The proliferation of digital technologies has ushered new possibilities for connection, communication, and facilitation of life tasks such as work and studying but also concerns about potential issues associated with time spent in front of screens, including excessive gaming and social media use. Problematic use of digital media and technologies is a multifaceted issue encompassing different behaviors and different impacts for various age groups. Although professional organizations have recommendations regarding the use of various types of digital media [1,2], clinically diagnosable problematic digital media use has been characterized as a disorder only regarding video games [3,4]. However, research continues to address other problematic digital behaviors, including addiction to the internet broadly [5,6], social media [7-11], and smartphone use [11-13]. Although there are disagreements about what constitutes normative use and how it differs between media types, media affordances, and generations of users [14], most researchers agree that, for some users, technology- and digital media–related behaviors can cause significant life interference at times, leading to the need for treatment and intervention approaches to promote healthy digital media use.

A growing body of research has started to address this concern through interventions to prevent problematic use and promote digital well-being, which we define as the healthy use of digital media and technology and the absence of problems resulting from excessive use, consistent with other definitions that focus on balancing the benefits and drawbacks to various life domains, such as mental, physical, and social [15,16]. However, the diverse nature of problematic digital media use and the heterogeneity of these interventions make it challenging to evaluate their efficacy [5,17]. While much of the earlier research summarized in previous reviews focuses on excessive involvement in online video games, multiple forms of digital media are now used extensively in daily life, providing a new population norm that differs from that in earlier research on technology addictions.

Previous reviews synthesizing evidence on interventions to prevent problematic use and promote digital well-being face several challenges. Reviewers must contend with a field in which new measures of problematic use continue to proliferate even with 2 very specifically defined technology or media use disorders [17]. In addition, research often takes place outside of disciplines such as public health that have rigid standards for evidence synthesis, leading to incomplete or selective outcome reporting [18]. Context is also a challenge; public and scientific concern about the extent of problematic digital media use varies across global regions, leading to differences in intervention targets and approaches. For example, policy approaches differ between global regions [17,19], and studies about problematic digital media use often downplay contextual factors that may contribute to the feasibility of interventions in various settings [20,21]. Successful public health interventions hinge on a deep understanding of the social and environmental factors that contribute not only to health but also to intervention implementation [22]. As such, understanding context is vital for evaluating what might work to prevent problematic digital media use and promote digital well-being in global settings.

Using frameworks that account for constellations of potential risk factors can provide insights into evidence on prevention interventions [23,24]. To better contextualize potentially modifiable factors that affect the development of digital well-being or problematic use across the life span, we adapted the development of a digital well-being framework [25], which starts from a social ecological model and incorporates cognitive theories [26], media context [27], traditional approaches to the development of addictions (ie, incentive sensitization theory [28]), and uses and gratifications [29].

The aim of this rapid review was to describe the recent literature on interventions used in children, adolescents, and youth (aged <25 years) to prevent problematic digital media and technology use and promote digital well-being, with a focus on understanding the context of interventions and theoretical approaches from a developmental framework. The review methods, including the research question (RQ), search strategy, inclusion and exclusion criteria, and risk-of-bias assessment, were developed a priori and described in the registered protocol (PROSPERO CRD42023444387); these are also available in Multimedia Appendix 1 [8-13,30-58].

The registered PROSPERO protocol; description of the population, intervention, comparator, outcome, and study design (PICOS) framework; search strategies; and other supporting documentation are available in Multimedia Appendix 1 [8-13,30-58]. The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) checklist is available in Multimedia Appendix 2. Data extraction forms and databases, including a list of articles excluded at the full-text screening stage with reasons for exclusion and all extracted data, extraction of all data, and results of the tool named A Measurement Tool to Assess Systematic Reviews (AMSTAR) checklists, are available on our Open Science Framework (OSF) website [59]. As this study is a systematic review of the literature, it is not considered human participant research.

Our review included both systematic reviews (henceforth called reviews) and primary studies (henceforth called studies). As this was not human participant research, institutional ethics approval was not necessary. We used a pragmatic approach to systematically review and synthesize recent literature rapidly (approximately 9 months) with a focus on contextual factors that can aid in understanding translatability, making trade-offs in our methods appropriate for rapid reviews as per recommendations from the Cochrane Rapid Reviews Methods Group [60]. For example, we limited the number of outcomes to two (first, digital wellbeing or problematic use, and second, time spent on digital media); we restricted the search dates; and we used a single-extractor approach to data extraction, with 10% checked by a second team member. We adapted a conceptual model that combines theoretical frameworks focusing on individual development of addictions and media effects as well as models of behavior that combine individual-level and contextual factors [25-29,61].

We addressed the following RQs developed with stakeholder input:

We conducted searches in March 2023 combining terms as described in this section. Citations were downloaded into Zotero (Corporation for Digital Scholarship) [62] and then uploaded into Covidence (Veritas Health Innovation) [63] for screening and inclusion. To update previous literature reviews and allow for a rapid review focus on methodological rigor in data extraction, synthesis, and risk-of-bias assessment, we searched for empirical studies and systematic reviews published since 2017 in all languages in PsycINFO, Web of Science, and PubMed, as well as the gray literature databases of the World Health Organization (IRIS database) and ClinicalTrials.gov. We also hand searched the reference lists of review articles. We piloted combinations of various search terms that included subject and text keywords to retrieve papers related to the following: (1) children, adolescents, and young adults (aged ≤25 years) who did not meet the criteria for technology use disorders (participants); (2) interventions related to promoting healthy digital media use or preventing excessive or problematic use (interventions or exposures); (3) any comparator or no comparator (comparators); (4) digital well-being or the prevention of problematic or disordered use, including symptoms of problematic use and time spent on digital media (outcomes); and (5) primary studies and reviews of interventions (study type).

The final search terms can be found in Multimedia Appendix 1 [8-13,30-58]. We included intervention studies of various designs, protocols, and systematic reviews from all settings that (1) addressed primary or secondary prevention (ie, used a sample that was not formally diagnosed with digital media use disorders, such as internet gaming disorder, where the aim was prevention rather than treatment) and (2) contained outcomes related to time using digital media or technology or problematic or excessive use even if these were not primary outcomes. We excluded studies that (1) focused on clinical treatments or a sample with disordered levels of behavior or symptoms (eg, focused on “addicted” individuals), (2) focused primarily on other behavioral addictions, (3) lacked outcomes measuring digital media or technology use or addiction symptoms, or (4) had a mean sample age of >25 years. Titles and abstracts and then full texts were screened independently by 2 raters; discrepancies were resolved through discussion. In cases where study samples seemed to overlap, we contacted the study authors to ensure that the studies were distinct.

We developed and pilot-tested a data extraction form based on the standards for rapid reviews [64] and evaluation of complex interventions [65], AMSTAR 2 criteria [66], and the Cochrane Eyes and Vision work group [67]. Data extracted from primary studies included PICOS criteria; theoretical approach, study aims, and design; facilitators or barriers for the study context; possible sources of bias; harms reported; authors’ conclusions; and evaluation of quality and translatability using adapted forms of the Critical Appraisal Skills Programme (CASP) checklists [68,69]. One investigator extracted data from each study, and then a second investigator reviewed the data extraction for all studies, including reviews.

We classified the outcome as the broadest category when categories were combined; for example, “excessive media use” (“exzessive Mediennutzung” [8]) and “GD [gaming disorder] and unspecified IUD [internet use disorder]” [30] were designated as problematic internet use. Where social networking–related categories were combined with another category but were the subject of the intervention, we designated social networking as the focus, for example, “Internet addiction to the social networks” (emphasis added) [31].

Efficacy outcome data were extracted from tables, text, and figures in Microsoft Excel (Microsoft Corp) and Stata (StataCorp) to create counts for vote counting for controlled trials only. Where possible, use of digital media was converted to minutes per day (eg, from hours per day) to facilitate comparisons. For studies reporting univariate and multivariate outcomes, we extracted all outcomes.

Data extracted from reviews included citation information, type of review, aims and objectives, target population and setting, eligibility criteria for inclusion and exclusion, methods for specification of outcomes chosen for inclusion, measurement of outcomes, results, authors’ conclusions, and quality and risk-of-bias criteria. The AMSTAR 2 checklist was completed for each review and can be found on our OSF website along with the review data extraction database. Due to time constraints, missing data were not requested from the authors.

Consistent with Cochrane recommendations for a rapid review approach [60], we synthesized evidence through multiple tables and figures and provided a narrative synthesis of findings structured around intervention characteristics and approach, level of prevention, and population and setting. Due to the rapid nature of this review, we conducted vote counting to aggregate results from randomized controlled trials (RCTs) and other controlled studies that directly compared an intervention to no intervention. We compared the number of outcomes favoring the intervention versus control groups (ie, beneficial vs harmful effects), focusing on the direction of the effect rather than on statistical significance per revised Cochrane guidelines (a departure from our registered protocol) [70]. We then conducted a binomial proportion test using the bitesti command in Stata to determine the probability of observing the distribution of beneficial versus harmful effects by chance. We used the cii proportions command in Stata to estimate the SE and Jeffrey CI for this proportion.

The studies took place in 19 countries and included between 9 and 243,957 participants, with an average of 8282 (SD 43,744). One national study in South Korea accounted for 243,957 of the participants [32]; when that study was excluded, the maximum number of participants was 2997, and the average was 426 (SD 613). The systematic reviews included between 11 and 204 studies (mean 58, SD 73); 33% (2/6) of the systematic reviews conducted meta-analyses. The systematic reviews included a range of countries, and the total numbers of participants in the systematic reviews ranged from 5627 (calculated) to 162,494 (calculated). The studies focused on children aged 0 to 12 years (3/35, 9% studies), adolescents aged 13 to 18 years (8/35, 23%), and youth aged 19 to 25 years (5/35, 14%), whereas an additional 31% (11/35) of the studies combined more than one age group, 17% (6/35) focused on parents or children and parents, and 6% (2/35) focused on other groups [8,13].

Table 1 provides an abbreviated overview of the characteristics of the studies of individual interventions included in our review; a full summary-of-findings table can be found in Multimedia Appendix 1 [8-13,30-58]. Overall, we found 33 different interventions in 35 studies. A total of 54% (19/35) of the studies were controlled trials that compared interventions to a control group (waitlist or no intervention), 20% (7/35) were before-and-after studies, 23% (8/35) were observational studies or protocols, and 3% (1/35) were unclear [31]. Of the controlled trials, some (3/19, 16%) did not compare extracted outcomes directly between intervention and control groups using statistical tests (eg, compared only within-group changes) and were not included in the vote counting. Of the 27 studies (77%) using causal language in their aims (eg, aiming to “analyze or examine effectiveness or investigate or examine [the] effects” of interventions), several (12/27, 44%) did not use a before-and-after design that compared an intervention group and a control group directly. Most reviews (4/6, 67%) required prospective study designs, but 33% (2/6) included protocols [5,71]; no review included only RCTs.

We conducted a rapid systematic review of the recent literature (since 2017) to investigate the current evidence on prevention approaches to promote digital well-being, defined as the healthy use of digital media and technology and the absence of problems resulting from excessive use, among children, adolescents, and youth. The final selection included 41 studies (n=35, 85% primary studies and n=6, 15% reviews). In the primary studies, internet and game use were the most common outcomes, and most interventions (22/33, 67%) included education about self-regulation. Digital well-being education, media literacy and effects, or mental health symptoms and coping (psychoeducation) were other common intervention approaches. Most interventions (26/33, 79%) also combined several approaches. Vote counting of controlled trials showed benefits for all interventions. A total of 33% (2/6) of the reviews conducted statistical analyses and found that the interventions studied had a small but beneficial impact on decreasing the amount of time that people spent on screens.

Similar to previous reviews in this area [5,17], this review also found that most programs for universal and selective intervention to promote digital well-being and prevent problematic technology use were implemented in educational settings. Many of these interventions can be classified as complex interventions that contain multiple active ingredients and tap into different potential behavior change strategies. Intervention approaches that aimed at improving the self-regulation of participants were the most common in the included studies, and this self-regulation closely resembled goals, feedback, and planning in a widely used taxonomy of behavior change strategies [76]. However, a close investigation of the individual studies showed that self-regulation, even if mentioned using that term, could look very different. For example, the 6-week intervention by Apisitwasana et al [33] contained 5 weeks of instruction specific to self-regulation, whereas the study by Affouneh et al [43] delivered 1 week of “self-discipline” instruction for avoidance of addictions specifically, with other weeks targeting “problem solving strategies for dealing with problems to prevent addictive behaviors” and “effective use of time to prevent addictive behaviors.” While grouping of intervention approaches is important for systematic investigation, these limitations in comparability of approaches across studies should be kept in mind, and detailed reporting of intervention approaches in future studies is needed.

Moreover, the interventions included in this review varied widely in intensity, with some interventions requiring a single session and others requiring 1:1 weekly clinician visits over weeks. Peer-delivered interventions may reduce overall costs. These have been successful for other addictions and are a common method of task sharing for psychosocial interventions, especially in middle-income settings [77]. However, this depends on the intensity of training; if extensive training of teachers is needed for them to be able to train peers and teenage peers then age out, these extensive trainings may need to be repeated with high frequency. Although our review was not designed to provide a quantitative evaluation of the studies, the included review by Jones et al [73] found that interventions with shorter durations were more likely to be effective in reducing screen time. Interventions with shorter durations are more likely to be feasibly implemented, and future studies should aim to examine associations among intervention intensity, feasibility, and efficacy.

Harms are not usually assessed in the reporting of studies of interventions for behavioral addictions [74], but it is important for digital well-being. Interventions involving abstinence from addictive behaviors may result in rebound effects after the intervention or may lead to engagement in compensatory behaviors that could be more harmful (eg, searching for gaming-related pornography when restricted from gaming [78]). We found that some comparisons included in our vote counting (7/51, 14%) had results that favored the control group; however, these were usually due to baseline differences (6/7, 86%). In contrast, individuals struggling with more harmful addictive behaviors such as alcohol or drug use may use digital behaviors—even excessive ones—as a less harmful compensatory behavior to deal with cravings and withdrawal symptoms [79]. Findings on both the potential benefits and negative consequences of digital well-being interventions underscore the need for RCTs with appropriate treatment of any baseline differences as well as comprehensive and consistent reporting of outcomes in future research.

As suggested by other scholars [17], our review included an explicit aim to report aspects of published work that impact how the findings can be translated and applied to different contexts (eg, countries and areas of the world that are starting to see technology use problems but have not developed or systematically evaluated digital well-being interventions themselves). Many studies included in this review (20/35, 57%) were conducted in Europe and the Global North. Whether these studies can translate to other contexts is another question. Our review focused on identifying factors that can help interventionists decide whether an intervention will work in their setting. Things to consider are information on intervention costs, intervention manuals, and information on scalability. Only a minority of studies (7/35, 20%) reported some of these aspects. Future studies would benefit from including greater detail on factors that would help inform translation to other contexts.

Moreover, formulating a well-defined RQ; using a clear study design; and providing comprehensive information about the studied population, intervention, comparator, and measured outcomes are crucial for understanding study quality. Reporting guidelines for empirical studies and systematic reviews provide clear recommendations for elements of study design, methods, and outcomes to report and should be used in future work. We hope that, by explicitly focusing on these topics, our review can inspire future studies to report this crucial information and, thus, make a contribution to developing the international evidence base in this important and growing area of research.

This review has some limitations. As this was designed to be a rapid review, we deferred some elements that are common in other types of reviews (eg, double data extraction, contacting authors for missing data, performing meta-analyses or otherwise evaluating effect sizes, and conducting a formal risk-of-bias assessment [64]). Moreover, we included the most recent literature published since 2017, which is appropriate for a rapid review and quickly changing technology platforms and use patterns. We excluded studies that described “treatment” of “individuals with disorder” even if the screening instrument scores of the included samples fell below the cutoffs used for the disorders. Thus, we may have missed other interventions that could be helpful to improve digital well-being. Although we searched 5 databases, we missed interventions such as the Digital Balance survey and website and Ithra Sync, which were delivered through websites and social media and did not show up in our searches [16,80]. Outcome-level limitations include that we did not limit our review to studies with specific outcome measures or scales or even to studies with validated outcome measures. This allowed us to include a wide range of studies but runs the risk of including weak outcome measures (eg, inconsistent measures for the same outcome across the studies and problems with self-reporting screen time). At the study level, we found that poor reporting of key information, including fidelity of intervention implementation or missing data, presented significant limitations to our ability to synthesize data into evidence. In some cases, study design and statistical analysis were difficult to ascertain [44] or missing [31]. Only a subset of RCTs (14/19, 74%; Table 3) were included in vote counting because they reported direct comparison between the intervention and control groups on the outcomes selected for this review. In addition, the dates of our review, 2017 to 2023, included the 3 years of the COVID-19 pandemic (ie, March 2020 to May 2023), making it possible that empirical studies occurring during this time could contain cohort effects. Finally, most reviews (4/6, 66%) lacked protocols or outcome specification, leading to low quality scores.

The strengths of our review include that all methods were guided by standards, including the Cochrane rapid review recommendations and AMSTAR 2 checklist. To enhance the ability of interventionists to use these findings, we focused on critical appraisal of quality and translatability through CASP checklists. We preregistered our protocol, and all accompanying data are supplied on an OSF website, which promotes reproducibility.

As digital technology use is increasing, interventions to prevent problematic use and promote digital well-being will be of growing importance. This review demonstrates positive findings for effectiveness of prevention interventions for behavioral outcomes and describes factors that may contribute to translation and implementation. Digital well-being is a global need, and understanding contextual factors that influence healthy use as well as limitations of the current evidence is vital for informing future research. Future research would benefit from following appropriate reporting guidelines, reporting both the benefits and harms of interventions, and including greater detail on factors informing translation.