Gender differences in symptom structure of adolescent problematic internet use: A network analysis

The present study is part of a national project in mainland China. This project covered thirty-seven primary and secondary schools across six different regions of the Chinese mainland. The regions included the regions of North China, Central China, South China, East China, Northwest China, and Southwest China, accounting for 25.5%, 16.9%, 15.0%, 24.1%, 7.4%, and 11.0% of the total participants, respectively. The invited schools were in the average socio-economic levels of each region. To ensure that our results were based on valid responses, we embedded two response quality items that require the participants to choose the “forced” answer in the survey and used them to exclude inattentive respondents, resulting in an effective recovery of 87.5%. The final sample included 4884 adolescents: 2519 females (51.6%) and 2365 males (48.4%). Their ages ranged from 10 to 19 years (M_age = 13.83, SD = 2.41). Approximately half of the participants (50.4%) were the only child in the family and nearly all the adolescents (96.5%) lived with their parents.

We collected data through an e-questionnaire website (Questionnaire Star) from April 23 to May 7, 2020. First, we obtained permission from the participating schools for data collection. Second, we sent the website link of the e-questionnaire to teachers at the schools; we explained the purpose of our study and the ethical principles of scientific research. Finally, the teachers sent the link to students and their parents after obtaining their informed consent. We included our email address (in case further support or assistance was needed by the participating families) and an informed consent statement on the first page of the e-questionnaire. All the participants and their parents were informed about the purpose of the study, and they were free to withdraw from the study at any time. The Research Ethics Committee of Beijing Normal University approved the present study.

We assessed PIU symptoms using the 26-item Chinese Internet Addiction Scale-Revised which has achieved high split-half reliability, test-retest reliability, and convergent validity and are widely used in Chinese adolescents [4, 22]. The scale comprises five symptom clusters: compulsive use, withdrawal symptoms, tolerance symptoms, interpersonal and health problems, and time-management problems. The detailed descriptions of PIU symptoms appear in Table 1. The responses were made on a 4-point Likert-type scale, ranging from 1 (“not like me at all”) to 4 (“completely like me”). Higher scores indicated higher levels of PIU. The internal consistency coefficient of CIAS-R was 0.97.

We used SPSS version 23.0 and R version 3.6.3 in RStudio 1.2.5033 for data analysis: the former for descriptive analysis and the latter for network analysis. The option setting of the e-questionnaire did not allow participants to skip any items. Thus, no item-level data were missing. We applied the qgraph package in R to estimate network structure. We employed a Gaussian graphical model [26], using the graphical lasso in combination with an extended Bayesian information criterion [27]. Each node represented a symptom, and the edge connected different nodes; higher thickness of the edge represented a stronger connection between nodes [20].

To identify central symptoms that were highly connected with other symptoms, we used the centralityPlot function to estimate the centrality of each PIU symptom, including three indicators [28]. Strength centrality represents a greater influence on the individual owing to the coexistence of several symptoms [29]; closeness centrality represents how a symptom quickly spreads to other symptoms [30]; betweenness centrality represents possible interventions when examining comorbidity, which acts as a bridge connecting the whole network [31].

We employed the NetworkComparisonTest function to estimate the differences between female and male networks, indicated by global and local differences. The network structure invariance test (comparing how the nodes were connected) and the global strength invariance test (comparing the weighted absolute sum of all edges) quantified global differences; the edge and node invariance tests (examining which specific edges and nodes were different) quantified local differences [25].

We applied the bootnet package to investigate the accuracy and stability of the networks. We examined edge-weight accuracy using the 95% confidence intervals (CIs) of the edge-weight bootstrap: higher degrees of overlapping between the CIs indicated better accuracy. Centrality stability was investigated using a case-dropping bootstrap, adopting a centrality stability coefficient as a reference index. A value higher than 0.25 is acceptable; a value higher than 0.5 is excellent [32]. For example, the value 0.5 suggests that if our sample decreases from 95 to 25% of the original sample, the correlation estimates between the subsample and the original entire sample remained 0.5.

Networks of PIU symptoms for females and males appear in Fig. 1. There were 26 nodes and 325 edges (26*(26–1)/2) in both networks. Of the edges, 53.5% and 58.2% were non-zero-weight in female and male networks, respectively, indicating that most PIU symptoms are closely intercorrelated with each other with partial correlation estimation.

Figure 2 shows the centrality indexes for females and males. The red dots represent the strongest indexes. In both networks, the highest strength centrality was reluctant to stop (PIU6). For the female network, the highest strength centrality was “increasing time for satisfaction” (PIU24); the highest closeness centrality was “uncontrollable checking” (PIU19); and the highest betweenness centrality was “reluctant to stop” (PIU6). For the male network, the highest strength centrality was “feeling depressed” (PIU10); the highest closeness and betweenness centrality was “failure to stop” (PIU22).

The female and male networks were moderately accurate (Fig. A1): there was considerable overlap between the 95% CIs of the edge weights. The stability of both networks was excellent (Fig. A2): the female network’ strength was 0.75, closeness 0.75, and betweenness 0.60; the male’s strength was 0.75, closeness 0.60, and betweenness 0.52.

The global strength invariance test showed that the network strength was significantly stronger in males (global strength = 13.04) than in females (global strength = 12.40, p = .046). The network invariance test indicated significant differences in the network structures between genders (p = .048).

Table A1 presents the results of the edge invariance test. The edge invariance test revealed two notable points. First, the edge between “uncontrollable checking” (PIU19) and “failure to stop” (PIU22) was stronger in females (p = .019); conversely, the edges between “uncontrollable checking” (PIU19) and “empty life” (PIU20) (p = .026) as well as between “empty life” (PIU20) and “failure to stop” (PIU22) were stronger in males (p = .009). Second, “fewer interactions with families” (PIU17) was more strongly connected with “backache” (PIU13) (p = .007) and “physical deterioration” (PIU21) (p = .025) in males, whereas “fewer leisure activities” (PIU18) was more strongly connected with “sitting up online” (PIU26) (p = .033) and “longer than intended” (PIU3) (p = .019) in females.

Table A2 presents the results of the node invariance test. The node invariance test revealed significant differences in the centralities. The strength centralities of “too much time” (PIU1), “impaired relationships” (PIU7), and “backache” (PIU13) were significantly higher for males than for females (p = .004; p = .036; p = .048). The closeness centralities of “reduced involvement with friends” (PIU12), “fewer interactions with families” (PIU17), “fewer leisure activities” (PIU18), and “uncontrollable checking” (PIU19) were significantly higher in females than in males (p = .037; p < .001; p = .002; p = .010). The betweenness centrality of “fewer interactions with families” (PIU17) was significantly higher in females than in males (p = .006).

The symptoms with the highest strength centrality were either the same or different in female and male adolescents. “Reluctant to stop” (PIU6) exerted the largest effect in both genders, indicating a cross-sample central symptom. Although the participants just wanted to surf the Internet for a little while, they felt reluctant to discontinue, which suggests increasing over-dependence [4]. One reason could be that the Internet has emerged as a vital platform to fulfill social needs for adolescents, and adolescence is a critical period for searching for relatedness [34]. Another possibility is that adolescents nowadays rely greatly on the Internet for recreation. Therefore, they may experience negative feelings, such as restlessness and loneliness when disconnected from the Internet [1]. This symptom could have a severe impact on adolescents because it could allow other symptoms to occur [29].

Some core symptoms were differentiated by gender. “Increasing time for satisfaction” (PIU24) was a particularly central symptom in female adolescents, whereas “feeling depressed” (PIU10) was a particularly central symptom in male adolescents. The differential symptoms by gender may reflect different motivations and types of Internet use between female and male adolescents. One meta-analysis found that females were more likely to exhibit social media addiction and less likely to display Internet gaming disorder than males [35]. Since female adolescents often value relationships more than males, they tend to use the Internet for social networking in response to feelings of emptiness when their social needs are not otherwise fully met [36]. Once the threshold of Internet use increases, individuals require more time to reach the degrees of satisfaction previously achieved [37]. In this case, “increasing time for satisfaction” (PIU24) may greatly trigger other symptoms in females, such as “less sleep” (PIU23) found in the present study. The excessive time on Internet use may reduce adolescents’ sleep time, and this phenomenon is more common among females [38]. Thus, the central symptom of “increasing time for satisfaction” (PIU24) is more likely to lead to sleep problems in female adolescents.

In addition, we found that male adolescents feel depressed when not online, which is consistent with previous studies that depressive feeling was dominant to males’ PIU [17, 18]. One explanation is that male adolescents are more likely to engage in PIU to gain feelings of success and achievement that can be provided by online gaming [39]. While Internet use decreases or is discontinued, males may be prone to depressive feelings. Consequently, “feeling depressed” (PIU10) could severely influence other symptoms of PIU among adolescent males.

Our results support the hypothesis that global connectivity would be stronger among adolescent males than among females, which is in line with previous findings about males being at a higher risk of PIU [10, 14]. A highly connected network is more likely to develop strong self-reinforcing loops of symptom interactions [24]. Our results indicate that once male adolescents begin to engage in PIU, they tend to display a more strongly connected symptom network than females, and ultimately, males might be prone to involve in long-term PIU. One possible explanation is that males easily become addicted to Internet gaming, which impairs executive control [13]. Worse, decreased impulse control can dramatically exacerbate PIU [40]. Another explanation could be that compared with females, males tend to suppress or avoid emotions and apply fewer adaptive emotional regulation strategies [41]. Therefore, males are more likely to use the Internet as a coping strategy—although it may trigger vicious circles of negative emotion and PIU, in turn aggravating PIU symptoms. In summary, our findings help explain that chronically higher levels of PIU are more commonly observed in males than in females [42].

In addition to global strength, gender differences also exist in the network structure of PIU, which reflects different symptom associations by gender. In general, our findings indicate significant gender differences in the global connectivity of PIU networks. Below, we discuss more detailed results from the perspective of local connectivity.

We observed that local connectivity (indicated by symptom associations and central symptoms) varied by gender, which extends the literature on gender differences in PIU symptomatology. We found that female and male adolescents differed in the interacting ways and the importance levels of PIU symptoms. The following discussion focuses on two of the gender differential features of PIU symptoms: direct and indirect associations, and different effects of interpersonal symptoms as well.

First, the edge between “uncontrollable checking” (PIU19) and “failure to stop” (PIU22) differed by gender. If female adolescents are unable to resist going online, they are unable to reduce Internet use. Alternatively, if they fail to cut down their Internet use, they may feel that they are losing control. Among males, this edge existed through an irrational belief that “life is boring and empty without the Internet” [21], which is a core belief among individuals who have difficulty controlling their Internet use. Because they believe that life is boring without the Internet, they become loss of control and fail to leave it. Hence, the treatment for changing this irrational belief could be helpful to adolescent males.

In addition, interpersonal symptoms of PIU co-occurred quite differently between female and male adolescents. Among females, “fewer leisure activities” (PIU18) was highly connected with time-related problems, such as “sitting up online” (PIU26) and “longer than intended” (PIU3). Specifically, when female adolescents spend too much time on the Internet and even through the whole night, they may have not enough time and effort to put in the real life and have few leisure activities offline. As for males, “fewer interactions with families” (PIU17) was highly connected with health problems, including “backache” (PIU13) and “physical deterioration” (PIU21). The results indicate that when male adolescents indulge in the Internet and have few interactions with families, they are inclined to select the Internet for recreation instead of sports exercises, which may be harmful to their physical health. Thus, their interpersonal and health problems because of long-time Internet use are strongly connected. These gender differences extend the findings of previous studies on gender differences in the patterns and consequences of the PIU [11, 12].

Moreover, regarding core symptoms of PIU, despite some common areas across gender, several symptoms differed between female and male adolescents, most of which were related to interpersonal and health problems owing to PIU. Gender differences in the association between PIU and interpersonal impairments have been reported previously [19]. We found more specific symptom differences that female adolescents showed greater symptoms of social withdrawal than males, including “reduced involvement with friends” (PIU12), “fewer interactions with families” (PIU17), and “fewer leisure activities” (PIU18). Conversely, male adolescents presented more interpersonal conflicts than females, including “impaired relationships” (PIU7) and “complaints by others about too much time online” (PIU1). The gender differences could be explained by the different ways females and males express conflicts [43]: when adolescents experience interpersonal problems with families or friends because of their over-dependence on the Internet, females tended to express conflict covertly (e.g., fewer or avoidant interactions) and males tend to express conflict overtly (e.g., conflictive interactions).

Specifically, the Internet creates excellent opportunities for adolescents with relational resource deficits to be socially efficacious and develop meaningful relationships online [37]. On the one hand, female adolescents whose social needs are not fully met in the real life tend to use the Internet for compensation [13]. However, such evasion could further decrease their real-life interactions and quickly activate other PIU symptoms. On the other hand, male adolescents who use the Internet as a way of avoiding interpersonal conflicts may instead aggravate conflicts with families and peers and damage real-life relationships. Thus, the conflictual relationships that develop because of PIU may trigger other PIU symptoms and severely impair male adolescents [29].

The symptoms of PIU in female and male adolescents were either the same or different, which leads to gender-specific differentiated intervention strategies. For both genders, “reluctant to stop” (PIU6) should be the target symptom in interventions. Specifically, it is necessary to help adolescents understand why they do not want to reduce the time they spend on the Internet and then provide related strategies. For example, social skills deficit theory asserts that adolescents with poor social competence may be more likely to establish relationships in the anonymous virtual world than those occurring in real life, which may result in the vulnerability of PIU [44]. Therefore, interventions focused on enhancing adolescents’ social skills are critical to decreasing their PIU symptoms [45].

In addition, it is also important to focus on “increasing time for satisfaction” (PIU24) for females and “feeling depressed” (PIU10) for males. For the symptom of “increasing time for satisfaction” (PIU24), some educational programs aiming at organizing daily activities and developing new healthy Internet use concepts and habits may be effective [46]. In terms of the symptom of “feeling depressed” (PIU10), mindfulness-based cognitive behavioral intervention can regulate an addiction-related distressed emotional state [47]. Moreover, female and male adolescents showed different features of PIU symptom clusters (e.g., interpersonal problems): social withdrawal symptoms for females and interpersonal conflict problems for males, which requires differential interventions on their motivations. Meanwhile, interpersonal and time management problems are more likely to co-occur among female adolescents, whereas interpersonal and health problems are more likely to co-occur among male adolescents. Thus, cognitive behavioral therapy focused on improving time management skills may be more beneficial to female adolescents [48] and sports intervention may be more beneficial to male adolescents [49].

Despite its novel findings and significant implications, this study has some limitations. First, we used self-report measures to determine participants’ PIU symptoms, which could not indicate a formal diagnosis and might be limited by self-report biases [50]. Future studies on PIU are encouraged to use structured clinical interviews and collect real-world behaviors. Second, considering the investigation’s cross-sectional design, directionality between PIU symptoms should not be causally interpreted. A future longitudinal study should identify the symptoms that exhibit the highest out-strength and may strongly influence other symptoms over time.

Below is the link to the electronic supplementary material.

Approval was obtained from the ethics committee of Beijing Normal University (No. 202109280046).

The authors have no relevant financial or non-financial interests to disclose.

The datasets and analytical code during the current study are available from the corresponding author upon reasonable request.