In April 2022, during the lockdown caused by the COVID-19 pandemic, a cluster sample of children in Baoding City, China, was investigated. Inclusion criteria: (a) children from the elementary grades of primary school; (b) isolated at home for at least 1 week due to the COVID-19 epidemic; (c) clear consciousness and normal perceptual function; (d) volunteered to participate in the study and provided the informed consent. Exclusion criteria: (a) cognitive dysfunction or communication disorder; (b) incomplete questionnaires. The Spence Children’s Anxiety Scale-Short Version (SCAS-S) and Internet Gaming Disorder Scale (IGDS) powered by www.wjx.cn were used to survey the children after explaining the purpose and method of the research to the children and their parents. The parents were required to read the questions and response options of each item to their children, and then the children answered according to their own feelings. A total of 722 questionnaires were distributed, and incomplete questionnaires were deemed invalid. Finally, 667 valid questionnaires were collected for an effective recovery rate of 92.38%. The present research followed the Helsinki Declaration and was approved by the Ethics Committee of Xijing Hospital of Air Force Medical University (Project No. CHiCTR1800019761).

Statistics of demographic characteristics and scale scores were calculated by SPSS 22.0 software. Network model construction, bridge centrality analysis, and robustness tests were conducted by R4.1.1 software.

The demographic characteristics of the sample are displayed in Table 1. The means and standard deviations of the variables in the anxiety-IGD network are displayed in Table 2.

The anxiety-IGD network is displayed in Figure 1A. Logically, there was a maximum of 45 edges across the communities in the network; in the present research, there were 23 cross-community edges (weights ranged from −0.03 to 0.12). Overall, positive edges accounted for the vast majority of cross-communities edges (86.96%). The positive cross-community edges contained A1 “separation anxiety”—IGD3 “giving up other activities” (edge weight = 0.04), A1 “separation anxiety”—IGD5 “escape” (edge weight = 0.02), A1 “separation anxiety”—IGD9 “withdrawal” (edge weight = 0.01), A2 “social phobia”—IGD1 “preoccupation” (edge weight = 0.04), A2 “social phobia”—IGD2 “tolerance” (edge weight = 0.04), A2 “social phobia”—IGD3 “giving up other activities” (edge weight = 0.06), A2 “social phobia”—IGD4 “continuing despite problems” (edge weight = 0.01), A2 “social phobia”—IGD6 “negative consequences” (edge weight = 0.04), A2 “social phobia”—IGD8 “deception” (edge weight = 0.02), A3 “panic disorder”—IGD5 “escape” (edge weight = 0.12), A3 “panic disorder”—IGD6 “negative consequences” (edge weight = 0.06), A3 “panic disorder”—IGD7 “loss of control” (edge weight = 0.02), A3 “panic disorder”—IGD9 “withdrawal” (edge weight = 0.01), A4 “physical injury fear” —IGD5 “escape” (edge weight = 0.01), A5 “generalized anxiety”—IGD1 “preoccupation” (edge weight = 0.003), A5 “generalized anxiety”—IGD4 “continuing despite problems” (edge weight = 0.03), A5 “generalized anxiety”—IGD5 “escape” (edge weight = 0.06), A5 “generalized anxiety”—IGD6 “negative consequences” (edge weight = 0.03), A5 “generalized anxiety”—IGD8 “deception” (edge weight = 0.01) and A5 “generalized anxiety”—IGD9 “withdrawal” (edge weight = 0.04). The negative cross-community edges contain A1 “separation anxiety”—IGD6 “negative consequences” (edge weight = −0.01), A4 “physical injury fear”—IGD6 “negative consequences” (edge weight = −0.03) and A4 “physical injury fear”—IGD8 “deception” (edge weight = −0.02). The correlation matrix of the network model can be found in Supplementary Table 1.

The accuracy test of edge weights is displayed in Supplementary Figure 1. The 95% confidence interval of edge weights in the anxiety-IGD network was relatively narrow, indicating that the edge estimation was accurate. The difference test of edge weights is displayed in Supplementary Figure 2.

The BEI is displayed in Figure 1B. The nodes with the top 80th percentile BEI were A2 “social phobia” (0.20), A3 “panic disorder” (0.21) and IGD5 “escape” (0.22). As shown in Supplementary Figure 3, the BEI indices of A2 “social phobia,” A3 “panic disorder” and IGD5 “escape” were statistically larger than those of most other nodes. As described in Supplementary Figure 4, the average correlation of BEI indices of the subsample and the original sample showed a relatively gentle downward trend with the reduction of subsample. The value of the CS-coefficient on the BEI was 0.52, which was larger than 0.5, indicating adequate stability.

The detailed relationships between variables revealed by network analysis can provide indications of the correlation pathways through which different aspects of anxiety affect IGD (48, 49). These correlation pathways may represent underlying interaction mechanisms of anxiety and IGD. In light of this, we discussed the most powerful pathways bridging various aspects of anxiety and contents in IGD.

Among the four pathways through which “separation anxiety” correlated with IGD, the strongest pathway was “separation anxiety”—“giving up other activities,” while among the six pathways through which “social phobia” correlated with IGD, the strongest pathway was “social phobia”—“giving up other activities.” Separation anxiety is described as the irrational fear or anxiety of being separated from family or close attachment (12, 50). According to Bowlby’s attachment theory, the quality of parental care a child receives in early years has profound effects on future mental health (51). Children who do not receive effective care and attention from parents are incapable of dealing with separation anxiety. Many children have been separated from their relatives and even parents due to the COVID-19 lockdown, and separation anxiety caused by the disruption of close relationships drives children to online games, which may meet their emotional requirements. Social phobia is described as the excessive fear of social occasions and interactions with others and may result in social isolation (52). Baker and Hudson (53) found that children with social phobia had narrower social networks and poorer friendship quality and were more likely to experience social alienation than children with other anxiety symptoms. This may be due to increased sensitivity to social exclusion and social withdrawal in adolescents with symptoms of social phobia leading to deficits in interpersonal skills (54). Therefore, children with social phobia tend to abandon offline social activities and indulge in virtual worlds. The reason is that the internet provides a platform for them to communicate anonymously.

Among the four pathways through which “panic disorder” correlated with IGD, the strongest pathway was “panic disorder”—“escape,” while among the six pathways through which “generalized anxiety” correlated with IGD, the strongest pathway was “generalized anxiety”—“escape.” Panic disorder is characterized by misinterpretations of explosive sensory and emotion, i.e., heart palpitations, dyspnea, dizziness, and near-death experience (55, 56). Inconsistent with panic disorder, generalized anxiety is characterized by persistent, relatively mild symptoms on most days for at least 6 months, such as nervosity, irritability, fatigue, attention and memory problems, and insomnia (12, 57, 58). Both panic disorder and generalized anxiety lead to evident negative emotional experiences, and when there is no appropriate outlet, such as during the COVID-19 lockdown, children often turn to online games to escape these emotional experiences.

Among the three pathways through which “physical injury fear” correlated with IGD, the strongest pathway was “physical injury fear”—“negative consequences.” The symptoms of physical injury fear are similar to those of specific phobia; they all manifest as extreme and persistent fear of a certain object or situation, such as heights, tunnels, darkness, or worms (59). COVID-19 may cause death and sequelae (60), and related reports have led to increased sensitivity to physical injury fear. In addition, the children in the present study were too young to identify the authenticity of information, so they were excessively afraid of dangerous information and were prone to online gaming addiction, which could lead to a series of negative consequences.

Anxiety is comorbid with IGD (61), and the activation of bridge symptoms can increase the risk of transferring from one disorder to another (37). In network analysis, bridge centrality explained the comorbidity and reciprocity of different disorders (62). The current study identified bridge symptoms between anxiety and IGD in children. Bridge symptom identification helps reveal fecund information on comorbidities and provide prioritized clinical targets to prevent co-occurrence (63, 64). The BEI indices of “social phobia” and “panic disorder” were the largest in the anxiety community. From the perspective of the network, the intervention of “social phobia” and “panic disorder” has a greater impact on IGD than other nodes of the anxiety community, so they are potential best targets for interventions. Previous studies have confirmed the relationship between social phobia and IGD (65). In fact, children with social phobia have more personal space and time, which is a crucial risk factor for IGD (66). Given the acuteness and recurrence of panic disorder symptoms (67), individuals may have persistent worries, and online games may provide a distraction. Gaming online is an escape strategy to alleviate anxious emotions, and people may spend excessive amounts of time on online games as a coping mechanism to escape from the reality of their worries and difficulties (24–26). However, online gaming does not actually solve the problem, and wasting too much time on games makes things worse. A vicious cycle may further exacerbate psychological distress in people with IGD. Therefore, there is a need to implement targeted interventions to improve intervention efficiency and save medical resources.

Many intervention methods have been proven effective for anxiety. Cognitive behavioral therapy is the most experienced and effective psychosocial therapy for treating social phobia in adults and children (68); furthermore, the well-known Gestalt therapy is also effective in reducing anxiety in elementary school children (69). Finally, parent–child interaction therapy can prevent IGD by reducing the child’s level of anxiety by establishing a good parent–child relationship (70, 71).

Several limitations should be pointed out in the present study. First, since cross-sectional data were used in our research and the relationships in networks were bidirectional, the determination of causality needs further support from longitudinal research. Second, the sample scope was limited to students in elementary school. Although we aimed to detect and intervene as early as possible, the results may not be suitable for generalization to older children. Finally, there was no control group, and the comparison of networks of students in lockdown and those in school may provide more meaningful findings.