We recruited all individuals online through convenience sampling in June 2021. The inclusion criteria were as follows: (1) age ≥13 years, (2) Chinese speakers who could understand the questionnaires, (3) consent to participate (adolescents with parental consent), and (4) played games at least one hour every week (excluding online gambling) in the last 12 months.

In total, 996 individuals participated in this survey. We excluded 88 individuals whose answer was “No” to the item “Are your answers to this questionnaire true and reliable?”, 37 who gave the same answers to more than 50% of the questions and whose time spent on the questionnaire was less than the mean minus 3 SD, and 99 who were younger than 13 years old. The final sample included 772 participants.

Before starting the anonymous online investigation, participants were informed about the purpose and rights of the study and signed an online informed consent form. Those younger than 18 years needed to inform their guardians and obtain consent before filling out the questionnaire. The ethics committee of the Second Xiangya Hospital of Central South University approved this study (protocol code 2020004; dated March 1, 2020).

The MOGS was translated into Chinese by 2 professional translators using a standard translation and back-translation method [66]. For some controversial items (eg, “Online gaming makes me lose control,” “Online gaming makes my worries more bearable”), we consulted the author of the original scale. Considering the original scale, 2 bilingual psychologists revised the translated version and checked its face validity. A pilot study was conducted with 5 adults and 5 adolescents to test the understandability. Based on their feedback, some descriptions of items were modified, and the final Chinese version of the MOGS (C-MOGS) was created.

We conducted this survey online using Questionnaire Star, a professional online survey platform. By reading recruitment advertisements posted on social networking sites (eg, WeChat, Weibo, and other webcast platforms), individuals could open the questionnaire link. On the first page, participants could read the objectives and content of this research and confirm their participation (minors, those younger than 18 years, had to obtain the consent of their guardian). Each IP address can only be used once to avoid repeated participation. After the questionnaire was submitted, all the data were sent to the researcher's account, and only the researcher could view the data.

Data analyses were conducted using SPSS version 25.0 (IBM Corp) and AMOS version 24.0 (IBM Corp). First, basic statistical analyses (eg, descriptive analysis, independent samples t tests, chi-square tests) were performed on sociodemographic variables and Internet gaming characteristics. To analyze the construct of the C-MOGS, the total sample was randomly split into 2 subsamples. Sample-1 (n1=390; 229/390, 58.7% male; age: mean 22.25, SD 9.05 years) was used for EFA, and sample-2 (n2=382; 229/382, 59.9% male; age: mean 21.14, SD 8.52 years) was used for CFA. Except for EFA and CFA, all analyses were conducted on data from the entire sample. Independent samples t tests and chi-square tests showed that there were no significant differences between the 2 subsamples regarding age (t₇₇₀=–1.754, P=.08), gender (χ²₁=0.121, P=.73), average weekly gaming time (t₇₇₀=0.775, P=.44), gaming devices (χ²₁<1.012, P=.31), long-term gaming partner (χ²₃=4.046, P=.26), and other demographic variables (P=.33-.88). An EFA with principal component analysis (PCA) and varimax-rotation method was conducted on the C-MOGS items. To validate the models derived from the EFA, a CFA was completed with AMOS 24.0 using the maximum likelihood method. The model fit was appraised using multiple fit indexes, including chi-squared:degree of freedom ratio (χ²/df<5), goodness-of-fit index (GFI>0.90), Tucker-Lewis Index (TLI>0.90), comparative fit index (CFI>0.90), standardized root of the mean square residual (SRMR<0.08), root mean square error of approximation (<0.05=close fit; <0.08=acceptable fit; <0.1=mediocre fit) [67]. The reliability of the C-MOGS was examined by assessing the internal consistency of the scale and subscale. Acceptable values for the Cronbach α and Guttman split-half coefficients are >0.70, while values >0.80 are considered good [68]. Finally, to test the concurrent and incremental validity, correlation analyses and hierarchical multiple regression analyses were conducted between the C-MOGS and Internet gaming characteristics (gaming time, IGD, gaming motives) as well as anxiety and depression.

In this study, analysis was conducted on data from 772 participants (458 men, 59.3%) aged between 13 years and 57 years (age: mean 21.70, SD 8.81 years; participants aged 13-17 years: 281/772, 36.4%). The majority of them were students (555/772, 71.9%). Smartphones were the most popular device for gaming (705/772, 91.3%). Of the sample, 69.4% (536/772) had one or more long-term gaming partners. Participants spent an average of 13.43 (SD 10.88) hours every week playing games. More details on the sample characteristics are shown in Table 1.

The Cronbach α coefficient and split-half reliability were calculated for the scale and its subscales in the total sample (n=772). The α coefficient for the total scale was .894, and it was .823 for the N-MOGU, .799 for the N-MOGD, and .925 for the P-MOG. No item deletion improved the internal consistency. The Guttman split-half coefficient of the overall scale was 0.942, and for each dimension, the coefficients were 0.776, 0.754, and 0.841. These findings confirmed that the C-MOGS and its subscales exhibit adequate internal consistency.

Moreover, we calculated the correlation coefficient between each item and its relative factor scores. The results showed that the item-total correlations for all items were high (r≥0.551).

We further analyzed the correlation between the 3 factors of the C-MOGS and IGD, gaming motives, anxiety, and depression to test the concurrent validity. Based on the Shapiro-Wilk test, these variables did not follow a normal distribution (all Ps<.05). Therefore, Spearman correlation analysis was chosen to explore the relationships between the variables. Table 4 shows the descriptive statistics (median and IQR), and Table 5 shows the correlations between the variables. Each factor of the C-MOGS showed positive correlations with the IGDS9-SF, weekly gaming hours, every dimension of the MOGQ, the PHQ-9, and the GAD-7 (r=0.153 to 0.759, all Ps<.01). Moreover, the correlation matrix showed positive correlations between the IGDS9-SF and the other variables (r=0.352 to 0.700, all Ps<.01).

We conducted a hierarchical multiple linear regression analysis to identify the incremental effect of metacognitions about online gaming on IGD. The IGDS9-SF was the outcome variable, and the 3 factors of the C-MOGS were predictor variables, along with other variables related to the IGDS9-SF (gender, weekly gaming hours, the 7 factors of the MOGQ, and the total PHQ-9 and GAD-7 scores). Each variable was input in the following order: step 1: age and gender (0=female, 1=male); step 2: weekly gaming hours; step 3: the 7 factors of the MOGQ; step 4: GAD-7, PHQ-9; step 5: the 3 dimensions of the C-MOGS.

The Durbin-Watson statistic showed that the observed values were independent of each other (D-W=2.077). All tolerance values were above 0.1 (0.180-0.878), indicating no multicollinearity. The results are presented in Table 6. The 3 factors of the C-MOGS accounted for 13.0% of the variance in the IGDS9-SF (P<.001). In step 5, the final model indicated that gender, weekly gaming hours, the PHQ-9 score, the MOGQ-Escape score, the MOGQ-Competition score, and the factors of the C-MOGS were significant positive predictors of the IGDS9-SF (R²=0.729, P<.001, adjusted R²=0.724), and the most important predictor was the N-MOGU (β=0.326, P<.001).

To investigate the psychometric properties of the Chinese MOGS and its association with IGD, this study translated and tested it in China for the first time. In general, the results suggested that the C-MOGS could potentially serve as a valid and reliable tool to assess specific metacognitions about online gaming and it may have the capacity to predict IGD independently.

First, factor analyses were used to explore the structural validity of the scale. The EFA suggested a 2-factor solution (N-MOG and P-MOG), which was consistent with the first assumption of the original scale [13]. By setting 3 factors to be extracted, the EFA also obtained the same 3-factor solution as the final version of the original scale (N-MOGU, N-MOGD, and P-MOG) [13,54]. Through CFA, the 3-factor model was later proved to have the best data fit. Moreover, we attempted to build a bifactor model that included a first-order factor (P-MOG) and a second-order factor (N-MOG: N-MOGU and N-MOGD). This model had the same goodness of model fit as the 3-factor structure. To maintain consistency with the original scale, the 3-factor structure is recommended for measuring specific online gaming metacognitions in the Chinese population. For studies that compare N-MOG and P-MOG, the bifactor model can be considered.

The 3-factor structure of the C-MOGS demonstrated adequate internal consistency, with Cronbach α coefficients ranging from .799 to .925 for each factor and the full scale, along with Guttman split-half coefficients ranging from 0.754 to 0.942. The current findings also provide evidence for the concurrent and incremental validity of the C-MOGS. Each subscale was significantly positively correlated with IGD, weekly gaming hours, gaming motives, depression, and anxiety. Moreover, the C-MOGS accounted for 13.0% of the variance in IGD while controlling for other variables. These findings highlight the utility of the C-MOGS as a reliable and valid tool to assess metacognitions about online gaming among the Chinese population.

Furthermore, this study explored the effects of metacognitions about online gaming, gaming motives, anxiety, and depression on IGD using hierarchical multiple linear regression analysis. After adding metacognitions about online gaming to the regression equations, the final model accounted for 72.9% of the variance in IGD. In addition to metacognitions, gender, weekly gaming hours, escapism motives, competition motives, and depression significantly predicted IGD, suggesting that these factors collectively contribute to the development and maintenance of IGD [69-71]. Importantly, the inclusion of metacognitions led to a reduction in the standardized regression coefficients of these variables, and the predictive effect of anxiety on IGD became nonsignificant. This indicates that metacognition may partially mediate or explain the impact of these factors on IGD. This finding is consistent with previous research, suggesting that metacognitions about online gaming may mediate the influence of other psychological factors, such as psychological dependence, anxiety, and depression, on IGD [55,72-75]. These results indicate that specific metacognitions about online gaming are important predictors of IGD, which is consistent with previous studies [13,54,55]. However, the mechanisms underlying the role of metacognitions in IGD seem to be interrelated with other psychological factors, which remains inconclusive.

In a hypothesized model, metacognitions about online gaming may promote problematic gaming engagement by increasing gaming time and disrupting normal emotion and cognition [76]. Consistent with this view, our study found that people with more metacognitions about online gaming would spend more time playing games and feel more anxious and depressed. P-MOG increases gaming time by promoting online gaming as a self-regulation method for emotion and cognition [13,21]. N-MOGU will maintain problematic gaming engagement by destroying one’s confidence in self-control, while N-MOGD can induce negative reinforcement and compulsive gaming engagement by triggering negative emotions such as anxiety and depression [21,76]. Furthermore, gaming motives may be an intermediate factor, as our study found: Gaming motives were simultaneously significantly correlated with MOGS and IGD. Dysfunctional metacognition activates maladaptive coping strategies and motivation, which causes negative emotions to persist and eventually leads to IGD [77]. Moreover, other studies have different views. For example, metacognitions have a mediating effect on the association between emotional dysregulation and problematic Internet use [78], and online gaming thought suppression and impulsiveness mediate the relationship between metacognition and IGD [79]. Therefore, the association between metacognition and IGD cannot be summarized by simple causality. Other psychological variables, such as motives, coping style, impulsiveness, and emotional regulation, should be considered in future research.

Since maladaptive metacognitions are an important predictor of IGD, interventions specifically addressing maladaptive metacognitions, such as metacognition therapy (MCT), may be beneficial for the prevention and treatment of IGD. MCT, an intervention aimed at modifying dysfunctional metacognition, is effective for treating psychiatric and psychological diseases such as anxiety, depression, and schizophrenia [80-84]. Although MCT is not widely used in the treatment of addictive behaviors, researchers are attempting to prove its efficacy [11]. In some pilot studies, MCT was used to effectively treat alcohol abuse and gambling disorder [85,86]. However, the specific efficacy of MCT for treating IGD needs to be further verified in clinical research. This study provides evidence for the potential value of MCT in the clinical treatment of IGD and offers an effective tool for conducting MCT for IGD specifically in the Chinese population.

Although this study has the advantages of a large sample size with people of different ages, it has several limitations that should be considered. First, this study adopted convenience sampling instead of random sampling, and only gamers were included. Therefore, it does not sufficiently represent all Chinese people. Second, collecting data using an online self-report questionnaire may increase the probability of participants giving false answers. However, this procedure is reported to be as reliable as pencil-and-paper surveys [87], which is likely to reduce social desirability and increase levels of honesty [88]. Third, this study lacked test-retest reliability of the C-MOGS; further research is required to test its stability. Finally, as a cross-sectional study, we could not infer the causality of the studied variables. Thus, longitudinal research is needed to further explore the relationship between metacognition and IGD.

In summary, this study offers some evidence that supports the satisfactory psychometric properties of the C-MOGS and highlights the possibility of metacognition as an independent risk factor in gaming behavior. It may be a useful and prospective tool for exploring psychological mechanisms of IGD and helping health professionals identify risky gamers (eg, individuals with more metacognitions about online gaming, specifically negative metacognitions about the uncontrollability of online gaming). Additionally, MCT may be beneficial for the prevention and treatment of IGD. This study may support more attention for metacognitive beliefs in addictive behaviors.