This umbrella review was conducted in accordance with the registered protocol (Open Science Framework Registry) and the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 statement [13]. Our study protocol has been registered with PROSPERO (registration number CRD42024594911).

This study followed the umbrella review process established by Aromataris et al [12], searched databases such as PubMed, Web of Science, the Cochrane Library, and Embase, and selected systematic reviews and meta-analyses on the influencing factors of digital addiction included in the databases up to September 24, 2024. The keywords used included digital addiction, digital overuse, internet addiction, internet-based addiction, problematic internet use, gaming disorder, social media addiction, influencing factors, risk factors, causative factors, contributing factors, meta-analysis, and systematic review, etc, and the search was conducted using a combination of subject words and free words. Taking the PubMed database as an example, the search strategy was shown in Multimedia Appendix 1. We also hand-searched the references of eligible included studies and google scholar to identify further relevant articles.

We included studies on the influencing factors of digital addiction if they met the following inclusion criteria: (1) study type was systematic reviews and meta-analyses; (2) studies that focused on the various determinants of any type of digital addiction across different populations; (3) studies that reported the overall effect size, including but not limited to pooled Pearson correlation coefficients or Spearman rank correlation coefficients (r), standardized mean difference, and odds ratio (OR) along with their 95% CIs or credible intervals.

Meta-analyses focusing exclusively on a single influencing factor were selected, prioritizing the most recent study with the largest number of included studies when duplicates were present [14]. When multiple systematic reviews addressed the same influencing factor, only the most recent review with the largest number of individual studies providing study-level estimates was included to prevent sample duplication. No language or date restrictions were applied. Articles lacking study-level effect sizes and 95% CIs for systematic reviews, or those based on animal models, were excluded. In addition, we excluded conference abstracts, editorials, narrative reviews, and systematic review protocols.

This study was conducted by 2 researchers who screened the literature and extracted data according to the inclusion and exclusion criteria. In case of disagreement, consulting a third researcher (ZFF) assisted in making the decision. Duplicate literatures were eliminated through the title and abstract, and the full texts were carefully read to extract information such as the first author, publication time, number of studies included and sample size, exposure factors, main outcome indicators, meta-analysis effect model, effect size and its 95% CI. When the same meta-analysis contained multiple health outcome indicators, each outcome indicator was extracted separately.

In this study, the modified version of the systematic review methodology quality assessment tool (AMSTAR 2 [A Measurement Tool to Assess Systematic Reviews 2]) was used to assess the methodological quality of systematic studies [15]. AMSTAR 2 evaluated the methodological quality and credibility of the study from 16 items in 3 dimensions (including 7 key items and 9 non–key items) such as research design and registration, literature search and selection, quality assessment of the study, and data extraction. It demonstrated good inter-rater consistency, practicality, and a wide range of applications. Referencing the interpretation by Zeng et al [16], it was considered that ≤1 non–key item not met was high-quality research, ≥2 non–key items not met was medium-quality research, 1 key item not met was low-quality research, and ≥2 key items not met was extremely low-quality research.

The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) tool was used to evaluate the evidence quality level of the literature [17]. The GRADE grading system evaluated the outcome indicators based on 5 downgrade factors such as limitations, heterogeneity, indirectness, imprecision, and publication bias. This study stipulated that no downgrade factors were high-quality evidence, the existence of 1 downgrade factor was medium-quality evidence, the existence of 2 downgrade factors was low-quality evidence, and downgrade factors ≥3 were extremely low-quality evidence [18,19]. While the GRADE system typically assigns an initial “very low” quality rating to evidence derived from observational studies due to inherent limitations such as potential confounding and bias, it allows for upgrading or downgrading based on additional factors such as the magnitude of effect, dose-response relationships, and the presence of plausible confounding factors. In this umbrella review, we applied the GRADE framework to evaluate the quality of evidence across meta-analyses, which often include a mix of observational and other study designs. This approach enables a more nuanced assessment of the evidence, considering factors such as risk of bias, inconsistency, indirectness, imprecision, and publication bias. By doing so, we aimed to provide a comprehensive evaluation of the credibility and robustness of the findings related to digital addiction.

This study conducted a methodological assessment using AMSTAR 2 and evaluated the quality of evidence with GRADE, aiming to identify high-quality and significant results from the most up-to-date and comprehensive available evidence. Significant heterogeneity was defined as present when the I²>50% or the P<.1 [20]. A meta-analysis was considered to have a small sample size if there were fewer than 1000 participants in the meta-analysis [21,22]. Notably, a significant publication bias was inferred if the largest weighted effect size in the meta-analysis was subordinate to the overall effect size, indicating a significant small study effect (SSE) [21,23]. Publication bias was quantitatively assessed using the Egger test [24] and the Begg test [25], with a P value <.05 [26] indicating the presence of statistically significant bias.

Through a database search and additional manual search, a total of 912 relevant articles were initially identified. After removing duplicates, 536 articles remained. Following a review of titles and abstracts, 387 articles were excluded (266 due to mismatched study types and 151 for being irrelevant). After full-text screening and reviewing in Multimedia Appendix 1, 101 articles were further excluded due to incomplete data or lack of effect size. As illustrated in Figure 1, the final selection process resulted in the inclusion of 18 articles that met the inclusion criteria.

The included meta-analyses were published between 2020 and 2024, reporting on 24 potential influencing factors related to digital addiction, with personality traits further subdivided into 8 categories. The sample sizes of the studies ranged from 300 to 159,425 participants. The included literature mainly used the New Castle-Ottawa scale (NOS) and the Joanna Briggs Institute (JBI) systematic review checklist to assess research quality. A total 6 studies did not specify the methods used for quality evaluation or risk of bias assessment. The basic characteristics of the included literatures are seen in Table 1.

Through evaluation with the AMSTAR-2 tool, all included meta-analyses exhibited some methodological deficiencies. Specifically, 6 were of moderate quality, 4 of low quality, and 8 of very low quality, with none classified as high quality. For key items, the compliance rates were as follows: item #2 (whether the review reported that its methods were established a priori and explained any major changes to the protocol), item #4 (whether a comprehensive literature search was conducted), item #7 (whether a list of excluded studies and reasons for exclusion were provided), and item #9 (whether a satisfactory technique for assessing the risk of bias in individual studies was used) all had a compliance rate of 83.33% (15/18); the compliance rate for item #11 (whether appropriate methods were used to combine results) was 72.22% (13/18); item #13 (whether bias was explained) had a compliance rate of 33.33% (6/18); and item #15 (whether the authors properly investigated and discussed the potential impact of publication bias) was 55.56% (10/18). The results of the methodological quality evaluation are presented in Table 2.

Based on the GRADE evaluation system, of the 24 outcome indicators, 1 study provided high-quality evidence, 5 provided moderate-quality evidence, 11 were of low quality, and 7 were of very low quality. The detailed results of the evidence quality assessment are presented in Table 3.

This umbrella review identified a range of intra-personal, interpersonal, and social factors influencing digital addiction. In a comprehensive examination of the aforementioned factors, several studies were noted for their methodological assessment. The studies conducted by Ding et al [32] on self-esteem and self-control, by Zewde et al [39] on male gender, and by Wan et al [37] on social support, all achieved moderate rankings for methodological rigor. However, the evidence for these factors was considered circumstantial or questionable in accuracy, resulting in a classification of low quality. On the other hand, study by Zhuang et al [8] concluded that a strong parent-child relationship significantly mitigated digital addiction, while study by Zewde et al [39] identified urban living as a risk factor. In addition, study by Hao et al [42] on childhood adverse experiences, and study by Wu et al [43] on social anxiety, findings from both of these studies received ratings of moderate methodological quality.

Intrapersonal factors significantly influence digital addiction, with our study evaluating 21 such factors, of which only 3 received moderate ratings on the AMSTAR 2 and 8 on the GRADE scale, all focusing on the associations between personality traits and digital addiction. Despite some methodological limitations, our findings were consistent with finding by Geng et al [44], indicating that self-control inversely associated with digital addiction, suggesting that individuals with lower self-regulation are more prone to emotional and impulsive behaviors, thereby fostering harmful digital usage patterns. In addition, specific personality traits can serve as protective factors against digital addiction [44,45]. As previous research has indicated, self-esteem, which is a reflection of an individual’s sense of perceived value, is associated with higher self-efficacy and self-control, ultimately reducing addiction risk [46]. Traits such as self-consciousness, characterized by self-discipline and organizational skills, tend to promote caution in digital interactions, favoring real-life experiences [47]. Furthermore, individuals exhibiting higher levels of agreeableness and openness are less likely to encounter risks associated with digital addiction [36,47]. Conversely, traits such as narcissism, characterized by self-centeredness and lack of empathy, may drive individuals to seek internet-based validation, increasing addiction susceptibility [48]. Consistent with previous research [49-53], impulsivity is a notable risk factor for digital addiction, as highly impulsive individuals often struggle to resist internet-based temptations. Shyness may prompt individuals to engage more in digital socialization to avoid in-person interactions, potentially heightening addiction risk [54]. Thus, our results suggest that while positive personality traits may buffer against digital addiction, negative traits can increase vulnerability. At present, cognitive behavioral therapy and psychosocial interventions are recognized as efficacious strategies for mitigating the symptoms of internet addiction. Cognitive behavioral therapy has been shown to substitute impulsive behaviors with more reasoned actions and improve self-control [55], thereby contributing to the prevention and management of digital addition [56]. Psychosocial interventions have been effective in addressing internet addiction and enhancing self-esteem. By focusing on the psychological and social dimensions of the individual, these approaches promote the development of a positive personality and can be effective in curbing digital addition behaviors [57,58]. Furthermore, gender differences are crucial, as males are more likely to develop digital addictions, likely due to their preference for strategy, role-playing, and action genres, which are inherently more addictive [59].

Among the 4 interpersonal factors associated with digital addiction, 1 study was rated as moderate on both the AMSTAR 2 and GRADE scales, indicating that it provided reasonably convincing evidence through robust methodology [8]. This finding is consistent with previous research, which has demonstrated that enhanced parent-child relationships can effectively reduce the risk of digital addiction among adolescents [60,61]. Research suggests that there is a link between excessive parental screen time and problematic screen behaviors in their offspring. Excessive use of digital devices by parents reduces interaction with their children; this, along with their children’s emulation of their parents’ screen-using habits, contributes to the development of screen-using behaviors in adolescents. This pattern of behavior has the potential to evolve into digital addiction [62,63]. Therefore, providing guidance on the use of digital devices, especially for parents of young children, may become a viable intervention strategy to counteract the emergence of digital addiction in children.

Six of the included studies explored examined the influence of social factors on digital addiction, with 4 rated as moderate quality based on AMSTAR 2 criteria. However, only 3 of these studies received a moderate rating from GRADE. Evidence indicates that urbanization uniquely affects addiction; urban residents often face environmental pollution and discomfort associated with city life, which can heighten psychological stress and contribute to digital addiction [64]. In addition, a lack of social support emerges as a critical risk factor for internet addiction. Insufficient social connections can lead to unhappiness, prompting individuals to seek emotional fulfillment through social media, potentially worsening addiction issues [65,66]. Late childhood, a pivotal period for social and emotional development, shows a marked negative association between internet engagement and various well-being dimensions. This correlation suggests that the rising levels of unhappiness and anxiety in children are closely linked to increased internet use. Furthermore, adverse childhood experiences related to home environments and parental pressures significantly influence children’s development, leading to increased time spent on digital media [[67,68]]. In addition to the previously mentioned parental digital guidance, group counseling has also been reported to play a significant role in enhancing the self-control abilities of school-aged children with internet addiction [69]. However, when parents are also involved in group counseling, research has proven that it requires parents to have previously successfully dealt with challenges such as social anxiety, social isolation, and social skill deficits, or else group counseling will be much less effective [69].

Intrapersonal factors influencing digital addiction have been extensively studied, predominantly focusing on personality traits. However, most meta-analyses seem to suffer from methodological flaws, and even those without serious issues often reveal publication bias in original studies, thereby diminishing the quality of the evidence [32,39]. Therefore, it raises concerns about the risk of publication bias, where positive results are more readily published, potentially obscuring the true nature of the cause-and-effect relationship [70]. In contrast, fewer studies examining interpersonal and social factors demonstrate better methodological rigor and evidence quality. These factors often exert their influence on digital addiction indirectly by affecting psychological traits or personality characteristics. For example, Malak et al [71] used structural modeling to show that social support mediates the relationship between anxiety and stress, contributing to internet-based gaming disorders, a form of digital addiction [66]. Thus, it is misleading to solely consider interpersonal and social factors when exploring correlates of digital addiction. Instead, this indicates the need for further research to elucidate the complex interplay between these domains.

While this umbrella review evaluated numerous factors associated with digital addiction, several limitations persist. First, the focus on published meta-analyses may lead to the omission of factors that are less studied or have not garnered significant attention [72]. Second, the broad scope of the review complicates the generalization of results due to the diversity of research findings included. Finally, many systematic reviews analyzed similar exposure factors and were conducted within a relatively short timeframe, resulting in potential overlap among initial studies. To mitigate this, we prioritized the inclusion of the most recent meta-analyses with the highest number of studies and aimed to exclude those exhibiting considerable overlap with original studies.

In conclusion, a range of intrapersonal, interpersonal, and social factors significantly influences an individual’s susceptibility to digital addiction. A rigorous evaluation using the AMSTAR and GRADE frameworks identified negative parent-child relationships, urban residence, adverse childhood experiences, and social anxiety as credible risk factors. These findings are essential for informing effective public health strategies aimed at preventing digital addiction. However, further refinement of the methodologies used to analyze these factors is necessary to enhance the robustness of future research.