This umbrella review was conducted in accordance with the registered protocol (Open Science Framework Registry) and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement [25]. Our study protocol has been registered with PROSPERO.

We searched the following databases for English language studies published up to February 2024: PubMed, the Cochrane Library, Web of Science, and Embase. We used key terms related to digital addiction and interventions. The following search strategy was used: (“Internet” OR “digital” OR “screen” OR “net” OR “online” OR “media” OR “electronic device” OR “electronic gadgets” OR “computer” OR “mobile” OR “phone” OR “smartphone” OR “television” OR “TV” OR “video” OR “Facebook” OR “game” OR “gaming”) AND (“addict” OR “use” OR “dependent” OR “overuse” OR “abuse” OR “disorder” OR “excessive” OR “effects” OR “habits” OR “intervene” OR “treatment” OR “therapy” OR “compulsive” OR “heavy”) AND (“meta-analy” OR “meta-analytic” OR “metaanaly” OR “meta-regression” OR “meta synthesis” OR “meta-analysis” OR “meta-analysis”) AND (“intervention” OR “interfere” OR “intervene” OR “treatment” OR “therapy” OR “cure” OR “therapeutic” OR “exercise therapy” OR “counseling”). Studies were published in full peer-reviewed literature without language or date restrictions.

The studies included in the meta-analysis had to meet the following inclusion criteria: (1) research topics related to digital addiction intervention; (2) participants diagnosed with digital addiction; (3) a study design limited to RCTs, nonrandomized controlled trials, and quasi-experimental studies; and (4) different forms of digital addiction with similar effects.

The exclusion criteria were as follows: (1) duplicate publications; (2) full-text documents could not be obtained, and other experimental-related basic research was performed; and (3) meta-analysis without any effect size. When 2 or more meta-analyses were performed for the same association, only the most recent systematic review with the largest number of individual studies was included to avoid duplication of samples (Multimedia Appendix 1 [18,20,21,26-39]).

Documentation records were extracted and imported into Covidence (Veritas Health Innovation). After excluding duplicates, titles and abstracts were assessed independently by 2 researchers. The full texts of the relevant studies were independently reviewed by 2 researchers to ensure that the inclusion and exclusion criteria were met. In the event of discrepancies, a final assessment was made by a third independent researcher.

Data extraction was independently completed by 2 researchers. The information extracted from each meta-analysis was as follows: first author; publication year; design and number of included studies; statistical estimates of the meta-analysis; number of original studies; sample size and region; intervention; comparison; any relevant subgroup analyses; and intervention-related outcome indicators. Any disputes during data entry were resolved through discussion with a third researcher. When a meta-analysis contained multiple intervention outcome indicators, each indicator was extracted and analyzed separately.

Two reviewers independently assessed the methodological quality of each systematic review via the Assessment of Multiple Systematic Reviews version 2 (AMSTAR 2) rating scale [40]. AMSTAR 2 consists of 16 items and is a measurement tool used to evaluate meta-analyses. The scale has good consistency, reliability, construct validity, and feasibility. It covers key questions of the included studies, proposals, literature searches, literature screening, data extraction, basic characteristics of the included original studies, data analysis, and conflicts of interest. The overall quality of a systematic review is informed by assessing how well the study meets each of the criteria. Overall quality is classified as “high,” “medium,” “low,” or “very low” [40].

The umbrella review compiles the evidence of multiple clinical-related issues in the system reviews and meta-analyses and then comprehensively evaluates the strength of the meta-analyses related to specific issues. First, the combined effect values and their 95% CIs associated with digital addiction interventions in each meta-analysis and their 95% CIs were extracted, and the combined effect sizes, 95% CIs, and P values were reestimated after the normalization of the mean difference (MD) and Hedges g to the standardized mean difference (SMD) via a random effects model (DerSimonian and Laird) by referring to the analyses of the previous omnibus review. I² and Cochran Q were used to evaluate the heterogeneity of the evidence in the literature, with I²≥50% or Cochran Q test P value of less than .10 indicating significant heterogeneity in the evidence. The estimated values of the publication bias of various studies through Egger inspection and evaluation, P value of less than .10, indicate that there is a significant publication bias. When the results indicate that there is a significant bias, if the total estimated effect value of the meta-analysis is greater than the effect value of the original research with the largest proportion of its effect, it indicates that the meta-analysis has a small-scale research effect. By determining whether the number of statistical results (P value of less than .05) in each meta-analysis (P value of less than .05) is greater than its expected quantity (E), the meta-analysis is excessive E, and P value of less than .10 indicates excessive significance. The rest of the analyses used a difference of P value of less than .05 to determine whether the difference was statistically significant.

The results are presented in forest plots and tables, and all analyses and mapping were performed via Stata (version 17.7, StataCorp) and R (version 4.2.1; R Foundation for Statistical Computing).

The Grading of Recommendations, Assessments, Development, and Evaluations (GRADE) approach was used to evaluate the strength of the evidence in each included study. According to the statistical analysis results, the credibility of each piece of evidence is further divided into 5 levels: I, II, III, IV, and NS, which represent “convincing evidence,” “highly suggestive evidence,” “suggestive evidence,” “weak evidence,” and “insignificant evidence,” respectively [41].

We retrieved 284 meta-analyses from PubMed, 277 from Embase, 319 from Web of Science, and 75 from Cochrane. Two investigators independently screened the studies, and 5 meta-analyses were ultimately included in this umbrella review through 2 rounds of literature screening (Figure 1).

The number of studies included in the 5 meta-analyses ranged from 11 to 59, with a median of 7 (IQR 6-11). The sample size ranged from 580 to 4656, with a median of 407 (IQR 107-1490). The original research covered multiple regions, including Asia, Europe, Oceania, and the Americas. AMSTAR 2 conducted a methodological quality assessment of the 5 meta-analyses included in this umbrella review. Among them, 4 studies were rated as “high,” and 1 study was rated as “low” (Table 1).

Five meta-analyses evaluated 9 interventions: CBT, group counseling, exercise, integrated IA prevention programs, psychosocial interventions, reality therapy, self-control training programs, and interventions to reduce screen time in child and art therapy. The general characteristics of the studies included in the meta-analysis are presented in Table 1. Four of the selected meta-analyses were graded as “high quality” on the basis of the AMSTAR 2 score, and only 1 review was graded as “low quality.” However, the evidence class of most meta-analyses included evidence rated as “weak evidence” (19/21, 90%) and “not significant” (2/21, 10%). A summary of the quality appraisals of the meta-analyses and the AMSTAR 2 scores are presented in Table 1 and Multimedia Appendix 2 [18,21,22,26,42].

CBT is effective in treating substance abuse, gambling, and affective and eating disorders. CBTs are based on the cognitive behavioral model, which holds that thoughts determine feelings; thus, changing one’s thoughts can help with behavioral change [43]. CBT can reduce anxiety, depression, IGD symptoms, time spent gaming, and IA scores, as shown in this review. Two meta-analyses previously examined the effectiveness of CBT for patients with digital addiction [22,44]. For example, Stevens et al [22] report that CBT demonstrated high efficacy in reducing IGD symptoms (SMD 1.394, 95% CI 0.664-2.214) and depression (SMD 0.797, 95% CI 0.254-1.341) and moderate efficacy in reducing anxiety (SMD 0.939, 95% CI 0.311-1.586) and time spent on gaming (SMD 1.259, 95% CI 0.311-2.206). However, owing to the evidence base, there is a need for more rigorous studies to determine the potential long-term benefits of CBT for IGD [22]. The heterogeneity in the effects of CBT on reducing anxiety (I²=92.5), depression (I²=87.1), IGD symptoms (I²=96.6), and time spent gaming (I²=95.9) was high, although the effects of CBT on depression exhibited publication bias, albeit the evidence was weak. Another review explored the impact of CBT on IA levels [44] and found CBT to be significantly superior to routine intervention and paroxetine combined with buspirone drug therapy (P value of less than .05) on the IA test scale and also superior to routine intervention and no intervention (P value of less than .05) on the basis of the Revised Chen Internet Addiction scale and Young Diagnostic Questionnaire scale (P value of less than .05) [40]. Finally, the heterogeneity for the effects of CBT on reducing IA scores was high (I²=92.5; Figure 2 and Table 2).

The main components of group counseling include identifying previous solutions and acknowledging problems while also realizing that exceptions to the problem are key to the solution, with a focus on the present and future [45]. Two meta-analyses previously explored the effectiveness of group counseling on digital addiction. Yeun et al [18] reported that group counseling had a large effect on improving self-control among school-aged children with IA (SMD 1.296, 95% CI 0.269-2.322). The heterogeneity for the effects of group counseling on improving self-control was high (I²=66.9), but the evidence was weak. Another review reported that group counseling reduced IA levels (SMD –1.417, 95% CI –1.836 to –0.997) [44]. The heterogeneity for the effects of group counseling on reducing IA scores was high (I²=95.9; Figure 2 and Table 2).

Exercise is an important intervention method that positively impacts individual cognition, emotion, and physical fitness [5]. Two meta-analyses examined this intervention’s effectiveness in reducing digital addiction outcomes. One review reported that sports interventions reduced IA scores (SMD –2.322, 95% CI –3.212 to –1.431) [44]. The heterogeneity for the effects of exercise on reducing IA scores was high (I²=88.1), although the evidence was weak. Another review conducted by Zhang et al [21] included a total of 2408 students from 39 RCTs. Six exercise types (team sport, double sport, single sport, team + double sport, team + single sport, and team + double + single sport) were compared on the basis of their effectiveness in reducing IA and maintaining mental health. The meta-analysis results revealed that, compared with the control group, exercising significantly improved loneliness, anxiety, depression, and interpersonal sensitivity (P value of less than .05) [21]. The heterogeneity in the effects of exercise on improving loneliness (I²=93.1), anxiety (I²=87), depression (I²=70.7), and interpersonal sensitivity (I²=79.9) was high (I²=95.9; Figure 2 and Table 2).

Psychosocial interventions capitalize on psychological or social actions to produce changes in psychological, social, biological, and functional outcomes [46]. One review examined the efficacy of psychosocial interventions on school-aged children’s IA, self-control, and self-esteem and revealed that psychosocial interventions had a strong effect on reducing IA (SMD –1.111, 95% CI –1.552 to –0.67) and improving self-esteem (SMD 1.609, 95% CI 0.439-2.778) [18]. The effect on IA had high heterogeneity (I²=82.5) and published bias (P=.004), and the heterogeneity of self-esteem was high (I²=87.7; Figure 2 and Table 2).

The integrated IA prevention (IIAP) program combines 2 or more interventions, including CBT, play, art therapy, conflict, stress management, and CBT. [18] A meta-analysis of data on the efficacy of IIAP for self-control and IA revealed that IIAP had no effect on self-control (SMD –0.043, 95% CI –0.506 to 0.421), although the evidence was insignificant [18]. IIAP was found to reduce the IA, and while the evidence was weak, the effect showed high heterogeneity (I²=88.1), publication bias (P=.027), and excessive significance (observed number of studies with positive findings=3; expected number of studies with positive findings=0.23; P value of less than .01; Figure 2 and Table 2).

A meta-analysis of data on the efficacy of self-control training programs for self-control and IA reported in 5 studies revealed that self-control training programs improved self-control (SMD 1.873, 95% CI 0.158 to 3.588) and reduced IA (SMD –2.62, 95% CI –4.071 to –1.169) [18], but the evidence was weak. In addition, the effects were highly heterogeneous and highly significant (P value of less than .1). There was a publication bias in terms of the effect on IA (P=.02; Figure 2 and Table 2).

Weak evidence has indicated that in a review of the effect of reality therapy on IA based on two trials, reality therapy reduced IA (SMD –1.096, 95% CI –1.677 to 0.514) [18] without heterogeneity between the 2 trials (Figure 2 and Table 2).

A meta-analysis suggested that interventions for screen use and sleep can achieve a small reduction in screen time (SMD –2.125, 95% CI –3.233 to –1.017) [47]. Heterogeneity was high (I²=98.7), and the evidence was weak (Figure 2 and Table 2).

This umbrella review provides an overview of the relationships between digital addiction intervention-related outcome indicators and intervention methods derived from 5 meta-analyses. Four of the selected meta-analyses were of “high” quality. We found that most interventions (eg, CBT and exercise) can improve related outcomes, such as lifestyle, emotional health, mental health, and social relationships, among people with digital addiction. However, this conclusion was derived from weak evidence and has significant heterogeneity and excessive significance. The evidence in this review was rated as “weak evidence” (19/21, 90%) or “not significant” (2/21, 10%). The evidence was downgraded to “weak” considering that most of the included studies were RCTs, which generally involved a limited number of participants, that is, fewer than 1000 in total after pooled analyses, easily causing the small sample sizes limitation, and second, the articles were considered to have exhibited heterogeneity as well as publication bias, suggesting that more high-quality, credible research is needed to further clarify the associations between intervention methods and digital addiction outcomes.

On the basis of the results of this umbrella review, CBT not only reduces IA scores but also reduces IGD symptoms, depression, and anxiety. CBT, the psychotherapy most commonly used for digital addiction, can improve symptoms of depression and anxiety by reshaping negative thoughts to help patients understand routine and problematic behaviors and motivate the creation of more adaptive, goal-oriented routines [48]. Compared with other interventions (such as drug replacement therapy), CBT has the advantage of being less invasive and effective in the short term. Furthermore, CBT is the most effective intervention for reducing internet use and improving self-perception [49]. CBT emphasizes the connection among behaviors, thoughts, and emotions, prompting patients to pay attention to these behaviors [20]. A previous meta-analysis revealed that CBT was associated with positive changes in depression, anxiety, and psychoticism in patients with IA [20]. Ding KY et al [48] also reported that most interventions for digital addiction in children and adolescents were CBT- or CBT-based interventions, which can improve anxiety, depression, and digital addiction-related symptoms. However, another meta-analysis did not conclude that the CBT intervention had a significant effect on the severity of IA [50]. Our analysis of the effect of CBT on digital addiction suggested that genuine heterogeneity likely exists. The high heterogeneity might be related not only to the potential bias in the original studies but also to the differences across the studies included in this meta-analysis. These associations are both supported by a weak level of evidence; thus, more studies are needed to further document the effect of CBT on patients with digital addiction.

There is evidence that group counseling is one of the main treatments for addiction. One of the advantages of group programs is their economic value since they target many students simultaneously [51]. A previous study reported that groups of fewer than 14 people presented effect sizes exceeding 90% [52]. Group counseling programs enable group members to understand and help each other, increase social participation and interpersonal interactions among patients, improve adaptability, develop personalities, and eliminate the symptoms of digital addiction [53]. Group counseling programs can gain patient support and insight through experiencing similar cognitions and emotions, providing a major modality in the treatment of digital addiction [54]. In this review, we conclude that group counseling improved self-control and reduced IA scores among patients with digital addiction. A meta-analysis suggested that group counseling programs improve relationships, health problems, time management, tolerance, and compulsive internet use [20]. A previous meta-analysis by Park et al [55] identified the beneficial effects of group counseling programs for adolescents with IA. However, group counseling may not be truly beneficial until the barriers related to social anxiety, social isolation, and a lack of social skills are overcome. Therefore, developing a protocol for appropriate methods on the basis of the characteristics of the participants is necessary.

Exercise, including running, badminton, basketball, table tennis, and other outdoor sports games, was found to be helpful in reducing IA scores and improving loneliness, anxiety, depression, and interpersonal sensitivity in patients with digital addiction. Exercise is an important intervention method that positively impacts individual cognition, emotion, and physical fitness, in addition to psychological interventions. Digital addiction alters the neural structure, reduces the activity of the dopaminergic system, and limits neurocognitive functions [56]. However, exercise-based interventions can effectively improve the autonomic nervous system in patients with IA and normalize, to a certain extent, the structure of specific parts of the central nervous system [57]. Moreover, it can increase plasma glial cell-derived neurotrophic factor and glucocorticoid levels [58], neurotransmitter release [57], and telomere protection [59]. Exercise-based interventions have been shown to reduce the prevalence and symptoms of IA, and participation in physical activity may prevent IA. Exercise has replaced most internet experiences and improved an individual’s physical and mental health [19]. Long-term physical exercise has been found to significantly reduce the degree of IA and depression and improve sleep quality among college students with IA [44]. The meta-analysis included in this review of exercise interventions included only exercise type classification, that is, it did not compare exercise intensity or exercise strength. Thus, more studies on exercise interventions are needed to demonstrate their effects on digital addiction.

Digital addiction may constitute impulsive behavior, although the consequences are negative impacts and involve continuous participation [60]. The impulse theory suggests that the behavior of addiction is the result of the generally excessive active brain reward system. Individuals with excessively active brain reward systems may show strong reactions to the prompts of predicting potential rewards, thereby explaining the pursuit of novelty, impulses, and powerful motivation to obtain internet or other addictive stimuli [61]. Therefore, impulses have been proposed as potential marks and treatment targets of digital addiction [62]. Reality therapy helps individuals control their behavior and consider other options related to internet use [63]. As reality therapy is based on the theory of selecting, the theory holds that people are responsible for their own lives, their feelings, their actions, and their behaviors [64]. In a sense, reality therapy helps individuals reflect on their behavior, evaluate their choices, choose more effective choices, and directly target the choice and self-control of target guidance [64]. In this method, the participants are required to determine the goals of their behavior and realize what they are actually doing. They were then guided to determine whether their behavior promoted or hindered their ability to achieve their original goals. Finally, they are encouraged to seek more appropriate and healthier alternatives to replace their current behavior, achieve their goals, and formulate plans to change negative behaviors [63]. Real-world therapy has been widely used to treat addictive diseases, including diseases involving drugs, sex, and food [63]. For people with addiction and impulses, reality therapy can help them improve their self-control and thereby reduce their problems. Research has shown that reality therapy can increase self-esteem and decrease IA [64]. Similar impacts have also been proven in the study of material addiction in real-world therapy [65]. In this review, we found that real-world therapy reduces IA scores. Despite the study being graded as having high methodological quality, the level of research evidence was weak due to the small number of participants included in the study.

IIAPs are characterized by the simultaneous or sequential delivery of several different types of interventions, including pharmacology, psychotherapy, and group counseling. A composite approach may help solve complex problems, whereas combinations of different therapies with different therapeutic targets and mechanisms may produce synergistic effects. The effects of CBT in combination with other psychosocial or pharmacological therapies for IA have been studied and determined that the combination of CBT with other therapies affects IA severity, somatization, paranoid ideation, psychosis, obsessive-compulsive traits, interpersonal sensitivity, depression, and anxiety [66]. In this umbrella review, the assessments of IIAPs had high between-study heterogeneity, with public bias and excessive salience, and the evidence of the effect on IA scores was weak. The IIAP showed a null evidence class of effects on self-control, suggesting that there is no clear scientific evidence to support the link between IIAP and self-control in patients with digital addiction.

The evidence reviewed here suggests that CBT, group counseling, exercise, IIAP programs, psychosocial interventions, reality therapy, self-control training programs, and interventions to reduce screen time may be beneficial to patients with digital addiction. Nonetheless, more robust population-based studies are essential to substantiate these potential benefits, as the current evidence is largely inconclusive or weak. It is imperative to conduct further research to evaluate the long-term efficacy and cost-effectiveness of these interventions, as well as to investigate their suitability and generalizability across diverse cultural and socioeconomic settings.

This umbrella review has certain limitations. First, this review was based on the results of published and available meta-analyses and meta-analyses, so other results that have not been evaluated by meta-analysis and missing data in the original literature could not be considered. However, our results were not significantly affected because repeated meta-analytic evaluations yielded similar results. Second, throughout the study, original studies covering multiple regions were included in this analysis, and although we were not able to describe subgroups of regions based on the meta-analysis that was conducted, a cross-cultural analysis of multiple geographic sources provides a better description of the effects of various interventions on digital addiction. Third, studies suggest that gender plays a role in digital addiction, as evidenced by the greater susceptibility of males, but the meta-analysis included did not stratify gender to examine the differences in the efficacy of interventions for digital addiction and therefore cannot be discussed further in this study [67]. Fourth, the statistical methods we used to test for the presence of bias indicate only its existence, not its exact source. Finally, the methodological quality of the meta-analyses included in this umbrella review and the quality of evidence for most prognosis-related outcome indicators were not sufficient. Therefore, more high-quality studies are needed in the future to further verify the results of this study.