This umbrella review was registered in the PROSPERO (International Prospective Register of Systematic Reviews; CRD42024510602). The search strategy was amended from what was originally specified at registration and in the protocol. The search was performed by following the PRIOR (Preferred Reporting Items for Overviews of Reviews) [15] and the PRISMA-S (Preferred Reporting Items for Systematic Review and Meta-Analyses Literature Search Extension) [16] checklists (Checklists 1 and 2).

The present review is part of a larger systematic review (SR) protocol, which includes studies identifying digital interventions aimed to improve different lifestyles, such as PA, dietary behaviors, migraine, or sleep patterns, within the DARE (Digital Lifelong Prevention) Project. The DARE Project is an initiative funded by the Italian Ministry of Universities and Research within the National Recovery and Resilience Plan, aimed at using digital technologies to create a data-driven health care ecosystem that can contribute to promoting healthy lifestyles and disease prevention and cure.

A systematically conducted overview of reviews and meta-analyses was chosen as the best design for the search and is hereafter referred to as an umbrella review. An umbrella review can be viewed as a way to obtain evidence from different existing reviews that address a broader question and can be useful for comparing interventions and developing guidelines, thus providing a very high level of evidence [17,18]. As a huge number of studies have been conducted and published in the literature on this topic, most of them lacking methodological rigor, the present search was limited to capturing SRs and meta-analyses that collected experimental studies, that is, randomized controlled trials (RCTs). SRs included by the authors followed a strict, peer-reviewed protocol to ensure all available evidence was considered impartially, involving a comprehensive literature search, critical appraisal of study quality, and statistical synthesis of results to provide a clear, evidence-based answer.

The review was conducted on 5 databases: Scopus, PubMed or MEDLINE, Web of Science, the Cochrane Database of Systematic Reviews via the Cochrane Library, and SPORTDiscus via EBSCOhost. The search was conducted in November 2025, with record publication dates restricted to the years 2018‐2025, informed by evidence that research on second-generation technologies (such as smartphones and wearables) has risen sharply since 2013 [19]; a 2018 review would therefore be expected to encompass a larger set of previously used tools.

A manual search was conducted in parallel by browsing the reference lists of all included reviews, by searching for information on the websites, and from gray literature. For papers not freely available on the internet, the authors were contacted and asked to provide the full-text versions.

Keywords used for the search included a combination of terms addressing the type of digital tool and the main PA outcomes (Multimedia Appendix 1). In detail, the type of digital tools searched included apps, wearables, social media, messaging, exergames, digital assistants, and different devices such as pedometers or accelerometers (Multimedia Appendix 1). No language restrictions were applied to the search strategy. Filters used were the range publication year from 2018 to 2025, the study design being an SR or meta-analysis, and the age range between 6 and 65 years. With regard to age, a comprehensive search for the age ranges 6‐65 years was performed first, as this was the target of the authors’ pilot study within the DARE Project, and afterward, studies targeting only children and adolescents (6‐17 y) were selected for the purposes of this study. The decision to conduct separate analyses for younger populations and adults was motivated by important age-related differences in the use of digital technologies. Young people tend to engage with digital tools that are distinct from those commonly used by adults. Moreover, these tools may influence behaviors in different ways across age groups. In particular, the impact of digital technologies on PA practice may vary according to developmental stage. For these reasons, age-specific analyses were considered necessary to ensure a more accurate interpretation of the findings.

References were managed and duplicates removed using Zotero (version 6.0.37, Corporation for Digital Scholarship). The selection process, with a first screening of titles and abstracts and a second screening of full texts, was conducted by 2 independent reviewers. Possible divergences were solved by discussion between the two reviewers or with the intervention of a third reviewer.

The PICOS (Population, Intervention, Comparison, Outcome, and Study Type) criteria were used to derive the eligibility of the studies.

For the population, we selected healthy human participants in the age range of 6‐17 years. Healthy participants are those who are not diagnosed with any disease or who do not have disabilities that could affect PA outcome assessment. Reviews targeted to overweight or obese participants without another declared disease were retained. Moreover, populations aged 6‐65 years were initially included in the review, based on the needs of the project; afterward, studies targeting only populations aged between 6 and 17 years were selected for the present review.

For intervention, we had interventions using digital approaches to increase PA. Such digital approaches should have included the use of one or a combination of the following: IT devices (eg, smartphone, tablet, and PC); wearable activity trackers (eg, accelerometers and pedometers); applications (eg, stand-alone apps or apps associated with wearable devices); social media tools (eg, chatbots or conversational agents, and social networks); messaging tools (eg, email, SMS text messages, and podcasts); web-based tools (eg, websites, and video chats or video sharing); exergames, active video games, or gamification approaches; and synchronous or nonsynchronous PA coaching or training.

For comparison, we use any other intervention or no intervention.

For the outcome, we used PA as the primary or secondary outcome.

For study type, we used SRs and meta-analyses including RCTs. The following reviews were excluded: reviews of reviews or meta-analyses; scoping or narrative reviews; those not reporting PA-specific outcomes (eg, those reporting only SBs, only dietary or nutrition outcomes, or only weight-related outcomes); those focusing only on wearables that cannot be applied to huge populations for economical or organizational reasons (eg, virtual reality); those focusing on individuals (telemedicine); studies where the intervention was not delivered by at least 1 digital tool; papers where the digital component was used only to “measure” PA and not as a means to increase PA; those reporting only nondigital self-reported tools (such as paper questionnaires) for the assessment of PA; interventions for secondary or tertiary prevention (eg, targeted to individuals with preclinical signs); studies targeted to specific groups, such as smokers or pregnant women.

The data from the selected reviews were extracted by the same 2 independent reviewers that conducted the selection process, and the following information was reported: first author and year of publication; document type; design of studies included in the review; country; age group; special population; date range of the search; number of databases searched; digital device for intervention delivery; digital component for intervention delivery; focus of intervention or digital tool; outcomes; number of total studies included in the review; number of total RCTs in all reviews; number of RCTs of interest; and inclusion criteria, first author, and publication date of RCTs describing tools of interest.

For the primary studies identified within each review, the following information was extracted: first author and year of publication; country; age group; special population; school attendance; sample dimension and type; intervention duration, arms, or name; setting; focus task; theoretical foundation; digital component or components of intervention; nondigital component or components; digital device for intervention delivery; intervention group (IG) and control group (CG); device-based or self-reported PA measure; PA objective measurement tool; fitness tests assessed; other assessed measures; and effectiveness on PA or SB outcomes. Regarding outcome effectiveness, an intervention was classified as partially effective when positive effects were observed for some, but not all, PA outcomes.

Data regarding the characteristics of the different reviews and of the RCTs selected are presented as numbers and percentage frequencies. Effectiveness data were stratified by the type of device-based PA outcome and by the characteristics of the interventions. Differences between groups in relation to their effectiveness have been estimated through the Pearson chi-square test. All statistical estimates were conducted using STATA MP (version 12.1, StataCorp, 2011) software. As a meta-analysis was not feasible given the design of this review, heterogeneity between studies was explored using qualitative and descriptive methods combined with narrative synthesis. Sensitivity analyses were qualitatively performed as well to understand differences in populations, interventions, and outcomes that may explain variability in findings. Considered factors for sensitivity analyses were the quality of included reviews and RCTs, the study population characteristics, the intervention features, and the outcome measurement.

The extent of overlap of RCTs across the included SRs and meta-analyses was examined using the corrected covered area (CCA) approach [20]. As STATA does not provide a dedicated command for computing the CCA, a citation matrix was constructed, with primary studies listed in rows and the different SRs or meta-analyses in columns. A custom STATA script was developed to automate the calculation according to the formula: CCA = (N − r) / (r×c − r), where N represents the total number of publications, r the number of unique studies, and c the number of reviews or meta-analyses. The resulting values were converted to percentages by multiplying by 100. A CCA of 0% indicates no overlap (each review or meta-analysis includes only unique RCT); 0%‐5% reflects slight overlap; 6%‐10% moderate overlap; 11%‐15% high overlap; and values above 15% indicate very high overlap. A value of 100% denotes complete overlap, meaning that all reviews or meta-analyses include the same primary studies.

The methodological quality of included reviews and meta-analyses reporting randomized controlled interventions was assessed using the AMSTAR 2 (A Measurement Tool to Assess Systematic Reviews) tool [21]. This tool allows a quality assessment based on 7 critical and 9 noncritical domains, rather than a total score, generating an overall rating of “high” quality (those reviews or meta-analyses that meet 7/7 critical domains), “moderate” (6/7), “low” (those meeting all but one critical domain and missing few noncritical domains), and “critically low” (not meeting multiple critical domains).

The methodological quality of each included RCT was evaluated using the 5 domains of the revised Cochrane Risk of Bias 2 (RoB 2) tool for RCTs [22]. Overall judgments for individual studies were derived according to the RoB 2 algorithm, categorizing studies as having a low risk of bias, some concerns, or a high risk of bias. A total of 2 reviewers conducted the assessments independently, and any discrepancies were resolved through consultation with a third reviewer.

Figure 1 shows the flow diagram of the search.

The database search included a total of 726 records, which were added to the 100 records identified by the manual search. After excluding 136 duplicates, a total of 690 reviews were screened by title and abstract, and subsequently, the full texts of the selected 216 records were examined. A total of 154 studies were eligible, but they included all-age populations. A further selection finally included a total of 48 SRs and meta-analyses targeted specifically to children and adolescents [23-70].

The present umbrella review provides an overview of SRs and meta-analyses that collected RCTs aimed at evaluating the effectiveness of digital approaches to increase PA in healthy children and adolescents. It summarizes and critically evaluates the digital intervention components, as well as the possible determinants and methodological limitations of their effectiveness.

Slightly less than half of the interventions reported either total or partial effectiveness of the digital tools in increasing device-based PA outcomes. A wide variation in study designs, target age groups, intervention objectives, digital devices, and intervention components was identified across the reviews and trials.

Effectiveness over the population target of the present review appears to be influenced by multiple factors, including the focus task of the intervention, the type of digital component used to deliver the intervention and its delivery device, the device used for the objective measurement of PA, some specific population target, and other characteristics such as the geographic region, the intervention setting, and the device brand.

Nondigital components were used in slightly more than half of the interventions, and they were mostly goals, rewards, and incentives. Previous works have shown that multicomponent interventions, including nondigital components such as counseling for children and parents, class seminars, and health promotion curricula, could have a positive impact on PA outcomes [159]. Therefore, different communication channels should be combined with digital tools because they are based on the possible behavior change [164].

Most SRs or meta-analyses and RCTs included studies that used smartphones as devices for the intervention delivery, and this is consistent with the fact that they are the most common devices used to access the internet games, apps, or other digital components [165]. Effectiveness was not likely to be affected by the specific device. The same could be stated for the use of devices such as console or arcade machines, computers, tablets, or iPod touch in relation to the increase in PA outcomes. The use of a miscellaneous array of devices, instead, revealed a major proportion of ineffective interventions.

Other frequently used devices were the wearables, in most cases overlapping with the wearable that was used as a component of the digital intervention delivered. The main used wearables were accelerometers and pedometers, such as ActiGraph, Polar Active, or Tractivity, all tools that have been widely validated and broadly used [166-168]. Additionally, Yamax or Omron pedometers were mainly used for the step count measurement. The same type of device was used within the intervention for the assessment of PA outcomes. Greater significant effectiveness was found for the use of pedometers, compared for example to the accelerometers. Pedometers may be more effective than accelerometers in increasing PA among children and adolescents because they provide simple, easily interpretable feedback, such as step counts, that is readily understood by young users. This immediacy can enhance motivation and goal setting, particularly when daily step targets are used. Pedometers are also less complex and often more user-friendly, which may promote consistent wear and engagement. In contrast, accelerometers are typically designed primarily for measurement rather than behavior change and often provide limited or delayed feedback to participants [67,77,169]. The visibility of step counts can encourage self-monitoring and reinforce active behaviors throughout the day. As a result, pedometers may better support behavior change processes in youth populations.

Some considerations have to be made about the specific outcomes that were assessed through the devices. Step count was positively affected by the digital intervention in half of the considered interventions, while MVPA or other PA outcomes, such as light-to-vigorous PA, LPA, MVPA, VPA, counts per minute, and active minutes, did not increase in the majority of the interventions. The failure in increasing step or other PA outcomes could be explained by supposing that engagement with the technology declined over time, reducing sustained behavior change. Some interventions relied primarily on monitoring without incorporating sufficient behavior change techniques such as goal setting, feedback, or rewards.

Limited integration of social support from parents, peers, or schools may also have weakened intervention effects [170]. In addition, interventions of short duration may have been insufficient to produce measurable increases in habitual PA, or those that were too long may have led to a loss of interest [171]. Technical issues, poor usability, or lack of age-appropriate design could have further limited children’s and adolescents’ motivation to increase daily steps and other PA outcomes [44]. Finally, measurement challenges and high baseline MVPA levels in some samples may have limited the capacity to observe significant increases [172].

The observed slight degree of overlap indicates that most SRs and meta-analyses include different primary studies. This statement supports more robust and balanced conclusions in this umbrella review, as the risk of double-counting evidence is reduced, and the independence of findings across reviews is strengthened. Although a slight overlap enhances the credibility of the overall synthesis by ensuring that conclusions are not driven by a small number of repeated trials, it may also increase heterogeneity, as findings are drawn from diverse interventions and populations [173]. Interpretation, therefore, requires careful consideration of contextual differences.

The critically low rating of the major part of the SRs and meta-analyses was frequently due to the omission of a key methodological requirement, specifically the reporting of excluded reviews and the reasons for their exclusion. The lack of transparency regarding excluded reviews could limit the completeness of the evidence base and the potential for selection bias.

With regard to the quality assessment of the RCTs, the domain of the RoB 2 tool “deviations from the intended interventions” was particularly critical. Digital interventions strongly depend on user engagement, adherence, and correct use of the technology, which are often inconsistent in younger populations. Children and adolescents may discontinue use, use the tools incorrectly, or engage at varying intensities. When participants deviate from the assigned intervention, the estimated intervention effect may be biased, especially if these deviations differ systematically between IG and CGs [174]. When not properly addressed analytically, they may lead to overestimation or underestimation of intervention effects or heterogeneity in reported outcomes.

A limitation of the present umbrella review is that several included SRs and meta-analyses were of critically low methodological quality; however, excluding all such reviews would have resulted in the omission of a substantial proportion of relevant evidence. This implies a potential and substantial influence on the interpretation of evidence of the digital tools’ effectiveness on increasing PA in children and adolescents. For example, methodological weaknesses, such as unclear inclusion criteria, incomplete search strategies, poor transparency in data extraction and synthesis, may reduce reproducibility and confidence in findings; inadequate assessment of risk of bias can result in overestimation or underestimation of intervention effects; insufficiently explored heterogeneity may mask important differences between intervention types or populations; inconsistent outcome definitions and measurement methods further limit comparability across reviews [175]. Therefore, the findings of this umbrella review should be interpreted with caution.

Similarly, RCTs assessed as having a high risk of bias constituted a substantial portion of the evidence base; therefore, excluding them would have been detrimental to the review. This could have influenced the results of this umbrella review, in the sense of overestimating the effectiveness of interventions, masking real differences between types of digital tools, such as mobile apps, wearable devices, or exergames. Common sources of bias, including inadequate randomization, lack of blinding, and high attrition, can systematically skew outcomes. Overall, the presence of high-risk trials weakens the ability to grade the certainty of evidence and may ultimately lead to inappropriate or poorly informed recommendations for practice and policy. Moreover, the lack of certainty of assessment (eg, through the GRADE [Grading of Recommendations, Assessment, Development, and Evaluation] tool) was a deliberate choice; it is a common and currently recognized practice in the field that a significant portion of published umbrella reviews do not assess the certainty of evidence, particularly when authors focus instead on the methodological quality of the included reviews using the AMSTAR 2 tool [176,177].

Another limit of this review is that it did not quantitatively synthesize through a meta-analytical approach (meta-meta-analyses or meta-analysis of the RCTs) the overall effectiveness, the heterogeneity, and the effects obtained after stratification and sensitivity analyses. Digital tools such as apps, wearables, exergames, and gamified platforms often differ in content, intensity, duration, and implementation context, making direct comparisons challenging. This variability limits the ability to synthesize findings quantitatively and reduces confidence in pooled effect estimates. In this overview, heterogeneity was explored through qualitative methods combined with narrative synthesis, and this approach helped in the identification of patterns and potential moderators even when quantitative pooling is not possible. Heterogeneity may obscure true intervention effects, as positive outcomes observed in some interventions may not be generalizable across different settings or age groups [178]. It also complicates the identification of which specific digital components or delivery strategies are most effective for schoolchildren. Consequently, conclusions must be framed more descriptively, and recommendations for practice and policy should emphasize context-specific effectiveness rather than universal applicability. Therefore, the authors envisage that the next step is to conduct a meta-analysis of RCTs to more robustly assess the effects of digital approaches on increases in PA. Such analyses should be complemented by sensitivity analyses, for example, excluding interventions of lower methodological quality, to enhance the robustness and interpretability of the findings.

As some of the included RCTs were conducted during the COVID-19 pandemic (2020‐2022) or in the postpandemic period, changes in PA patterns and digital tool use during these years may have influenced the observed outcomes [179], and this should be considered for the interpretation of results.

This umbrella review was based on rigorous and systematic methodology, relying on reviews that collected RCTs and that used objective device-based measures of PA. These measures improve the reliability of the review by providing more accurate and unbiased outcome assessment than self-reported data. Moreover, they reduce recall and social desirability bias, enhance comparability across studies, and allow more precise detection of changes in activity levels. As a result, they strengthen the quality of the evidence base, reduce heterogeneity related to outcome measurement, and increase confidence in conclusions regarding the effectiveness of digital interventions [180,181].

Risk of bias was evaluated at both the review and primary study levels, given that only a small subset of interventions from each review met the inclusion criteria for inclusion in this umbrella review. As a result, assessing risk of bias exclusively at the review level may have led to misleading conclusions, because the original reviews incorporated a broader evidence base.

Another strength of this review is its innovative character, as it is among the first umbrella reviews to comprehensively synthesize evidence from recent SRs and meta-analyses on digital PA interventions in children and adolescents. It provides novel comparative insights into methodological features and population subgroups that may drive or limit the success of digital approaches.

The findings of this work have several important real-world implications and contributions. They provide evidence-based guidance for practitioners, schools, and policymakers on which digital tools are more likely to effectively increase PA in children and adolescents, supporting more informed decision-making. The results highlight the importance of selecting appropriate devices, outcome measures, and intervention designs to maximize impact in real-world settings. This work also identifies gaps in equity and population coverage, informing future interventions aimed at reducing disparities in PA. Finally, the findings contribute to the optimization and scaling of digital health interventions by clarifying which approaches are most promising for practical implementation.

While the authors envisage quantitatively analyzing the effect of the interventions on the specific PA outcomes in children and adolescents in a future study through meta-analytic procedures, they wish to guide the strengthening of rigor for future SRs and interventions and support greater engagement in PA during childhood and adolescence.

First of all, the selection of device-based measures, which are able to provide reliable and comparable data, should be stressed. A clarification on the way to use gamification should be provided, together with a clearer explanation of the term “serious game” when it is used within the interventions. The importance of incorporating gamification tools, wearables, and/or apps as integral components of interventions should be highlighted, as they may be particularly effective for school-aged children [74,163].

Future interventions should place greater emphasis on conducting RCTs within home-based settings [150]. PA management should be the primary aim of the intervention; limiting the focus of the RCT to PA, without going too far into other outcomes that may cause a loss of the focus, could help in obtaining accurate data and effective outcomes [182]. There is a need to intensify RCTs targeted at high school populations, also considering that this age range is particularly engaged with digital technologies. Sustaining engagement over time remains a key challenge [183], highlighting the importance of developing and evaluating strategies that support long-term participation and adherence to digital interventions.

Better planning of interventions targeted to low-income and to insufficiently active subpopulations of children and adolescents represents a key aspect [184]. Future research should aim to promote and support RCTs in underrepresented regions, particularly in Asian and other non-Western high-income countries.

A greater emphasis on methodological transparency throughout the review and intervention selection process is suggested; in particular, clear and explicit criteria for study inclusion and exclusion should be reported. Transparency is needed in quality assessment procedures, including the rationale for quality scoring and study weighting; in addition, future reviews should clearly describe how methodological quality influences the interpretation of findings. The need for systematic evaluation and deliberate selection of behavior change techniques and theoretical models that are explicitly tailored to specific target populations within digital intervention contexts is underscored [185].

The results of this umbrella review can suggest practical actions both to schools and policymakers. Schools should avoid embedding PA as a minor component within broad well-being or educational programs; can integrate wearable-based programs (eg, step challenges and activity feedback) into physical education or daily routines; should favor pedometers or basic wearables over self-report tools to monitor student activity; and should accurately choose the type of device when implementing digital PA programs. Schools in disadvantaged areas may require additional support, such as teacher training, parental engagement strategies, or longer intervention duration.

Policymakers should prioritize funding and endorsement of interventions where PA improvement is explicitly defined, measured, and prioritized; can justify investments in wearable technologies as scalable and effective tools for youth PA promotion; can recommend objective PA monitoring standards in school-based programs to improve evaluation quality and accountability; can develop guidelines specifying preferred assessment tools to improve program effectiveness and comparability; should avoid one-size-fits-all approaches and allocate additional resources to adapt digital interventions to low-income contexts.

This innovative study represents one of the first umbrella reviews describing digital PA interventions in youth, determinants of effectiveness, and methodological gaps. It provides evidence-based guidance for practitioners, schools, and policymakers on which features of the interventions are more likely to effectively increase PA in children and adolescents, supporting more informed decision-making. Interventions appear to be most effective when they prioritize focused PA objectives, incorporate wearables as both the core digital intervention component and the delivery device, use pedometers to measure PA outcomes, and choose accurately the device to assess PA as a quality criterion in program selection.

All these adaptations in future studies could be of great importance to generating high-quality primary RCTs and provide consistent conclusions about the usefulness of digital approaches to increase PA. The findings of this umbrella review have important implications for public health practice and policy. Digital PA interventions represent a scalable, cost-effective approach to addressing low PA levels among children and adolescents, a key modifiable risk factor for noncommunicable diseases across the life course. By identifying which types of digital tools show the greatest potential and which design and contextual factors influence effectiveness, this review supports more strategic investment in evidence-based digital solutions within schools, communities, and health care systems. The results also underscore the need for public health stakeholders to move beyond one-size-fits-all approaches and to ensure that digital interventions are developmentally appropriate, engaging, and equitably accessible, particularly for underserved populations. Strengthening methodological rigor in future research will improve the reliability of evidence used to guide policy decisions and large-scale implementation.

Collectively, these insights can help inform integrated public health strategies that leverage digital innovation to promote lifelong PA habits, reduce health inequalities among youth, and the burden of chronic diseases in adulthood.