The prevalence of chronic health conditions in childhood is increasing [1,2]. Chronic health conditions are defined as “any physical, emotional, or mental condition that prevented him or her from attending school regularly, doing regular school work, or doing usual childhood activities or that required frequent attention or treatment from a doctor or other health professional, regular use of any medication, or use of special equipment”[3].

Behavioral interventions can support the treatment and management of chronic health conditions and can be effective in improving symptom management, reducing physical disability, and improving mental health [4-6]. These outcomes are particularly important in childhood because they have implications for children across their lifespan [7-13]. Behavioral or behavior change interventions are sets of techniques that aim to change health behaviors [14]. For children with long-term health conditions, these interventions typically focus on adherence to medical treatment, education about the medical condition, and improving aspects of medical care [15]. A specific example is the management of diabetes via behavioral intervention; glycemic control can be improved by encouraging behaviors of blood glucose monitoring, selection of healthy food choices, attendance at routine clinical appointments, and adherence to insulin therapy or other medications [16]. Improving the management of chronic health conditions at an early age can lead to immediate health improvements, but it also lays the foundations for health across the lifespan of the patient [7]. As such, it is important that younger children (and their families) are supported to improve understanding of their condition and develop self-management skills [17].

Digital interventions can deliver behavior change interventions using mobile phones, smartphones, portable computers, desktop computers, the internet, wearable technology, and television [18]. This is an emerging and rapidly developing field of research, and the potential advantages include increased cost-effectiveness, anonymity for users, appeal to younger people, and the ability for recipients to access interventions anywhere and at their own pace [19-22]. There is a growing body of evidence to suggest that digital interventions are potentially effective for adults and adolescents with chronic health conditions; they have beneficial effects on improving knowledge, self-management, self-care, quality of life, medication use, symptom control, and health service utilization [23-30]. However, there are some potential disadvantages that may affect the uptake, attrition, and efficacy of interventions. Some individuals may not be able to access the intervention because of technical issues, illiteracy, or the cost involved in obtaining the devices. Negative attitudes toward technology may also create barriers to use, and this includes concerns about data security. A lack of strong therapeutic relationships may discourage users and reduce engagement and efficacy of interventions. These potential disadvantages [31,32] should be carefully considered when planning and designing digital interventions. Furthermore, there are limitations with the evidence base for digital interventions, with systematic reviews highlighting the need for clearer reporting and higher quality research [23-29].

Despite the increasing availability of digital interventions and a growing body of evidence for adults and adolescents, there has been inadequate attention to designing and delivering these interventions to children. Children have different developmental characteristics and needs, and the developmental stage of children should be considered when designing interventions [33,34]. To the best of our knowledge, there are no systematic reviews that specifically investigate digital interventions for the management of chronic health conditions in children (aged <13 years). Furthermore, reviews investigating digital interventions for young people with chronic health conditions typically do not include children aged below 10 years [35], or, if they do, only a minority of the interventions included in the reviews include children aged below 13 years [15,36-38], recognizing that there are “fewer interventions targeting…the extreme pediatric age ranges of early childhood and emerging adulthood” [16]. The reviews spanning childhood and adolescence note important differences between these age groups. Three separate reviews of internet and computer-based cognitive behavioral therapy (CBT) for mental health problems found different treatment effects for older and younger children. The reviews found some positive effects for adolescents, but concluded that treatment effects were smaller or more uncertain for younger children [36,37,39]. Similarly, a review of electronic health interventions for young people with long-term physical conditions concluded that effectiveness was uncertain at this time, especially in children aged <10 years [15]. One review acknowledged, “we could not take the developmental stage of the patients…into account. As evidence is mounting, this issue should be addressed in future trials” [17].

Therefore, this review aimed to explore digital interventions for the management of chronic health conditions in children aged between 5 and 12 years.

Behavior change interventions are often complex [40], which can pose a challenge when synthesizing the effects of these interventions [41]. Advances in behavioral science have provided taxonomies and coding systems that help identify specific characteristics or active ingredients associated with effective interventions [42]. This includes the behavior change techniques (BCTs) taxonomy [43], which presents 93 discrete BCTs, “observable, replicable and irreducible component of an intervention designed to alter or redirect causal processes that regulate behavior”[43]. In addition to understanding what is being delivered (BCTs), it is important to understand how the content is delivered; this can be categorized using the mode of delivery taxonomy [18]. Identifying the theoretical underpinnings is possible with a coding frame [44]. Using these BCTs, mode of delivery and theory taxonomies in systematic reviews may result in more optimal evidence syntheses and health care practice recommendations [41].

This systematic review aimed to investigate digital interventions for the management of chronic health conditions in children aged between 5 and 12 years. We used an inclusive definition of chronic health conditions that included both physical and mental health. Conceptually, behavioral interventions for physical and mental health conditions are the same; they are designed to change the child’s behavior to improve the clinical outcome. Furthermore, there is a strong overlap between physical and mental conditions; comorbidity of physical and mental health conditions is common [45], and many conditions involve both mental and physical health issues (eg, chronic fatigue syndrome or myalgic encephalomyelitis, pain, and obesity), thus developing integrated approaches toward mental and physical health is increasingly becoming a priority [46]. In this review, we aimed to answer the following questions: (1) Which of these interventions are effective in promoting behavior change for the management of the chronic health condition? (2) What are the characteristics of effective interventions, considering the following: recipients, what is being delivered (BCTs), how this content is being delivered (the mode of delivery), the theoretical basis, and the modality of the intervention? and (3) What are the users’ experiences of the digital intervention?

The review was prospectively registered in the Prospective Register of Systematic Reviews (PROSPERO) database.

We carried out a systematic search of relevant databases: Excerpta Medica Database (EMBASE), PsycINFO, Medical Literature Analysis and Retrieval System Online (MEDLINE), and the Cochrane Library (January 2019). The search strategy included keywords and Medical Subject Headings (MeSH) for (1) children aged between 5 and 12 years, (2) behavior change, (3) randomized controlled trials (RCTs), (4) digital interventions, and (5) chronic health conditions (we used a mixture of generic terms, ie, “Chronic disease,” and also search-specific terms, informed by the most common chronic illness in childhood; Multimedia Appendix 1) [47].

To be included in this review, studies had to fulfill the following criteria:

Titles and abstracts (stage 1) and full-text papers (stage 2) were independently double screened against the inclusion and exclusion criteria using the data management platform Rayyan (stage 1) and Covidence (stage 2). AB screened all papers, and CL, LS, and EB were responsible for the independent second screening. Reasons for exclusion were recorded at stage 2. Discrepancies at both stages were discussed and resolved in meetings by the reviewers. Papers were tracked using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram [50].

For data extraction, papers were reviewed independently by 2 researchers and conflicts were resolved in regular meetings (AB reviewed all papers, and CL, LS, and EB were responsible for an independent second review). Two researchers independently coded BCTs (EA and AB, a health psychologist and health psychology trainee, respectively). We extracted information that allowed us to answer the 2 primary research questions, as described in Table 1. If the full text did not contain the information needed, we made 2 attempts to contact the authors by email.

Due to the clinical and methodological heterogeneity, we synthesized data using narrative synthesis [51,52] to answer our research questions. We reported effectiveness based on whether interventions were deemed as very promising, quite promising, possibly promising, nonpromising, or unable to assess effectiveness, determined by change in the outcomes and the quality of the science (Table 1 defines these categories).

As all studies in this review were RCTs, the Cochrane risk of bias tool for randomized trials (RoB 2.0) [58] was used to assess the scientific quality of each study. Two researchers reviewed each paper, and the researchers then compared their quality assessment and resolved conflicts (AB reviewed all papers, and ML, LS, and EB were responsible for an independent second review). Following this, each paper was given a score of either low risk of bias, some concerns, or high risk of bias. Where available, we reviewed trial registries and published protocols. If needed, we also requested further information from the authors, including statistical analysis plans.

After deduplication, 2643 papers were identified from the database searches, of which 18 papers were identified as eligible for inclusion. Two of these papers reported on the same intervention; therefore, we identified 17 digital interventions for the management of chronic health conditions in children aged between 5 and 12 years. Figure 1 displays the PRISMA diagram.

The digital interventions targeted a range of chronic health conditions, including overweight or obesity (n=7), anxiety and preoperative anxiety (n=3), cerebral palsy (n=3), attention-deficit/hyperactivity disorder (ADHD; n=1), type 1 diabetes (n=1), asthma (n=1), and social-emotional problems (n=1). All the interventions included children of key stage 2 age (8-11 years), 13 included children of key stage 3 age (12 years), and 9 included children of key stage 1 age (5-7 years).

Table 2 details the characteristics of the population, intervention, and outcome data (Multimedia Appendix 2). This is presented by their potential effectiveness, based on outcomes and quality assessment (Multimedia Appendix 3).

No interventions were identified as very promising.

A total of 5 interventions were identified as quite promising; 3 of these were interventions targeting overweight or obesity. An intervention involving exergaming used Kinect and Xbox with additional components of Fitbit step count monitoring and parent-child telehealth sessions with a fitness coach. Compared with the control group, the intervention group showed an increase of 11.4 min of moderate-to-vigorous physical activity (MVPA) per day (95% CI 2.25- 20.55) at 6 months. However, there was no clear evidence of effect on the primary outcome; the reduction in BMI Z-scores was −0.08 (95% CI −0.16 to 0.003) at 6 months, which did not reliably meet the authors’ threshold for change (−0.09). An exergaming plus a family-based pediatric weight management program led to an increase of 8.0 (95% CI 0.5-15.4) min of MVPA per day at 4 months, with the trial powered to detect a change of 4.0 to 4.6 min of MVPA per day (MVPA was the primary and behavior change outcome) [54]. A further trial involved 4 training units (2 face-to-face and 2 via Facebook) plus weekly support through a parental WhatsApp group for 12 weeks. This led to a decrease in BMI Z-scores of 0.14 (95% CI −0.28 to −0.003) at 6 months, with the trial powered to detect a decrease of 0.24 (no behavior change outcome was available). The third and fourth quite promising interventions were both internet-delivered CBT for children with anxiety disorders, both offering completely web-based modules for parents and children in conjunction with web-based therapist contact [60,63,64]. Both led to an improvement in anxiety as assessed by the clinician severity rating, equating to an estimated change of −077 (95% CI −1.15 to −0.40) at 3 months [60] and −1.16 (95% CI −1.55 to −0.77) at 10 weeks [63].

In all, 3 interventions were identified as possibly promising. A PC game led to improved balance control in children with cerebral palsy on 2 of the 7 measures of balance at 3 months [66]. An internet-based serious game for ADHD led to an improvement in parent- and teacher-rated time management skills at 5 months, but no evidence of improvement on parent- and teacher-rated planning and organization skills or social skills. A tablet app that included an educational animated video, along with games for distraction and to encourage relaxation or breathing exercises for preoperative anxiety [67], led to reduced anxiety scores on the modified Yale Preoperative Anxiety Scale of −7.71 (95% CI −14.27 to −1.15) immediately after the intervention. Although there was evidence of an effect, these studies were limited in scientific quality. There was a lack of transparency around randomization processes, a combination of nonblinded participants, and the use of self-report measures, and none of these trials were prospectively registered.

Three interventions showed no promise; 2 of these were targeting overweight and obesity, 1 was the exergaming plus classroom curriculum, and the other was motivational interviewing delivered via one-way text messaging [68,69]. Neither lead to improvements in behavioral outcomes (screen time, physical activity, and diet) or the primary outcome (BMI Z-scores) at 6 and 3 months, respectively. The other intervention was a video game for social-emotional problems [70], which did not lead to changes in bullying perpetration behavior at 9 weeks.

Six interventions were pilot studies, and they only reported acceptability or feasibility data [72,73,76] or involved small sample sizes (6, 15, and 14) that were not powered to determine effectiveness [71,74,75]. Of these studies, 3 reported that there were no further plans for investigation [72-74] and 1 reported that a larger, fully powered trial was planned for the future [75]. Information on the remaining 2 studies is unknown [71,76].

Adverse events for each study are reported in Table 3. Three studies reported adverse events; these were not severe and or there were similar numbers in the intervention and control arms. Four studies monitored adverse events and reported that no adverse events occurred during the trial. Most studies (n=10) failed to capture adverse events.

We considered the 8 interventions that were classified as promising, quite promising, and possibly promising to represent promising interventions.

A total of 7 of the 8 interventions had a digital component for the child, and all the interventions involved the child in some capacity (either digital or human component). In all, 5 of the 8 interventions involved the parent in some capacity (either digital or human component).

Table 4 provides the definitions of the BCTs, and Table 5 provides a summary of the BCTs used in promising interventions.

All the promising interventions used more than one BCT. Digital components for the child typically included techniques coded into the following BCT categories: feedback and monitoring, shaping knowledge, repetition and substitution, and reward and threat (we note that none used threat, but this is the overarching BCT taxonomy category label). Digital components for the parent typically included goals and planning, social support, and natural consequences.

The most promising interventions were for overweight or obesity (3 studies) and anxiety (2 studies). All 3 of the promising overweight or obesity interventions included a face-to-face component for both the parent and the child. Two interventions included a digital component for the child, both using the BCT repetition and substitution. Only 1 intervention had a digital component for the parent.

Both promising anxiety interventions included digital and face-face elements, all of which involved both the child and the parent. Both interventions used the following BCTs in the digital component: goals and planning (child and parent components), shaping knowledge (child and parent components), feedback and monitoring (parent component), and associations (child component).

We acknowledge that there may have been more BCTs included in the intervention; however, we were unable to code these as they were not explicitly reported in the paper. Furthermore, it was often unclear as to whether the BCT was delivered to the parent or the child and by what means it was planned to take effect. In some cases, we believe that the BCTs were directed at the parent, with the parent then eliciting behavior change in the child. However, none of the papers addressed this level of complexity; they did not describe this mechanism of change nor did they include a parent behavior change outcome measure.

A total of 5 of the 7 interventions with child digital components used gaming features. All the parent digital components and 5 of the child digital components were guided. In all, 3 digital interventions involving parents and 1 digital intervention for the child were tailored.

Half of these papers reported the use of theory in the intervention: social cognitive (n=2) and CBT (n=2).

A total of 6 of the 8 interventions were first wave (purely behavioral) interventions, and 2 were second wave (cognitive-behavioral) interventions. There were no third wave interventions.

Only 3 of the studies included qualitative data on users’ experiences and views of the intervention [68,72,73]. One study evaluated the family experience in a preceding pilot study [77]. A table of the raw qualitative data and themes are available (Multimedia Appendix 4).

To the best of our knowledge, this is the first review to identify effective digital interventions for younger children, report the characteristics of promising interventions, and describe the user’s experience of digital interventions. Of the 17 eligible interventions, we only identified 5 that had a beneficial effect and had a low risk of bias; 3 targeted overweight or obesity, using exergaming or social media with additional human support, and 2 targeted anxiety, using web-based CBT with therapist support.

Characteristics of promising digital interventions included gaming features in the child digital component and having additional therapist support (guided digital interventions). Digital components for the child typically used the BCTs [43] feedback and monitoring, shaping knowledge, repetition and substitution, and reward. Most were purely behavioral interventions (first wave), with only a quarter using CBT (second wave) and none using third wave approaches; half of the interventions had a theoretical basis (social cognitive theory and CBT). Over 60% involved the parents in the intervention.

Only 3 papers used qualitative methods to explore the users’ experience of digital intervention. These studies reported the affordances of digital interventions, including ease of use, integration into daily life, and the ability to enhance communication with a health professional. However, a lack of personalization, technical problems, and cost issues posed challenges to families. The qualitative data indicated how the content (eg, language and concepts) and design could be improved for younger users.

We included a range of chronic health conditions, which enabled us to review a larger number of interventions and identify patterns or commonalties of promising interventions. Spanning health conditions makes these findings relevant to a wide audience of researchers working in the field of digital interventions. We focused on RCTs because they have the strongest study design and are most likely to be adopted in clinical care [78]. This review focused on the outcomes that were most important to our research question (behavioral outcomes) and most important for that particular study (the primary outcome). It was outside the scope of this paper to review all the possible outcomes, such as health status or symptoms of the disease, quality of life, and knowledge.

Guidance was followed on how to report effectiveness in narrative reviews [51]. We extracted a common statistic to show the size and direction of effect, and where possible, we placed results in the context of clinically meaningful change [79]. Strengths of narrative synthesis include richer exploration of more complex questions, exploring both effectiveness (aim 1) and what “might explain differences in direction and size of effect... how and why interventions have or do not have an effect” (aim 2) [51]. We increased the rigor of presenting characteristics of interventions by using established coding systems and taxonomies for BCTs [43], modality [56], mode of delivery [18], ages [53], and population type [80]. We also considered parental and child components separately, which is important for this younger population.

A limitation of this review is that we only included RCTs. Although observational studies and nonrandomized trials could have provided additional information on the characteristics and effectiveness of digital interventions for this population, we excluded these study designs as they have a greater potential for risk of bias [81]. Although we believe that our search strategy (which included the terms “Randomized Controlled Trial,” “Trial,” and “Clinical Study”) was broad enough to identify different RCT designs, it is possible that we may not have identified some designs specifically used in the evaluation of digital interventions, such as micro randomized trials. We also restricted our search to papers published after 2015. We chose this strategy as digital health is a rapidly changing field, and recently conducted studies are likely to be the most relevant. We excluded studies that included our target age group (5-12 years) but also included older and younger children (eg, 5-18 years). Although it is possible that these studies could have been stratified by age, it was not feasible to contact authors to request these stratified data. As expected, the broad scope of this review led to heterogeneity across studies (in terms of population, intervention, and outcome), meaning formal meta-analysis was not possible; therefore, we selected the most appropriate method, narrative synthesis. Although potential limitations to narrative synthesis include a lack of transparency and reproducibility and being subject to author interpretation [52], we mitigated this by prospectively registering our protocol, with specified outcomes, and following narrative synthesis guidelines [51]. To identify the characteristics of effective interventions, we reviewed both quite promising and possibly promising interventions and acknowledged that the possible promising interventions were of poorer scientific quality. Due to the small number of qualitative studies, we did not conduct full meta-ethnography [82] to synthesize qualitative data, and we did not undertake critical appraisal. However, to increase the transparency of our qualitative summary, we reported the raw data from the papers along with the themes developed by us.

This work highlights characteristics that may be beneficial when developing digital interventions for younger populations. The finding that purely behavioral interventions (first wave, not including cognitive components) are common in promising interventions is consistent with developmental theory; children tend to be limited to concrete thought [57]. There were fewer CBT (second wave) interventions, possibly because elements of CBT require abstract thinking, which may be beyond the cognitive abilities of children aged <8 years [57]. Similarly, third wave interventions also include abstract concepts such as metacognition. The lack of third wave approaches may also be explained by the fact that this is a relatively new approach for children. As such, concrete interventions focused on behavioral recommendations may be more appropriate [84]. Caregivers are commonly involved in promising interventions. This is also consistent with developmental theory, which highlights the important role of caregivers in structuring the child’s environment and shaping the child’s behavior [84,85]. Gaming features have been used in many promising interventions. Digital games can be adapted to the developmental level and can effectively engage younger users in medical education and treatment, as they are typically more visually oriented, involve appealing exploration, and are perceived as fun [17]. Consistent with the literature, guided interventions were common in promising interventions and have been identified as a moderating factor that can influence therapeutic outcomes and engagement [86,87].

Guidelines encourage standardized reporting of interventions to ensure transparency and reproducibility [43,88]. On the basis of our findings, we have developed recommendations for increasing the clarity of interventions with parental involvement. Interventions with both a child and a parent recipient have a complex model of behavior change; it is likely that the therapist aims to shape the behavior of the parent, with the expectation that the parent will change the behavior of the child. Studies in this review failed to explicitly differentiate the BCTs used by the therapist for parental behavior change and the behavior techniques used by the parent for child behavior change. Furthermore, none of the studies in this review captured a parental behavior change outcome measure, when this may be on the causal pathway to the child’s behavior change. This recommendation is consistent with guidelines on process evaluation; outcome measures should be used to test the causal mechanism of the intervention. Figure 2 illustrates the relationship between therapist, parent, and child, detailing our recommendations for how these interventions could be conceptualized and reported.

The low number of promising interventions demonstrates the need to better understand the perspective of those receiving interventions. Few studies have conducted qualitative research to explore the user’s experiences. Qualitative methods, such as the person-based approach [89], base the development and evaluation of digital interventions on an in-depth understanding of the perspectives of the people who will use the intervention. This can lead to interventions that are more meaningful, usable, feasible, and engaging in improving uptake and adherence and maximizing effectiveness [89].

Of the 17 interventions, we only identified 5 with promise (those with a beneficial effect and low risk of bias). Using qualitative methods during digital intervention development and evaluation may lead to more meaningful, usable, feasible, and engaging interventions, especially for this under-researched younger population. Promising interventions were exergaming and social media for obesity or overweight and a web-based CBT platform for anxiety. We identified characteristics that could be considered when developing digital interventions for younger children: involvement of parents, gaming features, additional therapist support, behavioral (rather than cognitive) approaches, and particular BCTs (feedback and monitoring, shaping knowledge, repetition and substitution, and reward). We suggest a model for improving the conceptualization and reporting of behavioral interventions involving children and parents.