Edited by: Ofir Turel, California State University, Fullerton, United States
Reviewed by: Hamed Qahri-Saremi, DePaul University, United States; Elisa Wegmann, Universität Duisburg-Essen, Germany
This article was submitted to Psychopathology, a section of the journal Frontiers in Psychology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Playing video games is a common pastime activity among adolescents that, for the majority, provides hours of fun, challenge, relaxation and socialization (Hoffman and Nadelson,
Longitudinal studies might investigate whether mental health problems first and foremost are predictors of gaming disorder, whether mental health problems mainly are consequences of gaming disorder, or whether the relationship between mental health problems and gaming disorder are of a cross-lagged nature. Cross-lagged effects estimate the reciprocal relationship between variables over time, describing their mutual influence on each other (Kearney,
Previous research has shown that gaming disorder is associated with an array of health-related and social problems (Wittek et al.,
Across several studies, sex seems to be a robust predictor of video gaming, as males are more likely to engage in video games (Mentzoni et al.,
Previous longitudinal studies have assessed pathological gaming as a unidimensional construct (Lemmens et al.,
Against this backdrop, the aim of the first study was to identify antecedents and consequences, as well as sex differences, of video game problems. An unidimensional conceptualization of gaming disorder used in previous studies (Lemmens et al.,
The conceptualization of gaming disorder emphasizes functional impairment and psychological distress to distinguish the disorder from high involvement in gaming (Charlton and Danforth,
Applying a typological perspective enables investigation of whether there are any similarities or differences between “addicted gamers,” “problem gamers,” and “engaged gamers.” In line with such distinction, one study found gaming engagement to be more weakly associated with mental health outcome than addiction (Loton et al.,
The aim of Study 2 was thus to investigate antecedents and consequences of the three typologies (addicted, problem, and engaged) of gamers over time. We expected to find a larger number of antecedents, and a larger number of consequences associated with “addicted gamers,” than for “problem gamers,” and “engaged gamers.” Due to the explorative nature of this study, and lack of previous studies investigating different gaming behaviors applying a typological approach, all variables (mental health and gaming) were investigated both as antecedents and consequences.
In addition to exploration of causes and consequences, longitudinal studies provide the possibility to investigate stability of a condition over time. The temporal stability of gaming disorder provides an indication of whether the disorder is a transient problem that resolves spontaneously, due to for instance maturation, or if the condition is rather persistent. To date, results from studies investigating the stability of gaming disorder have been mixed. One study found a high temporal stability of 84% after 2 years (Gentile et al.,
Therefore, the aim of study 3 was to investigate the temporal stability of “addicted gamers,” and the transitions occurring between “problem gamers,” “addicted gamers,” and “engaged gamers” over time.
All three studies employed data from the same large longitudinal survey of gambling, gaming, and drug behavior in adolescents. A nationally representative sample of 3,000 adolescents (50% female) aged 17.5 years was drawn from the Norwegian Population Registry in 2012 (Wave 1). The adolescents were informed about the purpose of the study, that all data would be treated confidentially, and that data would be used only for research purposes. Written informed consent was obtained from all the participants. Parental consent was not required on account of the adolescents being above the age of 16. All who responded at Wave 1 received annual follow-up surveys by postal mail (2013 and 2014) with up to two reminders for each wave. The survey could be answered on paper and returned via an included prepaid envelope or answered online. All participants received a gift certificate with a value of 200 NOK (~18 UK £) upon completion of each wave. The study, including the consent procedure aforementioned, was approved by Regional Committee for Medical and Health Research, Ethics, South East Region (Project Number: 2012/914).
Data from all three waves (2012, 2013, 2014) were used in the three studies included in the current paper. Of the 3,000 adolescents who were invited in 2012, 54 were not reachable due to invalid addresses, whereas 23 were not able to respond due to other reasons such as cognitive disability, reducing our sample to 2,923. In the first wave, 2,059 adolescents responded (response rate 70.4%, 53% female). Four cases were excluded because they were younger than 17 years old, and four cases did not indicate their sex and were excluded. In the second wave, a total of 1,334 individuals responded (retention rate 64.9%, female 58.7%); and in the final wave, 1,277 responded (retention rate 62.1%, female 61.7%).
The questionnaire contained sociodemographic questions including sex.
Pathological gaming was assessed using the Game Addiction Scale for Adolescents (GASA) (Lemmens et al.,
To measure symptoms of anxiety and depression the Hospital Anxiety and Depression Scale (HADS) (Zigmond and Snaith,
To measure loneliness we administered the Roberts UCLA Loneliness Scale (RULS) (Roberts et al.,
Alcohol consumption was assessed with the Alcohol Use Disorder Identification Test-Consumption (AUDIT-C) (Bush et al.,
The Physical and Verbal aggression subscales of the Short Form Buss-Perry Aggression Questionnaire (BPAQ-SF) (Diamond and Magaletta,
Preliminary analysis and attrition analysis was conducted using SPSS, version 25 (Corp,
Further analysis was carried out using the multi-group path analysis in Mplus, version 7.4 (Muthén and Muthén,
In the first study, a cross lagged path model with observed indicators was tested to measure the cross-lagged effect of mental health outcomes and gaming across the three waves (see Figure
Cross-lagged path modeling of pathological gaming (GASA) against mental health variables (outcome measures).
In the restricted models for
The restricted models were compared with the unrestricted model using the Satorra-Bentler chi square test (Satorra and Bentler,
To test the predictions for a typology perspective, we classified gaming status into four groups: (1) Engaged gamers, (2) Problem gamers, (3) Addicted gamers, and (4) Non-problem/non-engaged/non-addicted contrast group (hereafter denoted “contrast group”) using the Core 4 approach (Brunborg et al.,
In a series of subsequent regressions models, we investigated whether gaming status predicted mental health outcomes (hereafter called “consequences”) and whether mental health outcomes predicted gaming status (hereafter called “antecedents”). As gaming status was a nominal variable, multinomial regression analysis was used to investigate the antecedents of gaming status. To examine the consequences of gaming status, gaming status was pacifier coded and used as independent variables with mental outcomes as the dependent variable. We controlled for sex, typology identification, and scores on the previous wave on the outcome measures. Both analyses were conducted in the same model, and the analysis was repeated for each outcome measure. Categories of gaming and outcome measures were first analyzed with 1 year intervals (T1–T2, T2–T3), and then in a new model investigating the effects over 2 years (T1–T3). Results of the regression analyses will be discussed according to the findings in study 1, to further investigate the identified associations between pathological gaming to mental health over time.
To investigate the stability and trajectories of addicted gamers, problem gamers and engaged gamers, we estimated a hidden Markov model of transition probabilities between the three typologies of gamers, and the contrast group. Hidden Markov models are used to estimate transition probabilities between categorical variables for time series. Observed values are used to estimate the underlying and unobserved Markov process, also known as a Markov chain, which rests on the assumption that the probability of a current state is dependent on the previous state (Zucchini et al.,
Of the 2,055 participants, 21 were excluded due to missing items on GASA at Wave 1. Of the remaining participants, 999 participated at all waves; 256 were missing at Wave 2, 309 were missing at Wave 3, and 470 were missing at both Waves 2 and 3. In general, predictors of being missing were weak with a few exceptions. Being missing at Wave 3 was predicted by being male (
The results of the association between pathological gaming and health outcomes in the unrestricted model are reported in Table
A cross-lagged path model of the antecedents and consequences of gaming problems.
Boys | 0.14 |
0.07 | 0.03 | 0.01 | 115.33 | 0.907 | 0.674 | 0.114 | 0.050 | 963 |
Girls | 0.13 |
0.12 |
0.11 |
0.12 |
1,088 | |||||
Boys | 0.11 |
0.07 | 0.03 | −0.02 | 93.93 | 0.936 | 0.776 | 0.102 | 0.042 | 963 |
Girls | 0.07 |
0.07 |
0.05 | 0.05 | 1,088 | |||||
Boys | 0.07 | 0.05 | 0.04 | 0.01 | 92.47 | 0.933 | 0.767 | 0.101 | 0.044 | 962 |
Girls | 0.07 | 0.08 | 0.10 |
0.08 |
1,088 | |||||
Boys | −0.03 | −0.05 | −0.05 | −0.06 | 76.62 | 0.938 | 0.784 | 0.091 | 0.038 | 963 |
Girls | 0.01 | 0.01 | −0.001 | −0.04 | 1,087 | |||||
Boys | 0.09 |
0.02 | 0.04 | −0.04 | 103.73 | 0.923 | 0.730 | 0.108 | 0.043 | 963 |
Girls | 0.03 | 0.03 | 0.02 | −0.003 | 1,088 | |||||
Boys | 0.05 | −0.03 | 0.05 | 0.05 | 87.91 | 0.938 | 0.782 | 0.099 | 0.040 | 963 |
Girls | 0.04 | 0.06 | 0.08 |
0.05 | 1,088 |
Results of the Satorra-Bentler test of chi square differences between an unrestricted
Satorra-Bentler chi square test comparing the restricted path models regarding consequences of pathological gaming, antecedents for pathological gaming, stationary assumptions, and sex differences to the unrestricted model.
Depression | 37.84 |
20.47 |
6.16 | 11.52 |
Anxiety | 19.51 |
6.18 | 13.27 | 5.99 |
Loneliness | 12.82 |
16.92 |
8.50 | 10.43 |
Alcohol | 2.18 | 5.41 | 12.14 | 13.27 |
Verbal aggression | 7.28 | 2.38 | 8.86 | 7.86 |
Physical aggression | 7.31 | 10.66 |
12.03 | 7.44 |
Table
Multinomial regression analysis showing antecedents for “engaged gamer,” “problem gamer” and “addicted gamer.” The contrast group comprises the reference category.
T1–T2 | 1.11 |
1.11 |
1.08 |
T2–T3 | 1.04 |
1.11 |
1.22 |
T1–T3 | 1.15 |
1.05 |
1.09 |
T1–T2 | 1.08 |
1.05 |
1.07 |
T2–T3 | 1.09 |
0.98 |
0.93 |
T1–T3 | 1.06 |
1.05 |
0.97 |
T1–T2 | 1.11 |
1.07 |
1.06 |
T2–T3 | 1.08 |
1.05 |
1.07 |
T1–T3 | 1.08 |
1.08 |
1.16 |
T1–T2 | 0.90 |
0.97 |
1.19 |
T2–T3 | 0.87 |
0.78 |
1.46 |
T1–T3 | 0.94 |
0.78 |
0.96 |
T1–T2 | 1.16 |
1.11 |
1.15 |
T2–T3 | 1.00 |
0.95 |
0.66 |
T1–T3 | 0.96 |
1.04 |
0.75 |
T1–T2 | 1.12 |
1.10 |
1.19 |
T2–T3 | 1.01 |
1.05 |
0.91 |
T1–T3 | 1.04 |
1.02 |
0.94 |
Table
Regression analysis showing consequences of being “engaged gamer,” “problem gamer” and “addicted gamer,” compared to the contrast group with 1 year intervals between the measures.
Engaged | 0.13 | 0.30 | 0.38 |
Problem | 0.42 |
0.12 | 0.33 |
Addicted | 0.52 | 0.33 | 0.58 |
Engaged | 0.04 | 0.07 | 0.19 |
Problem | 0.15 | 0.15 | 0.13 |
Addicted | 0.24 | −0.04 | 0.38 |
Engaged | 0.15 | 0.29 | 0.06 |
Problem | 0.30 |
0.14 | 0.30 |
Addicted | 0.11 | −0.06 | 0.08 |
Engaged | −0.07 | 0.26 | −0.02 |
Problem | −0.01 | −0.22 | −0.02 |
Addicted | −0.08 | −0.26 | −0.01 |
Engaged | 0.10 | 0.01 | 0.25 |
Problem | 0.19 |
−0.03 | 0.11 |
Addicted | 0.21 | 0.08 | −0.15 |
Engaged | 0.09 | −0.04 | 0.11 |
Problem | 0.06 | −0.01 | 0.13 |
Addicted | −0.18 | 0.38 | −0.12 |
The distribution of “engaged gamer,” “problem gamer,” “addicted gamer” and contrast group over the three waves is found in Appendix
Latent transition probability of the four categories of gamers based on a Hidden Markov analysis reported in percentage.
Engaged | 52 | 20 | 02 | 26 |
Problem | 16 | 59 | 08 | 17 |
Addiction | 00 | 53 | 35 | 12 |
Contrast | 00 | 00 | 00 | 100 |
Sankey chart depicting the estimated transitions between the three typologies of gamers, and the contrast group of the adolescents that did not fall into the three gaming typologies. The estimation is based on transitions between T1–T2–T3.
The aim of study 1 was to identify antecedents and consequences of pathological gaming over a timespan of 2 years. We identified, as expected, cross-lagged association between mental health symptoms in terms of loneliness and depression to pathological gaming. Regarding loneliness, our findings were consistent with the findings reported by Lemmens et al. (
Consistent with previous studies (Lemmens et al.,
The aim of study 2 was to investigate the specific associations between mental health and the three typologies of gaming, using non-engaged/non-problem/non-addicted gamers, as the reference. We investigated typologies both as consequences and as antecedents of mental health. We expected to find that the addicted group of gamers would be associated with a higher number of antecedences, and a higher number of consequences than the other groups of gamers, which proved not to be the case in the current study. Building on the findings from study 1, the relevant results from study 2 include depression as a antecedent for engaged and problem gamer, loneliness and physical aggression as antecedents for all typologies. Regarding consequences, the relevant associations include depression as a consequence of all typologies, anxiety as a consequence of gaming addiction, and loneliness as a consequence of problem gaming typology.
As discussed in study 1, we identified cross-lagged effects between pathological gaming and loneliness and depression. When investigating the same variables in association to the typologies, loneliness was identified to be an antecedent and a consequence of problem gaming', and depression was found to be both an antecedent and a consequence of problem gaming as well as engaged gaming. The weak but significant, reciprocal association between depression and engaged gaming was somewhat surprising as previous research have not reported any detrimental effects of video game engagement (Brunborg et al.,
There were also some interesting findings specific to addicted gaming, which might explain why more severe psychopathology is found to be associated with this group (Loton et al.,
Our findings show that physical aggression predicted all categories of gaming compared to the reference group. This suggests that the association between gaming and aggression might be explained by adolescents with aggressive tendencies and related psychological characteristics (Kim et al.,
The aim of study 3 was to investigate the temporal stability of gaming typologies, and the transitions occurring between such typologies and the contrast group. The temporal stability of addicted gamers was estimated to be 35% which is in the middle of ranges reported by other studies (< 1–84%) (Gentile et al.,
The findings suggest that for all typologies of gamers, except gaming addicts, there was a higher probability to remain in the same category over time than to leave. Furthermore, no one from the contrast group transitioned to any of the gaming typologies, which suggests that symptoms of pathological gaming emerge early in the developmental history. This might explain the lack of new recruiting into the groups of engaged, problem, and addicted gamers.
The current paper applied a broad approach to investigate the natural course of gaming behavior over time, exploring directionality of associations with mental health, as well as trajectories between different gaming typologies. Applying a unidimensional and a typological conceptualization of gaming behavior in the same sample indicate that an unidimensional approach to gaming disorder (as used in study 1) without further exploration, might result in ignoring potentially important distinctions between gaming typologies. For instance, the direction between alcohol consumption and “problem gaming” (negative) and “addicted gaming” (positive) in study 2, was opposite of each other, while study 1 did not detect any association between alcohol consumption and pathological gaming. Hence, study 2 found that there are differences between the gaming typologies that could not be investigated in study 1, where pathological gaming was measured continuously. The distinctions between categories of heavily involved gamers might be important for identification of adolescents specifically in need of treatment, and for the development of interventions for prevention and treatment.
Results of study 2 and 3 combined provide knowledge of the natural course of different gaming typologies. Inspecting the trajectories of study 3 indicates that most addicted gamers (53%) transit to “problem gamers,” or remain addicted (35%). This is interesting with regards to findings in study 2, that point to several negative consequences of “problem gaming” and “addicted gaming.” In sum, the stability of addicted gamers and problematic gamers was quite high, and it seems fair to assume that many of these adolescents are in need of treatment or other kinds of support.
According to our findings in study 1, depression and loneliness seem to interact with symptoms of pathological gaming in a mutually self-enhancing and/or upholding cycle. This was by and large supported by the findings from study 2, with some distinctions between the typologies. An explanatory model of pathological internet use in adolescents (Strittmatter et al.,
Model depicting the proposed mechanisms acting between depression, loneliness, and pathological gaming.
Findings in study 3 indicate that engaged gamers are more likely than the other typologies to transit to the contrast group, while addicted gamers are the least likely to transit to the contrast group. Furthermore, there seem to be virtually no transitions between “engaged gamers” and “addicted gamers.” This suggests that the probability becoming addicted due to engaged gaming is small. Findings in study 2 show that engaged gamers are also are the category with fewer negative consequences than the other typologies. Worth noting when inspecting the mean scores of engaged gamers (see Appendix
The question of why some develop gaming disorder while other develop less severe gaming behaviors (problem or engaged gaming) remains an issue for further studies. In accordance with previous research (Lemmens et al.,
A major strength of the present study is the broad, longitudinal approach which provides insight into the trajectories between mental health variables and pathological gaming, and specifically, enables examination of the three typologies over time. Other strengths include the large sample size, random sampling from the National Population Registry, and high initial response rate. Previous longitudinal studies have been criticized by not taking the initial level of variables into account (Scharkow et al.,
One limitation of the present study is the issue of generalizability. The sample consists of adolescents between the age of 17.5–19.5 years, who are the most at-risk age group for addictive behaviors, and therefore results may not generalize to other age groups. Furthermore, the attrition analysis found several predictors for dropout at T2 and T3 (sex, alcohol consumption, and addicted gaming), which might indicate a certain selection bias. This might have affected the statistical power of our study, and it could therefore have been beneficial to have a larger sample. However, the consequential control of sex, and previous level of all variables do assumingly reduce attrition effects.
Another limitation is that model fit in study 1 is not optimal, indicating that that results from study 1 should be interpreted with some caution. One explanation for the mediocre fit might be that the theoretical assumption allowing for sex differences, and inequality across time in the unrestricted model was invalid. This is supported by the results of the Satorra Bentler test. Few degrees of freedom might also inflate the RMSEA, and rejection of models with poor fit is thus not necessary recommended (Kenny et al.,
Another limitation worth mentioning is that the reliability analysis showed somewhat low internal consistency (Nunnally,
The current study shows that mental health problems seem to interact closely with gaming pathology, both as antecedents and consequences over time. Several similarities were identified between engaged gamers, problem gamers and addicted gamers, and there seems to be a significant transition between the typologies, but not between addicted gamers and engaged gamers. “Engaged gamer” is associated with less severity in regard of negative consequences, whereas being an addicted gamer is associated with more severe psychopathology. Viewing gaming problems from a typological perspective might be useful in further assessment and conceptualization of video game addiction, both in research and in clinical settings. Further, the results suggest that gaming disorder has a relatively high stability, indicating that for a substantial group addicted gamers, their symptoms do not seem to resolve spontaneously, indicating the need for intervention or clinical management.
SP, RM, TT, HM, and DK stood for the conception and design of the work. All authors contributed to the acquisition, analysis, and interpretation of data. EK drafted the work. All authors revised the work critically in terms of important intellectual content. All authors approved the final version and are accountable for all aspects of the work in terms of ensuring that questions related to the accuracy or integrity of any part of the work were appropriately investigated and resolved.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The Supplementary Material for this article can be found online at: