Maladaptive Cognitions in Adolescents and Young Adults When They Play: The Dysfunctional Cognitions in Gaming Scale (DCG)

Abstract

Gaming is increasingly prevalent among young people, and Gaming Disorders are a growing concern. Maladaptive cognitions related to gaming may affect the psychological development of young people. We examined psychometric properties of the Dysfunctional Cognitions Gaming (DCG) Scale in Spanish adolescents and young adults. We applied 16 items of the DCG Scale in a sample of 2173 video gamers (age from 12 to 22; 28.8% female), extracted from random sampling in educational institutions. Three factors emerged from exploratory analysis (EFA): Preoccupation, Self-esteem, and Compulsion, accounting for 51.92% of the scale’s total variance. Confirmatory analysis (CFA) yielded a good fit, RMSEA = 0.040, 90% CI [0.034, 0.046]. However, several results (factor cross-loadings in EFA, a high eigenvalue for the first factor in parallel analysis, high correlation between latent factors in CFA, and high hierarchical omega and explained common variance—ECV—in a bifactor model) suggested the convenience of using the total score for evaluation and other applied purposes. The scale showed adequate reliability, ω = 0.908, 95% CI [0.900, 0.914], R_xx = 0.91. No gender or age invariance was found. The scale converged with the severity of internet gaming disorder scores, r = 0.697, p < 0.05; higher frequency of video gaming, F(1, 2165) = 474.9, p < 0.001, η² = 0.176; poorer mental health, r = 0.20, p < 0.05. We provided percentile ranks separated by gender and age. The DCG Scale seems to be a reliable and valid instrument to identify maladaptive cognitions about gaming in Spanish youth. These cognitions are a health-related problem; identifying and addressing them would be desirable to promote positive youth development.

1. Introduction

Gaming is an increasingly popular activity among adolescents and young people, becoming one of their main leisure activities. In Spain, in 2020, the sector generated 1747 million euros with 15.9 million estimated gamers, showing increases similar to worldwide figures [1]. Problematic gaming, in adolescents and young people, characterized by (a) a pattern of obsessive and uncontrolled video game play, (b) interference with the daily life of the person, (c) causing problems that affect both their personal life and their academic and professional performance [2], has become cause of social alarm. The importance of problematic gaming is reflected in the incorporation of the diagnoses of Internet Gaming Disorder (IGD) in the DSM-5 [3], and Gaming Disorder in the ICD-11 [4]. Among the criteria for these diagnoses are cognitive aspects, common in addictions with and without substances, including concern for gambling. However, such cognitive aspects have received less attention than in other forms of addiction.

The starting point of the recognition of these cognitions is Davis’s Problematic Internet Use model [5]. According to this, the positive reinforcement resulting from the use of the internet and the presence of psychopathological problems, such as depression, could lead to the development of maladaptive cognitions, which include thoughts about the self and thoughts about the world. These cognitions, together with social isolation, could lead to a pathological use of the internet (including problematic gaming). This pathological use could facilitate negative symptoms such as obsessive thoughts about its use, guilt about its use, or the inability to stop using it, which would feedback these maladaptive cognitions.

Caplan developed Davis’ model [6], including a preference for online social interactions, caused by psychopathological problems (depression, anxiety, and loneliness), as a trigger for problematic internet use (and problematic gaming), and the use of the internet as a mood regulation strategy. Furthermore, he identified maladaptive cognitions related to interpersonal communication skills.

In a systematic review of 36 studies, King and Delfabbro [7] sought to identify the common cognitions present in IGD. Based on this systematic review, the authors developed the Internet Gaming Cognition Scale (IGCS) [8], encompassing the maladaptive ones in four categories: (a) Beliefs about the reinforcements of the game and its tangibility. (b) Maladaptive and inflexible rules about behavior in video games. (c) Use of video games as a source of self-esteem. (d) Use of video games to gain social acceptance. King and Delfabbro [8] found that, in adolescents with maladaptive gaming beliefs, these were presented together with the criteria for IGD, but they did not appear in adolescents without symptoms of IGD, even though they spent a lot of time playing. Moudiab and Spada [9] also found a relationship between maladaptive cognitions and the presence of IGD. The best predictors of IGD were the overvaluation of the rewards of the game, and the motivation to play as a coping strategy. Bodi et al. [10] found these cognitions to appear in all types of video games, although to a greater degree when playing online.

Forrest et al. [11] looked at players aged 16–65 and found a direct relationship between the presence of maladaptive beliefs and problematic gaming. They identified four factors: (1) Perfectionism, the need to be better than other players, or not being able to stop playing until goals are achieved. (2) Cognitive salience, the inability to focus on other activities due to ruminative thoughts about video games. (3) Regret, the presence of feelings of guilt about the frequency of gambling and its negative consequences. (4) Behavioral salience, the difficulty of quitting video games due to the high amount of time spent playing. People with problematic gaming had significantly higher scores on perfectionism, cognitive salience, and regret. In their longitudinal study, Forrest et al. [12] found that the presence of these maladaptive cognitions could also predict changes in gaming behavior.

Although King and Delfabbro led the study on cognitive factors in IGD with their categories [7,8], they pointed out the need to specify the cognitive aspects (types and characteristics) associated with problematic gaming and its effects on it. However, taking into consideration the Chinese and French validations of the IGCS, a clear factorial structure of maladaptive cognitions related to video games is missing [10,13]. Yu et al. [13], in the Chinese version, proposed a new three-factor model (perceived rewards of internet gaming, perceived urges for playing internet games, and perceived unwillingness to stop playing without completion of gaming tasks). Bodi et al. [10], in the French version, identified five factors (positive emotions, need of completion, cognitive salience, virtual comfort, and need of social recognition). This lack of a clear factor structure would not be, by itself, enough to justify the development of a new instrument. In fact, discrepancies in factor structure are common in psychometric measures across cultures and languages. However, King and Delfabbro found that the adolescents with IGD had distinct problematic cognitions about gaming than those without IDG, with a large size of observed effects [8]. This association between gaming cognitions and IGD symptoms could be explained because the scale items were, precisely, developed from the common cognitions found in primary studies on IGD [7] (i.e., subjects with gaming problems). We think that the maladaptive cognitions do appear with some degree of frequency also in subjects without IGD. Given that they constitute the vast majority of the population, it would be useful to have a new instrument to detect problematic cognitions in this general population, and to create sorting scales to discriminate adequately these subjects. In this direction, it would be of great help to have a new instrument (a) to assess cognitions associated with problematic gaming or the risk of developing it, (b) validated in young people and adolescents, (c) with a clear factorial structure, and (d) developed from a situational framework common to all gamers (with or without IGD). This instrument could be administered independently or alongside IGD measures, to specifically assess maladaptive cognitions.

Therefore, the objective of this study was to develop brief dysfunctional cognitions in the gaming scale and find evidence for its reliability, validity, and factorial structure, in a large random sample of Spanish adolescents and youth. We aimed to assess the psychometric properties of the DCG Scale, including reliability (internal consistency and split-half method), construct validity (exploratory and confirmatory factor analysis, and measurement invariance across gender and age groups), and criterion validity (with theoretically related variables, such as IGD and frequency of gaming). Additionally, we provided percentile ranks for sorting and measurement purposes.

2. Materials and Methods

2.1. Participants

The initial sample was composed of 2887 students (both male and female), from a representative sample of educational institutions of Madrid (Spain). A final number of 65 classes of 41 different schools were assessed, which provided a total N of 2887 students, 1637 over the 1250 minimum goal [14]. This increase was evenly distributed across the sample so no district, school year, or type of school was overrepresented. The final sample comprised only video game players (those participants who had played video games in the previous twelve months), amounting to 2173 participants (75.3% of the initial sample), from the equivalent to 7th to 11th grade, and Basic and Advanced Professional Training in the US system. Thus, we considered two age groups: (a) 12- to 16-years-old (up to 11th grade), M = 13.82 (SD = 1.34), n = 1516; of them, 491 (32.4%) were female; (b) 17- to 22-years-old (Basic and Advanced Professional Training), M = 18.82 (SD = 1.34), n = 657; of them, 135 (20.5%) were female. The final sample comprised only video game players, amounting to 2173 participants, of whom 65 participants (3% of the gamers) were labeled as having an IGD with the endorsement of at least five of the criteria of IGD, by answering “very often” (the highest score) in at least five items of the IGDS9-SF.

2.2. Instruments

2.2.1. Dysfunctional Cognitions in Gaming Scale (DCG Scale)

This scale assesses gamers’ cognitions in four different situations: (a) What they think when they are doing another activity (e.g., “I wish I were playing right now”). (b) What they think when they make a mistake (e.g., “This game or these rivals can’t handle me, I won’t stop until I win”). (c) What they think when they are playing well (e.g., “I’m really good and I feel good”). (d) What they think when their parents or someone around them asks them to interrupt or stop playing (e.g., “I think that people who don’t play video games don’t understand me”). Five experts developed the DCG Scale, a 16-item scale. Some of the items were adapted from the proposals by King and Delfabbro [8] and Forrest et al. [11], as they fitted one of the four situations. Other new items were created directly from the proposed scenarios. A pilot study was carried out with a convenience sample of 93 players, based on whose comments the final scale was developed.

The DCG was conceived as a brief, screening measure. It included 16 self-reported items (four for each situation), evaluated with a 5-point Likert scale (with five levels: 0 “Never” to 4 “Always”), ranging from 0 to 64 points (a greater score indicates more problematic cognitions toward video games). The items present in the scale can be seen in

Table 1. DCG Scale items. Descriptive statistics.

#	Item	M	SD	Mdn	IQR	Sk	K
1	I find myself thinking about video games when I’m not playing [Me encuentro pensando en videojuegos cuando no estoy jugando]	1.0	0.9	1	2	0.8	0.3
2	I say to myself “I wish I was playing right now” [Me digo “ojalá en este momento estuviera jugando”]	1.0	1.0	1	2	0.9	0.3
3	I couldn’t stand to wait much longer without playing [No soportaría esperar mucho más tiempo sin jugar]	0.6	0.8	0	1	1.8	3.5
4	I plan or think about what I will do next in the game to advance [Planeo o pienso lo siguiente que haré en el juego para avanzar]	1.3	1.2	1	2	0.6	−0.7
5	I won’t stop until I get past it. Better now than tomorrow [No pararé hasta conseguir pasármelo. Mejor ahora que mañana]	1.3	1.2	1	2	0.7	−0.4
6	This game, or these rivals, are not going to be able to beat me, I am better [Este juego o estos rivales no van a poder conmigo, yo soy mejor]	1.5	1.3	1	2	0.5	−0.8
7	Stopping playing now will be a failure [Dejar de jugar ahora será un fracaso]	0.8	1.0	0	1	1.4	1.6
8	I cannot fail. Other players admire and respect my achievements in the game [No puedo fallar. Otros jugadores admiran y respetan mis logros en el juego]	0.8	1.1	0	1	1.3	0.8
9	I am very good and I feel good [Soy muy bueno y me siento bien]	2.1	1.2	2	2	0.0	−0.8
10	I am better than the others [Soy mejor que otros]	1.4	1.3	1	2	0.6	−0.7
11	I don’t want to know anything that happens around me, it’s secondary [No quiero saber nada de lo que ocurre a mi alrededor, es secundario]	0.7	1.0	0	1	1.4	1.7
12	It’s great to be able to spend time enjoying with these people [Es genial poder disfrutar de pasar tiempo con toda esta gente]	2.2	1.2	2	2	−0.2	−0.8
13	I think people who don’t play video games don’t understand me [Pienso que las personas que no juegan a videojuegos no me entienden]	0.9	1.1	1	1	1.2	0.6
14	I tell myself “just a few more minutes”, but then it’s much longer [Me digo a mí mismo “solo unos minutos más” pero luego es mucho más tiempo]	1.5	1.2	1	1	0.5	−0.7
15	I don’t care at all how they react, I’m not going to stop [Me da exactamente igual como se pongan no voy a parar]	0.4	0.8	0	1	2.2	4.7
16	If I stop playing I will feel bad [Si dejo de jugar voy a sentirme mal]	0.4	0.7	0	0	2.6	7.3

Notes. n = 2173. Minimum = 0, Maximum = 4. Sk = Skewness; K = Kurtosis. Spanish version in brackets [].

.

2.2.2. Internet Gaming Disorder Scale—Short-Form, IGDS9-SF

This is a Spanish validated translation [15] of the original scale [16]. It is based on the DSM-5 [3] criteria for IGD, consisting of nine 5-point Likert items (from 1 “Never” to 5 “Very Often”). It assesses the severity of IGD and the negative consequences of online and offline video gaming in the last 12 months. In several subsamples, it showed adequate internal consistency (ω from 0.778 to 0.828), split-half reliability (R_xx from 0.770 to 0.822), and a single-factor structure. The total score is the sum of the items scores; the higher the score, the higher the severity of IGD.

2.2.3. General Health Questionnaire, GHQ-12

This is a Spanish validation [17] of the original GHQ-12 [18]. This 12-item questionnaire assesses perceived health, with seven items composing the anxiety subscale (α = 0.776) and five items composing the dysfunction subscale (α = 0.771), and the total score (α = 0.569). Each item is a 4-point Likert scale. A higher score indicates poorer mental health.

2.2.4. Frequency of Gameplay

Participants separately reported their average days per week invested in VGs, and average hours per week (from “less than 1 h” up to “more than 30 h” playing VGs, increasing in orders of five hours; 1 to 5 h, 6 to 10, until reaching 30, with the further option of more than 30 h).

2.3. Procedure

Five evaluators with psychology degrees independently administered the Gamertest [19], an online evaluation tool that includes all the measurement instruments used in this study. A list of schools from the city of Madrid [20] was segmented by each of the 21 city districts, and type (public, private, or state subsidized school). For each segment of district and type of schooling, a random school was sent a letter and a follow-up call. School authorities were asked to provide access to the required number of participants. For those schools that refused (n = 95), another random school from that segment was contacted. Upon agreement, informed consent forms were delivered for the children’s guardians to sign. Only 1.3% of the participants did not return the informed consent form. The evaluators administered the assessments to all students in each selected class, which were chosen via stratified random sampling. The students took 30–40 min to complete the online anonymous assessments, in each school’s computer room. The responses were automatically coded in a database. The ethics committee of the University’s Faculty of Psychology audited the ethical issues for this study.

2.4. Data Analysis

Statistical analyses were carried out using R and several R packages: MVN [21] for Mardia’s multivariate normality analysis; lavaan, version 0.6-3 [22] for confirmatory factor analysis (CFA); MBESS, version 4.4.3 [23] for α and ω estimators and their CI; psych [24] for Bartlett’s sphericity test, KMO estimator, parallel analysis, split-half reliability, and exploratory factor analysis (EFA); BifactorIndicesCalculator [25] for the bifactor model statistics; semTools [26] for the invariance analysis.

2.4.1. Item Analysis

We calculated several descriptive statistics for the items and the total scores of the scale and subscales. As the variables have five ordered categories, we considered them as continuous for the purposes of all analyses.

2.4.2. Factorial Validity

We divided the final sample into two random subsamples of video game players, to study the factor validity of the DCG Scale.

The first subsample (n = 1130) was used to carry out an EFA, after assessing its adequacy in our dataset using Bartlett’s sphericity test and the KMO estimate.

As the main criterion for determining the number of factors to retain from EFA, we conducted a parallel analysis based on principal component analysis [27] on the final sample. We assessed the multivariate normality of the items of the scale via Mardia’s multivariate kurtosis and skewness coefficients [28]. As we rejected the hypothesis of multivariate normality, we selected an ordinary least-squares, for minimum residual solution (minres), as the extraction method. To improve factor simplicity, we used a Promax rotation method.

Next, we used the second subsample (n = 1143) to confirm the subjacent structure, via confirmatory factor analysis (CFA). We selected a diagonally weighted least squares (DWLS) estimation method, with a Pearson’s correlation matrix as input. To assess model fit, we used several indices, comparing them with the recommended values (RVs) [29,30]: chi-square-to-degrees-of-freedom ratio (χ²/df) (RV ≤ 3.000), root-mean-square error of approximation (RMSEA and its 90% confidence interval; RV < 0.060 to 0.080), standard root-mean-square residual (SRMR) (RV ≤ 0.080), comparative fit index (CFI), and Tucker–Lewis index (TLI) (both RV ≥ 0.950). The magnitude, direction, and statistical significance of the standardized parameter estimates were interpreted, as recommended by Brown [31].

Finally, we fitted a CFA bifactor model to assess the appropriateness of using factor scores or, on the contrary, using only the total DCG Scale score. We reported the explained common variance, ECV (i.e., the ratio of variance explained by the general factor divided by the variance explained by the general plus the group factors). The higher the ECV, the stronger the influence of the general factor and the greater the confidence in using a unidimensional measurement model, even when there are several underlying factors. In addition, we reported the coefficient omega hierarchical to assess the degree to which the factorial structure is interpretable as a single factor measure [32,33].

2.4.3. Measurement Invariance

We tested for group measurement invariance using multiple group confirmatory factor analysis (MGCFA), in order to examine the generalizability of the scale across age (12- to 16- and 17- to 22-years-old) and gender (male and female participants). For each variable, we compared three nested multiple group models to test for configural invariance (same number of factors), metric invariance (same factor loadings), and scalar invariance (same indicator intercepts). As fit indices for configural invariance, we interpreted RMSEA. As fit indices for metric and scalar invariance, we tested for χ² difference (Δχ²), CFI difference (ΔCFI), and RMSEA difference (ΔRMSEA). We would consider a value of ΔCFI < −0.010 supplemented by a ΔRMSEA < 0.015 as an indicator of invariance [34].

2.4.4. Reliability

We used two estimators for internal consistency, with 95% CI: Cronbach’s alpha (α) and, to address the problems described by McDonald [35], omega (ω). We also assessed reliability via the Spearman–Brown split-half method.

2.4.5. Other Evidence of Validity

We explored a linear trend between DCG Scale scores (total and subscales) with average hours spent playing video games using ANOVA polynomial tests, and the relationship between DCG Scale scores and weekly frequency using correlation tests. Then, we ran correlation tests between DCG Scale scores with GHQ-12 (total and subscales) scores and the IGDS9-SF scores.

2.5. Transparency and Openness

We adhered to the JARS statement for quantitative research [36].

3. Results

3.1. Item Analysis

Table 1 displays the descriptive statistics of the items. The items showed different degrees of negative and positive kurtosis; also, 14 out of 16 items showed positive skewness. The estimates of Mardia’s multivariate kurtosis and skewness coefficients were high, 10,760.99 and 101.90 respectively, and significant (both ps < 0.001).

3.2. EFA

The recommended number of dimensions, considering the mean of 100 random correlation matrices in a parallel analysis, was three (the scree plot and parallel analysis are shown in Figure 1). As the presence of a third factor was somewhat marginal, several factoring methods were used (maximum likelihood, minimal residual, unweighted least squares, and minimum rank factor analysis), combining them with Pearson’s correlation as input matrices, and using the mean and 95th percentile criteria. The results were all the same. In addition, when deciding how many factors to extract, we took into account theoretical criteria. However, the large difference between the eigenvalues of the first and second factors also suggests that the existence of a single factor should also be considered and discussed. The KMO test, 0.93, as well as Bartlett’s test of sphericity, χ²(136) = 15,124.26, p < 0.001, showed data adequacy for EFA.

Table 2. DCG Scale. Results from exploratory factor analysis.

	Factor Loading
Item	Preoccupation	Self-Esteem	Compulsion
1	0.801	0.461	0.585
2	0.808	0.467	0.569
3	0.706	0.387	0.649
4	0.748	0.484	0.490
5	0.568	0.616	0.589
6	0.421	0.788	0.498
7	0.520	0.571	0.661
8	0.505	0.644	0.625
9	0.425	0.759	0.399
10	0.365	0.688	0.446
11	0.421	0.515	0.647
12	0.562	0.589	0.399
13	0.569	0.509	0.675
14	0.562	0.434	0.579
15	0.456	0.415	0.727
16	0.483	0.348	0.686

Notes: n = 1130. Minimum residual solution (minres) extraction method. Explained variance 51.92%. The assignment of items to each factor according to their factor loadings is shown in bold type.

shows the factor loadings of each item after the factor extraction, ranging from 0.706 to 0.808 (Factor 1), 0.589 to 0.788 (Factor 2), and 0.579 to 0.727 (Factor 3). The correlations among factors ranged from 0.61 to 0.70. These three factors were labeled as Preoccupation, Self-esteem, and Compulsion, according to the analysis of the content of the items, accounting for 17.22%, 16.58%, and 18.12% of the total variance, respectively. This three-factor model accounted for 51.92% of the total variance. We observed cross-loadings in most of the items. In these cases, the decision to assign an item to a specific factor was made following a theoretical criterion. These results also suggest the presence of a global factor that encompasses all items.

3.3. CFA

The CFA model showed the adequate fit based on all the fit indices, χ²/df = 2.684; RMSEA = 0.040, 90% CI [0.034, 0.046]; CFI = 0.986; TLI = 0.983; SRMR = 0.056. Furthermore, the model showed positive item factor loadings (p < 0.001), ranging from 0.772 to 0.882 for Preoccupation, 0.629 to 0.673 for Self-esteem, and 0.569 to 0.635 for Compulsion (see Figure 2). The correlation among latent variables ranged from 0.717 to 0.819. These high, significant correlations among factors suggest, again, the presence of a global factor.

The bifactor model also showed an adequate fit, χ²/df = 1.730; RMSEA = 0.026, 90% CI [0.019, 0.033]; CFI = 0.995; TLI = 0.993; SRMR = 0.042. The ECV of the bifactor CFA model was 0.693, sufficiently high to consider the general factor “maladaptive cognitions” as stronger than the specific factors. In addition, the omega hierarchical coefficient for the general factor was high, ω_H = 0.832, especially compared to those of the specific factors, Preoccupation, ω_H = 0.258; Self-esteem, ω_H = 0.234; Compulsion, ω_H = 0.142. Considering both indices, the DCG can be considered essentially unidimensional [32].

3.4. Measurement Invariance

The MGCFA did not yield a good fit across age group subsamples (

Table 3. DCG Scale. Measurement invariance, by age (12- to 16- and 17- to 22-years-old) and gender groups.

Grouping Variable	Invariance Model	χ2	df	Δχ2	p (Δχ2)	CFI	ΔCFI	RMSEA	ΔRMSEA
Age	Configural	1738.7	202	-	-	0.902	-	0.082	-
	Metric	1803.8	215	60.034	<0.001	0.899	−0.003	0.081	−0.001
	Scalar	1966.7	228	235.978	<0.001	0.889	−0.010	0.082	0.002
Gender	Configural	1762.2	202	-	-	0.894	-	0.082	-
	Metric	1827.0	215	48.534	<0.001	0.890	−0.003	0.080	−0.001
	Scalar	2008.0	228	220.819	<0.001	0.878	−0.012	0.082	0.002

Notes: χ² = Standard test statistic. Δχ²: Robust difference test. Robust estimators for CFI and RMSEA. Deltas are compared to previous model.

). While ΔCFI and ΔRMSEA suggested invariance, Δχ² was significant. Moreover, RMSEA at the configural level did not allow us to assume invariance. This was also the case for gender. Thus, we could not assume invariance across age and gender groups, and we computed distinct scoring scales for each sub-group.

3.5. Reliability

The internal consistency (α and ω) and split-half reliability statistics seemed adequate, for the total scale and the three factors (see

Table 4. DCG Scale. Internal consistency estimators, α and ω, and average split half reliability (Rxx).

	α [95% CI]	ω [95% CI]	Rxx
All 16 items	0.907 [0.899, 0.915]	0.908 [0.900, 0.914]	0.91
Preoccupation (4 items)	0.843 [0.828, 0.855]	0.849 [0.832, 0.859]	0.85
Self-esteem (6 items)	0.829 [0.815, 0.838]	0.832 [0.819, 0.845]	0.83
Compulsion (6 items)	0.799 [0.779, 0.818]	0.800 [0.784, 0.816]	0.81

Note: n = 2137.

).

3.6. Other Evidence of Validity

3.6.1. Frequency of Playing Video Games

We found a significant linear relationship between the scores of the DCG Scale and weekly frequency of VGs, for the total score, F(1, 2165) = 474.9, p < 0.001, η² = 0.176; preoccupation, F(1, 2165) = 464.6, p < 0.001, η² = 0.174; self-esteem, F(1, 2165) = 389.1, p < 0.001, η² = 0.149; compulsion, F(1, 2165) = 215.2, p < 0.001, η² = 0.090. Higher scores were associated with more hours of playing per week.

Participants played video games between 1 and 7 days a week (M = 3.41, SD = 2.01 Mdn = 3.00, IQR = 3.00). We found a positive relationship between DCG scores and days of gaming per week for the total scores, r = 0.402, r² = 0.162; preoccupation, r = 0.396, r² = 0.157; self-esteem, r = 0.365, r² = 0.133; compulsion, r = 0.293, r² = 0.086.

3.6.2. GHQ-12 and IGDS9-SF

The DCG (the total and factor scores) showed a significant, positive correlation with IGDS9-SF scores (r² from 0.286 to 0.486), and with GHQ-12 total scores (r² from 0.011 to 0.029) and anxiety scores (r² from 0.008 to 0.041). The dysfunction scores of the GHQ showed a significant, positive correlation with the DCG total (r² = 0.004), preoccupation scores (r² = 0.006), and self-esteem scores (r² = 0.011), but not with compulsion scores (see

Table 5. DCG Scale. Convergent validity with IGDS9-SF and GHQ-12.

	IGDS9-SF	GHQ Total	GHQ Anxiety	GHQ Dysfunction
DCG Total score	0.697	0.200	0.165	0.060
Preoccupation	0.658	0.125	0.078	0.076
Self-esteem	0.535	0.199	0.134	0.107
Compulsion	0.658	0.182	0.210	−0.036 ns

Note: ^ns: non-significant, p > 0.050.

).

3.7. Scale Scores and Percentile Ranks

The scores of the DCG Scale ranged from 0 to 64. As we did not assume invariance for gender or age groups, we computed percentile ranks (based on the observed scores) separated by these variables (

Table 6. DCG Scale. Observed total score and percentile rank, by gender and age.

	Male		Female			Male		Female
	12 to 16	17 to 22	12 to 16	17 to 22		12 to 16	17 to 22	12 to 16	17 to 22
	(n = 1025)	(n = 522)	(n = 491)	(n = 135)		(n = 1025)	(n = 522)	(n = 491)	(n = 135)
Score	Percentile		Percentile		Score	Percentile		Percentile
0	1	1	6	4	33	88	93	97	98
1	1	3	8	10	34	89	93	97	99
2	1	4	14	15	35	90	95	98	99
3	2	5	18	22	36	91	95	98	99
4	3	8	25	26	37	92	95	98	99
5	5	9	29	33	38	94	95	98	99
6	7	11	34	36	39	95	97	98	99
7	10	13	41	43	40	95	97	99	99
8	12	16	45	48	41	96	98	99	99
9	15	19	51	56	42	96	98	99	99
10	18	22	57	58	43	97	99	99	99
11	20	26	61	61	44	98	99	99	99
12	23	28	62	65	45	98	99	99	99
13	27	32	66	66	46	98	99	99	99
14	30	37	69	70	47	98	99	99	99
15	34	39	72	73	48	98	99	99	99
16	37	42	75	76	49	98	99	100	99
17	41	46	76	77	50	98	99	100	99
18	45	49	81	82	51	99	100	100	99
19	49	54	82	83	52	99	100	100	99
20	53	57	84	84	53	99	100	100	99
21	57	63	85	84	54	99	100	100	99
22	60	69	86	85	55	99	100	100	99
23	63	72	87	86	56	99	100	100	99
24	67	75	88	90	57	99	100	100	99
25	70	77	90	93	58	99	100	100	99
26	74	78	91	95	59	99	100	100	99
27	77	83	92	95	60	99	100	100	99
28	80	85	93	95	61	100	100	100	99
29	81	87	93	95	62	100	100	100	99
30	83	89	94	96	63	100	100	100	99
31	85	91	95	97	64	100	100	100	100
32	87	92	96	97

). For the same reason, the statistics of the complete scale appeared segmented by gender and age (

Table 7. DCG Scale total scores. Descriptive statistics, by gender and age.

Gender	Age	n	M	SD	Mdn	IQR	Min	Max	Sk	K
Male	12 to 16	1025	20.88	10.95	20	14	0	64	0.8	1.06
	17 to 22	522	18.71	10.07	19	13.8	0	61	0.55	0.59
Female	12 to 16	491	11.85	9.84	9	12	0	60	1.4	2.55
	17 to 22	135	11.34	9.75	9	12	0	64	1.71	5.17

Note: n = 2173. Sk = Skewness; K = Kurtosis.

).

4. Discussion

This empirical study supports, through some evidence of validity and reliability, a model of the Dysfunctional Cognitions in Gaming (DCG) Scale in a large, representative random sample of video gamers (from 12- to 22-years-old), from educational centers in Madrid (Spain). The researchers administered an online evaluation tool [19] in schools that consented to participate. The collected data were analyzed using R and several R packages. We found evidence of reliability (using Spearman–Brown split-half method, and α and ω for internal consistency) and factor and construct validity (via exploratory and confirmatory analysis). Other analyses were conducted to assess whether the substantive three-factor model could be considered essentially unidimensional [32,33]. The sample size exceeded the minimum recommendations for all the analyses we used in this study [14]. The results of this study can be generalized to 12- to 22-years-old students of Madrid, thanks to the random selection of educational institutions. Caution is advised if this scale is applied to other populations.

This scale aims to provide a contextualized evaluation of cognitions (i.e., what thoughts players present in video-game-related situations). From the 16 items of the DCG Scale, three underlying factors were identified in adolescent and young students: Preoccupation, Self-esteem, and Compulsion, which are similar to the Chinese version of IGCS [13]. These factors are interesting for research purposes, and further studies could delve deeper into the factor structure and, if necessary, develop a theoretical model. However, the results obtained from this scale in this sample of video gamers suggest the presence of a general factor, which can be used to adequately assess and discriminate maladaptive cognitions about gaming in adolescents and young adults.

4.1. Maladaptive Cognitions, Gaming Frequency, Perceived Health, and Problems Using VGs

Scores on the DCG Scale are higher among those who play more VGs, both in days per week and total hours per week of playing. We also found worse perceived mental health among those who scored higher on the scale. In addition, a positive and reasonably high relationship was found between the presence of dysfunctional cognitions and problematic gaming scores, in line with the literature [7,8,9,11,12,37]. Although the correlation of the total score of the DCG and the IGDS9-SF is highest (reinforcing the idea of a single-factor structure), it is also significant and with a similar effect size in each factor separately.

The Preoccupation factor, which could be equated to the factor Cognitive Salience of Forrest et al. [11], refers to the difficulty to focus on other activities not related to the use of VGs. These concerns about VGs could be working as ruminations that have been identified as predictors of problematic gaming [12]. This factor, together with that of Compulsion, may be key to detecting the increased severity of problematic gaming, with an increase in the intensity of these maladaptive cognitions with increasing problematic gaming. The Self-esteem factor showed a somewhat smaller effect size. This result agrees with King and Delfabbro [8], who also reported a smaller effect size in the factor that encompasses those maladaptive cognitions related to the use of video games as a way of achieving social acceptance. Having said that, we must remember that the three factors have a high correlation between them, so they do not discriminate well on a practical level. We have kept them to work on a theoretical model in future works, but the global factor allows us to better discriminate subjects in their maladaptive cognitions.

4.2. Implications for Evaluation and Intervention

Based on these results, some considerations for the evaluation and psychological intervention in problematic gaming can be pointed out. Regarding the evaluation, the reliability and validity of the DCG to identify maladaptive cognitions associated with problematic gaming have been supported. In the absence of a gold-standard to consider a specific value as the threshold for dysfunctional cognitions, we provided a percentile rank for the total score, which allows us to evaluate a subject’s score in comparison to their reference group. Separate percentile rank scales were calculated for male and female, and younger and older students, as we did not find gender or age invariance. The scale for the subgroup of women aged 17 to 22 years was created on the basis of a small number of cases, and should be used with caution. However, the need for separate percentile scales, indicated by the absence of invariance, is reasonable. In the youth population, behaviors—including cognitions—are complex, and may show large differences between genders and educational stages (representing maturational or learning differences). Future work could explore the effect and interaction of gender and age in a MIMIC model, for example, by introducing gender and age as covariates in the overall CFA model [31].

From the point of view of therapeutic intervention, given the importance of the relationship of the cognitive factors with problematic gaming severity, including them in psychological therapeutic programs as a part of treatment aimed at modifying problematic gaming seems promising. Probably, an adequate psychoeducation on problematic gaming, together with cognitive restructuring techniques of the identified maladaptive cognitions, could be of particular help [38,39]. In addition, problem solving techniques and, in some cases, exposure techniques may be useful to modify both cognitions and gaming behaviors and their use as coping strategies. Knowing which situations triggered maladaptive cognitions may help the clinician.

4.3. Limitations

When working with the general population, the percentage of people with problematic gaming is low, and the percentage of those who have already developed IGD is even lower (thus limiting the variability of the scores). Having a larger sample of players with problematic gaming and IGD (and, therefore, a wider range of scores) would allow greater discriminant capacity to the DCG Scale. Even so, with the current sample, a good discriminant validity has already been obtained, allowing us to differentiate participants with the highest and lowest problematic cognitions about gaming.

All the data were obtained by means of self-report, with the limitations that this single source of obtaining information supposes. In addition, the anchoring of the response options of each item of the DCG could produce skewed scores. While “always” is the semantic counterpart of “never”, it does not make sense in this context. The statistic in Table 1 seems to support this idea, with positive skewness in 14 out of 16 items. In future research with this instrument, we will consider changing the anchor to “very often” to avoid this bias and, probably, obtain better discriminant validity.

On the other hand, although there are participants between 12- and 22-years-old, and invariance of the instrument was found when applying it in the age groups, it would be convenient to study the validity of the instrument in university students and older adults (i.e., over 22 years of age). Finally, caution is advised if the scale is to be used with different, non-Spaniard populations.

5. Conclusions

The DCG Scale has proven to be a reliable and valid instrument to identify maladaptive cognitions about problematic gaming in Spanish adolescents and youth. Its basic structure involves three factors: Preoccupation, Self-esteem, and Compulsion. However, the DCG Scale is essentially unidimensional, so we recommend the use of the total score for assessment and screening. For this purpose, we have provided practical scoring scales segmented by gender and age. The scores of the scale showed a significant relationship with the presence of problematic gaming and other health-related outcomes, adding empirical support for the utility of this scale. The importance of evaluating these disabling cognitions is highlighted, which will help to design and apply psychological treatments for problematic gaming.