A 12-month prospective longitudinal cohort study was conducted in Guangzhou and Chengdu, China, which had population sizes of 18.3 and 16.6 million, respectively. The baseline survey and follow-up survey were conducted in December 2018 (T1) and December 2019 (T2), respectively. Four junior middle schools from Guangzhou and 2 from Chengdu were selected by convenience sampling. All grade 7 and 8 (7 and 8 years of formal schooling) students of those schools were invited to join this study. Grade 9 students were not invited as they would leave the school before the end of the 12-month follow-up period. As we were investigating IGD conversion, those with IGD at baseline were removed from data analysis (an exclusion criterion). The structured questionnaire was self-administered anonymously using paper and pencil, in the classroom setting, and in the absence of the schoolteachers. Well-trained fieldworkers briefed the students about the objective, content, logistics, and voluntary nature of the study. The reporting of this study follows the Journal Article Reporting Standards for Studies Using Structural Equation Modeling [33].

The ethics approval of this project was obtained from the Survey and Behavioral Research Ethics Committee of the Chinese University of Hong Kong (SBRE-18‐430). This study used a parental opt-out procedure, and student participation was strictly voluntary and uncompensated. Due to the anonymous study design, written informed consent was waived; instead, the act of completing and returning the questionnaire indicated informed consent, which was explicitly stated on the cover page of the questionnaire. To facilitate longitudinal data matching while guaranteeing anonymity, a unique matching code was used combining participants’ birthdays, the last 4 digits of their parents’ mobile phone numbers, and the last 2 letters of their parents’ given names. Consequently, no personally identifiable information was collected, and no identifiable information could be identified in both the datasets analyzed and this manuscript. No incentives were provided to participants, parents, and schoolteachers in this study.

As path analysis was used as the key statistical method, the rule of thumb for its sample size planning requires a minimum of 10 cases per item or an observed variable [36]. Accordingly, this study requires a sample size of at least 140 participants, and the current sample size is adequate.

Attrition analysis was conducted by examining between-group differences in the background variables and 2 parental gaming behaviors at T1 between follow-ups and dropouts. The Little missing completely at random (MCAR) test was conducted to evaluate the pattern of missing data due to dropouts; multiple imputation (n=20) was used to address missing data. Descriptive statistics and normality tests (via the Kolmogorov-Smirnov and Shapiro-Wilk tests) were performed for the key studied variables. Spearman correlation coefficients were derived for the correlations between two parental gaming behaviors (T1), parental supportive attitude (T2), behavioral intention of increasing gaming time (T2), and IGD conversion (T2). Path analysis was performed to test the hypothesized mediation mechanisms, with the weighted least square mean and variance adjusted estimator. The models were adjusted for all studied background factors and respective baseline levels of mediators and outcomes (ie, parental supportive attitude at T1 for parental supportive attitude at T2). Satisfactory model fit indices for the path analysis models included comparative fit index ≥0.90, root mean square error of approximation ≤0.08, and standardized root mean square residual ≤0.08. Path analysis was performed by using Mplus (version 7.0; Muthén & Muthén) and other analyses by SPSS (version 21.0; IBM Corp). Statistical significance was indicated as P<.05 and 95% CI not involving zero.

A total of 2436 students completed the baseline questionnaire; 264 (10.8%) with IGD at T1 were removed. Of the remaining 2172 participants, 1721 (79.2%) were matched for their baseline and follow-up questionnaires (Figure 1). Those lost to follow-up were more likely to be a Chengdu student, older, female, not having moved to the city, and having parental educational levels of junior middle school or below; their parents also played internet games and invited them to cogame more frequently (Table 1). The Little MCAR test was statistically significant (χ²₃₀₂=599.5; P<.001), suggesting that data were not missing completely at random.

Of the 2172 participants based on the 20 imputed dataset, the mean age was 12.56 (SD 0.02, 95% CI 12.52-12.60) years. More than half were female (n=1102, 50.7%, 95% CI 48.7%-52.8%); about two-fifths of their fathers (n=953, 43.9%, 95% CI 41.8%-46.0%) and mothers (n=976, 44.9%, 95% CI 42.8%-47.0%) had received junior middle school or below education. Close to one-tenth self-reported a poor or very poor family financial situation (n=223, 10.3%, 95% CI 9.0%-11.7%). The prevalence of IGD conversion over a 12-month follow-up period was 5.2% (n=113, 95% CI 4.4%-6.1%; Table 2). The mean scores were 1.66 (SD 0.02; 95% CI 1.63-1.69; range 1‐4) for perceived parental gaming frequency, 1.21 (SD 0.01, 95% CI 1.19-1.23; range 1‐4) for perceived parental invitation for cogaming, 2.01 (SD 0.02, 95% CI: 1.96-2.06; range 1‐5) for perceived parental supportive attitude, and 1.91 (SD 0.02, 95% CI 1.87-1.95; range 1‐5) for behavioral intention of increasing gaming time.

Kolmogorov-Smirnov and Shapiro-Wilk tests showed that all studied key variables violated the normality assumption (all P<.001). Spearman correlation analysis showed that perceived parental gaming frequency, perceived parental invitation for cogaming, parental supportive attitude, and behavioral intention of increasing gaming time were significantly and positively correlated with IGD conversion. Perceived parental gaming frequency and parental supportive attitude were positively correlated with behavioral intention of increasing gaming time, but the correlation between parental invitation for cogaming and behavioral intention was statistically nonsignificant. The 3 variables related to parental gaming behavior and attitudes were positively correlated with each other (Table 3).

Figure 2A presents path analysis testing the mediation effects of parental supportive attitude and behavioral intention of increasing gaming time on the association between perceived parental gaming frequency and IGD conversion. The 1-mediator indirect path via parental supportive attitude (β=0.02, 95% CI 0.01-0.04) and the 2-mediator indirect path (first via parental supportive attitude then via behavioral intention; β=0.006, 95% CI 0.002-0.009) were statistically significant, whereas the 1-mediator indirect path via behavioral intention (β=0.01, 95% CI −0.01 to 0.03) was not. As the association between perceived parental gaming frequency and IGD conversion was statistically nonsignificant (β=0.07, 95% CI −0.06 to 0.19), full mediation via the 2 indirect paths was observed. Similar patterns were observed for the mediation between perceived parental invitation for cogaming and IGD conversion (Figure 2). The 1-mediator indirect path via parental supportive attitude (β=0.02, 95% CI 0.01-0.04) and the 2-mediator indirect path (first via parental supportive attitude then via behavioral intention; β=0.005, 95% CI 0.001-0.009) were statistically significant, whereas the 1-mediator indirect path via behavioral intention of increasing gaming time (β=0.01, 95% CI −0.01 to 0.03) was not. As the association between perceived parental invitation for cogaming and IGD conversion was also statistically nonsignificant (β=0.11, 95% CI −0.04 to 0.26), full mediation via the 2 indirect paths was observed. Both models support the original hypothesized models, and the other relevant statistics are presented in Multimedia Appendix 1.

This longitudinal study revealed the prospective associations between 2 parental gaming-specific factors of perceived parental gaming frequency and parental invitations for cogaming at T1 and IGD conversion at T2 among Chinese adolescents. Furthermore, the 2 individual path analysis models revealed that these prospective associations were fully mediated by a 2-mediator indirect path (first via parental supportive attitude toward the adolescent’s gaming behavior and then via adolescents’ behavioral intention of increasing gaming time) and the 1-mediator indirect path via parental supportive attitude, but not the other 1-mediator indirect path via behavioral intention. The results have implications for IGD interventions involving parents.

This study documented the prevalence of IGD conversion of 5.2% (113/2172) among previously non-IGD adolescents in China, which was slightly lower than 11.7% [22] and 1.7% [37] among participants aged 14 years or older in Western countries. Our results identify 2 parental gaming behaviors positively associated with IGD conversion among Chinese adolescents, highlighting the role of environmental triggers [12]. Specifically, higher frequencies of parental gaming and invitations for cogaming may serve as environmental endorsements of adolescent gaming and facilitate their IGD conversion. Aside from the observational learning perspective of SCT, these parental gaming behaviors can be cues to action, prompting intensive gaming that may increase adolescent IGD conversion [38]. Hence, health education should target not only adolescents but also their parents. Behavioral intention of increasing gaming time was also positively associated with IGD conversion, supporting TPB that behavioral intentions predict corresponding health behaviors [30]. A behavioral intention is often not translated into the actual behavior [39], but this “intention-behavior gap” could be reduced by implementation planning interventions [40]. For instance, the if-then intervention (“if” facing temptation and “then” implementing a mentally rehearsed preventive response) [40] has demonstrated strong effectiveness in achieving behavioral change goals [41].

In addition, the significant indirect paths shed insights into how to reduce IGD conversion. Both parental gaming behaviors were prospectively associated with parental supportive attitude that was positively associated with behavioral intention of increasing gaming time, which was associated with a higher risk of IGD conversion. The observed full mediation effects suggest that the mediation mechanism via parental supportive attitude alone and first via parental supportive attitude and then via behavioral intention of increasing gaming time may fully explain the associations between parental gaming behaviors at T1 and IGD conversion at T2. Notably, although both indirect paths reached statistical significance, the magnitudes of the involved associations were small. It is important to acknowledge that the statistical significance of these paths may be due to the large sample size of this study [42]. These small effect sizes reflect the multidimensional etiology of IGD conversion [43], wherein parental gaming behaviors and attitudes and personal cognition (behavioral intention) account for only a minute portion of the overall variance. Accordingly, while these variables (eg, parental gaming behaviors) are widespread and modifiable [12], they are better conceptualized as small, interacting components within a broader biopsychosocial network of risk factors, rather than stand-alone targets. This suggests that public health interventions should not rely on targeting single environmental factors but instead adopt holistic, multicomponent approaches that integrate family-level determinants alongside other individual and systemic factors. An additional remark is that the 1-mediator indirect path from parental gaming behaviors to IGD conversion via behavioral intention alone was statistically nonsignificant. This suggests that parental gaming behaviors were prospectively and directly associated with parental supportive attitude, but not behavioral intention, and that the association between parental supportive attitude and IGD conversion was mediated via behavioral intention of increasing gaming time. As TPB postulates that the association between attitude and behavior would be mediated via behavioral intention [30], this finding supports the TPB.

Overall, the earlier findings offer practical implications that prevention of IGD conversion should not neglect both parental and adolescent gaming behaviors and attitudes. Inevitably, parental gaming behaviors (eg, cogaming with children) and parental attitude are a double-edged sword, as they may both increase the risk of adolescent IGD [16-18] and improve parent-child relationships [44]. A delicate balance needs to be reached. Parents may use their gaming behaviors to set good examples in time management, nonexcessive gaming, and self-control for their children. Although excessive cogaming may be problematic, cogaming is a good opportunity for parent-child communication on healthy gaming [15]. Parents should be informed that parental supervision and monitoring are protective factors against adolescent IGD [14] and that parental supportive attitude toward the adolescent’s gaming behavior needs to be carefully conveyed to the children to avoid the wrong impression of absent parental supervision and monitoring. Hence, effective family-based IGD interventions on health education, parenting skills, and improving family relationships [45,46] may consider integrating these findings.

This study has several limitations. First, the attrition rate was 22.6%, and there were significant differences between the retained participants and dropouts. Specifically, those dropouts reported higher frequencies of parental gaming and invitations for cogaming. Furthermore, the MCAR test confirmed that the data were not missing completely at random. To reduce this selection bias and improve the validity of the estimates, multiple imputation was used in data analysis to account for these missing data patterns [47]. Second, convenience sampling may introduce sampling bias [48]. As there might be geographic variations, the generalizability of these findings to other regions and countries should be made with caution. Third, as the questionnaire was self-administered, social desirability bias might be present [48], and several measures were taken to mitigate this risk. The survey was anonymous, and no schoolteachers were present in the classroom during administration; data collection was managed entirely by external research staff to encourage honest reporting. Another notable limitation is the single-item measures to assess parental supportive attitude and behavioral intention of increasing gaming time. While single-item measures can reduce participant burden and are considered valid for concrete constructs [49], they preclude the estimation of internal consistency reliability. As path analysis assumes that observed variables are measured without error [50], the presence of potential random measurement errors in these single items may underestimate path coefficients [51]. Future studies should verify these findings using validated multi-item scales. Furthermore, IGD was assessed by the 9-item DSM-5 IGD checklist instead of the ICD-11 gaming disorder criteria, as the former was the prevailing diagnostic standard for large-scale epidemiological studies at the time of baseline data collection (December 2018). Although the clinical validity of the DSM-5 IGD checklist has been validated in Chinese populations [34], this self-report screening tool may overestimate IGD prevalence compared with a clinical gold standard administered by a psychiatrist [52]. Furthermore, this prospective study used a 2-wave design, and both mediators and IGD conversion were assessed concurrently at T2. Although the baseline variables were adjusted, this design precludes the establishment of a complete temporal relationship (ie, predictor to mediator to outcome). As such, the specific causal relationships between the mediators and IGD conversion cannot be statistically confirmed and should be interpreted as associations within a prospective framework rather than causal mechanisms. Finally, this study did not include data on gaming time or frequency, which are known predictors of IGD conversion [43]; this limits the ability to assess whether the observed associations exist independently of gaming time. Furthermore, this study did not assess peer factors, which are important determinants of adolescent gaming [17]. This limits the ability to compare the relative strength of parental versus peer influences. Future research should integrate peer variables to provide a more comprehensive ecological model of IGD conversion.

In conclusion, this longitudinal study is novel to identify the prospective associations between specific parental gaming behaviors and the risk of IGD conversion among Chinese adolescents. Grounded in SCT and TPB, these findings elucidate the mediation mechanisms involving family environment and individual cognitive processes. Specifically, the identification of parental supportive attitude toward the adolescent’s gaming behavior and adolescents’ behavioral intention of increasing gaming time as key mediators reveals a nuanced parental behavior–adolescent cognition pathway, suggesting that parental modeling not only provides a behavioral template but also shapes the internal cognitive appraisals and behavioral intentions that precede IGD conversion.

These results bear theoretical and practical implications. Theoretically, they extend the application of SCT and TPB to the field of behavioral addiction by highlighting how distal environmental factors (parental behavior) are internalized through proximal cognitive factors. Practically, the findings underscore the necessity of shifting the focus of IGD prevention from solely adolescent-centered approaches to family-based interventions. Prevention strategies should involve educating parents on their own gaming habits and fostering healthier cognitive appraisals of gaming within the household. Moreover, mitigating the risk of adolescent IGD conversion requires a holistic approach that targets both modifiable parental influences and the subsequent cognitive distortions in adolescents.