, given the low survey response rate (41/190, 21.6%) and differences in characteristics in survey completers compared with the overall parent sample, our acceptability survey data may have been influenced by response bias. The generalizability of the findings to ECEC services outside of the study region may be limited.

The intervention failed to decrease energy from discretionary foods in children’s lunch boxes. However, the use of apps was rated as an appropriate modality for delivering information related to the packing of healthier foods in lunch boxes. Process data suggest that the lack of impact may have been because of factors associated with the implementation of the intervention, such as the low parent uptake and use of the app, as well as the lower than expected levels of packing of discretionary foods at baseline. Given these limitations, the current feasibility of using mHealth interventions to target parent packing of healthier foods in lunch boxes is still uncertain. Future trials should invest time in collecting formative data on how ECEC services and parents currently use and engage with different digital platforms to best identify the ideal technology to deliver parent-focused nutrition interventions in this setting. In addition, given drop-offs in use over time, exploration of parent reasons behind poor ongoing use will also assist in enhancing program design and content to improve engagement.