Internet Gaming Disorder: Investigating the Clinical Relevance of a New Phenomenon
The American Psychiatric Association (APA) identified Internet gaming disorder as a new potential psychiatric disorder and has recognized that little is known about the prevalence, validity, or cross-cultural robustness of proposed Internet gaming disorder criteria. In response to this gap in our understanding, the present study, a first for this research topic, estimated the period prevalence of this new potential psychiatric disorder using APA guidance, examined the validity of its proposed indicators, evaluated reliability cross-culturally and across genders, compared it to gold-standard research on gambling addiction and problem gaming, and estimated its impact on physical, social, and mental health.
Four survey studies (N=18,932) with large international cohorts employed an open-science methodology wherein the analysis plans for confirmatory hypotheses were registered prior to data collection.
Among those who played games, more than 2 out of 3 did not report any symptoms of Internet gaming disorder, and findings showed that a very small proportion of the general population (between 0.3% and 1.0%) might qualify for a potential acute diagnosis of Internet gaming disorder. Comparison to gambling disorder revealed that Internet-based games may be significantly less addictive than gambling and similarly dysregulating as electronic games more generally.
The evidence linking Internet gaming disorder to game engagement was strong, but links to physical, social, and mental health outcomes were decidedly mixed.
The American Psychiatric Association (APA) has identified Internet gaming disorder as a potential psychiatric disorder that might merit inclusion in a future revision of the DSM (1). In line with this possibility, the APA Substance-Related Disorders Work Group (2) has called for basic research exploring Internet gaming disorder prevalence, validity of Internet gaming disorder diagnostic criteria, and cross-cultural reliability and criteria (Table 1).
|1. Preoccupation with Internet gaming||Preoccupation with Internet games. (The individual thinks about previous gaming activity or anticipates playing the next game; Internet gaming becomes the dominant activity in daily life.)|
|2. Experienced withdrawal||Withdrawal symptoms when Internet gaming is taken away. (These symptoms are typically described as irritability, anxiety, or sadness, but there are no physical signs of pharmacological withdrawal.)|
|3. Developed tolerance||Tolerance—the need to spend increasing amounts of time engaged in Internet games.|
|4. Loss of control||Unsuccessful attempts to control the participation in Internet games.|
|5. Continued use||Continued excessive use of Internet games despite knowledge of psychosocial problems.|
|6. Mislead others||Has deceived family members, therapists, or others regarding the amount of Internet gaming.|
|7. Use as escape||Use of Internet games to escape or relieve a negative mood (e.g., feelings of helplessness, guilt, anxiety).|
|8. Reduced interests||Loss of interest in previous hobbies and entertainment as a result of, and with the exception of, Internet games.|
|9. Risked opportunities||Has jeopardized or lost a significant relationship, job, or educational or career opportunity because of participation in Internet games.|
Work responding to this call is at an early stage, and extant studies rely on constructs not informed by APA’s guidance. For example, a number of these early studies do not distinguish between offline and online games (3), and the flexible criteria used to date have meant estimates of addiction range from as low as 0.2% (4) to as high as 46% (5). A comprehensive review of this literature estimated a prevalence rate of 3.1%, although the experts who conducted the review caution that the accuracy of this figure is not reliable because relevant studies do not distinguish between passionate engagement and pathology (6).
The APA call for new research and unified criteria might help address concerns raised with regard to the existing body of research. First, in addition to defining key features of Internet gaming disorder, this guidance acknowledges that dysregulated gaming is characterized by significant distress, a nuance that may discriminate passion from pathology. Many players may experience a feature of Internet gaming disorder, for example, a preoccupation with a new game that distracts from other responsibilities. Much in the same way a sports fan might feel distracted at work if his or her team reaches the finals, feeling this way may be typical among those for whom gaming is a favored hobby. Such experiences are not necessarily pathological if unaccompanied by significant distress. The presence of distress for diagnosis may be key to accurately distinguishing individuals with pathological features from those with nonpathological features (6).
Second, DSM-5 guidance underlines the need for improving the methodologies used to study the potential disorder. With few exceptions (4, 7), most of what is known about dysregulated gaming comes from studying samples of convenience (6). Polling online support communities may exaggerate the clinical relevance of problem gaming, as these support communities sample those who have pre-existing difficulties regulating their behavior and are therefore seeking community help. Similarly, data from Internet-based gaming forums might oversample highly invested and engaged players and may therefore not reflect the experience of most players, given between one-half and three-quarters of people play such games (8, 9).
The present research employed large-scale national cohort samples and used an open-science methodology to evaluate four research questions (listed below) key to the APA call:
Research question 1: What is the acute prevalence rate of the Internet gaming disorder criteria proposed in the DSM-5 and of Internet gaming disorder diagnoses?
Research question 3: To what extent do the assumptions behind an indicator-based method for evaluating Internet gaming disorder hold up psychometrically? In the DSM-5 guidance, all nine symptoms are thought to equally contribute toward a diagnosis of Internet gaming disorder providing significant distress is present. Is this the case across demographic and national groups?
Research question 4: To what extent might those with Internet gaming disorder vary in terms of their everyday behaviors and clinical outcomes, as compared with those who do not meet criteria?
We present data from four studies: 1) a cohort of young adults aged 18–24 years old from the United States (study 1: females, N=527; males, N=720); 2) a sample of adults aged 18 years and older from the United Kingdom (study 2: females, N=941; males, N=958); 3) four cohorts of young adults aged 18–24 years old from the United States, United Kingdom, Canada, and Germany (study 3: females, N=4,995; males, N=5,014); and 4) a sample of adults aged 18 years and older from the United States (study 4: females, N=3,328; males, N=2,449). Participants were recruited through Google Surveys using joint distributions of age, gender, and geographic location for studies 1–3, and YouGov omnibus panel platform was used for study 4. Demographic information inferred from web tracking data identified participants and informed demographic quotas in studies 1–3, and YouGov participants were selected based on self-reported panel data for study 4. Ages for studies 1 and 2 were bucketed for samples of young adults aged 18–24 years old, were grouped into six age cohorts for study 2 (18–24 years old [21.9%], 25–34 years old [20.3%], 35–44 years old [15.4%], 45–54 years old [18.2%], 55–64 years old [14.2%], and 65 years and older [10.0%]), and were continuous for study 4 (mean age=46.59 years [SD=17.80]). Because the surveys presented low participant burden, the weighted completion rate was 92.23%. Google Surveys have been shown to be particularly effective in reaching dispersed populations (11, 12), while YouGov samples are used to study issues in depth (13), and both have been used to study health behaviors (14–16) and technology (17, 18) use in the general population (13, 19).
In studies 1–3, participants completed a brief indicators checklist drafted in consultation with clinical and research psychologists active in the area, and measures of health and behavior focused on the previous 6 months were added in study 4 (20). The research presented minimal risk and was granted clearance by the University of Oxford (CUREC/C1A15–006). Study 1 was treated as an exploratory study, whereas studies 2–4 had confirmatory aspects registered prior to data collection (21–23), and all data and materials are available via the Open Science Framework (24).
More than one-half of participants had recently played Internet-based games (Table 2). Preliminary reliability analyses indicated that Internet gaming disorder indicators loaded well together (alpha ranged from 0.68 to 0.76), and exploratory chi-square tests indicated no statistically significant differences in Internet gaming as a function of gender in studies 1 and 2 (all p values >0.11). In study 3, the proportion of Internet gaming was higher among males (83%) than females (78%) (χ2=32.3, df=1, p<0.01). In study 4, the proportion of Internet gaming was higher among females (68%) than males (62%) (z=4.19, p<0.01) (log-linear model with weights considered), reflecting a general trend toward egalitarianism among those who play games (25, 26).
|Observation||Study 1 (N=1,247)||Study 2 (N=1,899)||Study 3 (N=10,009)||Study 4 (N=5,777)|
|%||95% CI||%||95% CI||%||95% CI||%||95% CI|
|Recent internet gaming|
|Internet gaming disorder prevalence|
|Preoccupied with Internet gaming|
|Loss of control|
|Use as an escape|
|Experienced significant distress due to gaming|
Internet Gaming Disorder Indicator and Diagnosis Prevalence (Research Question 1)
The proportion of participants reporting indicators of Internet gaming disorder is presented in Figure 1. More than one-half of players reported no indicators (68.1% [study 1], 69.8% [study 2], 58.5% [study 3], and 68.4 [study 4], respectively), and the proportion monotonically decreased as the number of indicators increased. The proportion of participants who endorsed five or more indicators was 2.8% in study 1 (95% confidence interval [CI]=2.0%–3.9%]), 2.7% in study 2 (95% CI=2.0%–3.5%), 2.6% in study 3 (95% CI=2.3%–2.9%), and 1.2% in study 4 (95% CI=0.8%–2.0%), indicating that nearly 2.4% demonstrated potentially dysregulated gaming, a level close to the 3.1% estimated in a comprehensive recent meta-analysis (6).
To assess the period prevalence of Internet gaming disorder, we estimated the proportions of participants who reported that they suffered significant distress due to gaming and endorsed five or more of the indicators. Diagnosis prevalences were 1.0% in study 1 (95% CI=0.6%–1.8%), 0.5% in study 2 (95% CI=0.2%–0.9%), 0.7% in study 3 (95% CI=0.5%–0.9%), and 0.3% in study 4 (95% CI=0.2%–1.0%). The number of indicators endorsed was positively correlated with the distress criterion across all four studies (rs=0.24–0.33), and those who endorsed five or more of the indicators were more likely to report distress compared with those who did not (17%−37% compared with 1.3%−3.0%). For online game players only, the observed prevalences for Internet gaming disorder were 1.0% in study 1 (95% CI=0.5%–1.9%), 0.6% in study 2 (95% CI=0.3%–1.1%), 0.8% in study 3 (95% CI=0.7%–1.1%), and 0.5% in study 4 (95% CI=0.3%–1.0%).
Comparison to Disordered Gambling and Problem Gaming (Research Question 2)
Three samples were drawn to compare Internet gaming disorder rates to gold-standard research on gambling disorder, the only nonsubstance addiction recognized as a psychiatric condition, and to general gaming. Two subsamples were drawn from the British Gambling Prevalence Survey (10), one of 7,536 adults aged 18 years and older and a second of 757 adults ranging in age from 18 to 24 years old. Results indicated that 5,574 participants (74%) aged 18 years and older and 557 participants (73%) aged 18–24 years old had engaged in some form of gambling in the past year. This included, but was not limited to, participation in online gambling, the lottery, pool betting, sports betting, bingo, or casino games. A total of 73 participants aged 18 years and older (1.0%) and 20 participants aged 18 to 24 years old (2.6%) met established criteria for gambling disorder (27, 28). To evaluate differences in prevalences between gambling disorder and Internet gaming disorder (those who endorsed five of nine indicators and identified gaming as a significant source of distress), we compared our data in studies 2 and 3 with samples of problem gamblers. Results indicated that the rate of gambling addiction among the general United Kingdom population in studies 2 (z=–2.08, p=0.038) and 3 (z=–3.53, p<0.001) were higher than what was observed for Internet gaming disorder. Results also showed that the prevalence of Internet gaming disorder was lower among those who had played Internet-based games in the past year, compared with those with gambling disorder who had engaged in any form of gambling in the past year, in study 2 (z=–2.71, p=0.006) and in study 3 (z=–3.93, p<0.001).
A third sample composed of 656 Germans aged 18 to 24 years was drawn from Festl’s study of video game addiction (4) to compare with our estimate of Internet gaming disorder. Results indicated that only a single participant (0.2%) qualified as addicted to games (3), and our data collected from German participants (N=2,477) in study 3 identified that five participants (0.2%) met DSM-5 criteria for Internet gaming disorder. These proportions were not different in our German cohort (z=0.12, p=0.904).
Exploring the Validity of Self-Report to Assess Internet Gaming Disorder (Research Question 3)
A key feature of the DSM-5 guidance on Internet gaming disorder is that diagnosis can be made, in part, on the endorsement indicators of problem gaming. The implicit idea behind this approach is that these criteria equally contribute toward the diagnosis of Internet gaming disorder. Statistically, this is the assumption of a Rasch model in item response theory (29) and can be tested by examining the fit of the model to the data using structural equation modeling. In all studies, our analysis showed a very good fit of the Rasch model to the data across gender, comparative fit index (0.97–0.99), Tucker-Lewis index (0.97–0.99), and root mean square error of approximation (0.017–0.029), as well as across the four countries (United States, United Kingdom, Canada, and Germany) in study 3 (χ2=408.3, df=140, p<0.001; comparative fit index=0.97; Tucker-Lewis index=0.97; root mean square error of approximation=0.028). These results suggest that items assessed Internet gaming disorder with the same sensitivity and difficulty across gender and across these countries.
Behavioral and Clinical Impact of Internet Gaming Disorder (Research Question 4)
Given that Internet gaming disorder is thought to have a practically significant influence on functioning, akin to psychiatric disorders, we tested a preregistered hypothesis that those meeting the diagnostic threshold would show more frequent gaming and less frequent physical exercise (physical activity) and quality social time with others (social activity), compared with those who did not meet the diagnostic threshold. A series of one-way Bayesian t tests using a default Cauchy prior of 0.707 for the effect size of the alternative hypothesis tested confirmatory relations between Internet gaming disorder and behavioral engagement with games and physical and social activity (30, 31). Bayesian t test was selected for our registered analysis plan in place of null hypothesis testing because it quantifies the relative evidence for the alternate hypothesis with moderately sized effects compared with the null (32, 33). In line with best practices, if observed Bayes factors were 3 or above, we considered our hypotheses to be supported; if Bayes factors were 1/3 or below, we considered the null hypothesis to be supported; and if Bayes factors observed were between 1/3 and 3, we considered the results inconclusive (34). Full results are presented in the data supplement accompanying the online version of this article. Internet gaming disorder was significantly linked to higher levels of regular gaming (Bayes factor=11.29), showing that engagement levels were higher for those meeting the Internet gaming disorder threshold (mean=4.00 [SD=1.04]) than for those not meeting the threshold (mean=2.80 [SD=1.67]). Those meeting the Internet gaming disorder threshold reported overall lower levels of physical activity (mean=2.92 [SD=1.49] compared with 3.26 [SD=1.39]) yet higher social activity (mean=3.92 [SD=1.00] compared with 3.61 [SD=1.14]), but evidence from Bayes factors indicated these differences were not significant.
Those meeting the Internet gaming disorder threshold reported marginally lower levels of mental health (mean=2.77 [SD=1.01] compared with 2.78 [SD=1.01]) and marginally higher levels of physical health (mean=2.33 [SD=1.23] compared with 2.31 [SD=0.94]) and social health (mean=2.64 [SD=1.03] compared with 2.23 [SD=0.96]), but evidence derived from Bayes factors indicated these differences were not significant.
The present research represents the first large-scale studies, to our knowledge, of Internet gaming disorder guided by an open-science approach and grounded in APA criteria. The studies addressed fundamental questions about this potential psychiatric condition and provided evidence regarding the acute symptom patterns, potential diagnoses, and clinical and behavioral impact of this condition.
Results indicated that Internet-based games are widely popular among adults in the United States, the United Kingdom, Germany, and Canada. At the same time, the great majority of players, nearly three in four, reported no indications of behavioral dysregulation. Specific indicators, such as increasing play time to maintain excitement, were reported roughly three times more frequently than other indicators, such as risking social relationships. Importantly, all criteria were relevant to a potential diagnosis, with the least common still reported consistently across studies and all appearing to be psychometrically sound. These findings are promising because they suggest the proposed criteria tap into less frequent or more extreme symptoms and are appropriate for characterizing the phenomenon.
Second, findings suggested the rates of potential Internet gaming disorder diagnosis estimates based on DSM-5 criteria are quite low. Our results indicated the acute period prevalence rate might realistically be as high as 1.0% among young adults (studies 1 and 3) and 0.5% among all adults (studies 2 and 4) in the countries we studied. Indeed, acute prevalence rates of gambling, the only behavioral addiction in DSM-5, were notably higher. This provides tentative evidence that despite being a new and popular activity, Internet-based games might be less dysregulating than gambling. Finally, our findings indicated that Internet gaming disorder classifications did predict gaming engagement, but there was little evidence for other behavioral or clinical effects.
This study informs ongoing debate about Internet gaming disorder between those who argue for (35) and against (36) an international consensus regarding gaming addiction. It may be that Internet gaming disorder can be detected in line with DSM-5 guidance, but such classifications might reflect self-regulatory challenges epiphenomenal to electronic game play. Our analyses suggest that there might be cross-cultural variability in Internet gaming disorder. Prevalence estimates in study 3 varied significantly across the countries studied (χ2=12.00, df=3, p<0.01), with Germans showing the lowest levels. Research comparing Internet gaming disorder prevalence and stability across a wider range of cultures, such as Asian countries, where gaming is widespread and played in different social settings will be useful (2). These studies relied on self-reported data, and evidence derived from convergent sources including peers, caregivers, and health specialists is needed.
The present study carries three important takeaways for developing reliable and robust research into Internet gaming disorder. First, unlike most research on technology addiction, the present work involved collecting data generalizable to adult populations in a number of countries. Maintaining this standard will allow for direct comparisons between datasets and studies in this new and developing area (4). Second, these data are publicly available (24). This affordance increases the robustness of research (30) and minimizes wasted resources in an area rife with duplicate efforts (37). Finally, this research identified its exploratory and confirmatory features, given some predictions were registered before the start of data collection (21–23). Because clinics in a number of countries are already claiming to treat gaming addiction (38), practitioners should be made aware that exploratory findings, which are especially susceptible to false positive results (39), should be weighted differently than confirmatory ones.
Internet-based games are currently one of the most popular forms of leisure, and researchers studying their potential “darker sides” must be cautious. If one extrapolates from our data, upwards of 160 million American adults play Internet-based games, and as many as one million of these individuals might meet the proposed DSM-5 criteria for addiction to online games (40). This represents a large cohort of people struggling with what could be clinically dysregulated behavior. However, because we did not find evidence supporting a clear link to clinical outcomes, more evidence for clinical and behavioral effects is needed before concluding that this is a legitimate candidate for inclusion in future revisions of the DSM. If adopted, Internet gaming disorder would vie for limited therapeutic resources with a range of serious psychiatric disorders.
8 Lenhart A, Smith A, Anderson M, et al: Teens, Technology and Friendships. Washington, DC, Pew Research Center, 2015. http://www.pewinternet.org/2015/08/06/teens-technology-and-friendships/Google Scholar
9 Duggan M: Gaming and Gamers. Washington, DC, Pew Research Center, 2015. http://www.pewinternet.org/2015/12/15/gaming-and-gamers/Google Scholar
10 : British Gambling Prevalence Survey, 2010Google Scholar
12 McDonald P, Mohebbi M, Slatkin B: Comparing Google Consumer Surveys to Existing Probability and Non-Probability Based Internet Surveys. http://www.google.com/insights/consumersurveys/static/consumer_surveys_whitepaper.pdfGoogle Scholar
16 : Adult and adolescent exposure to tobacco and alcohol content in contemporary YouTube music videos in Great Britain: a population estimate. J Epidemiol Community Health 2016; 70:488–492Crossref, Medline, Google Scholar
17 Müller H, Gove JL, Webb JS, et al: Understanding and Comparing Smartphone and Tablet Use: Insights from a Large-Scale Diary Study. http://static.googleusercontent.com/media/research.google.com/en//pubs/archive/44200pdf.Google Scholar
18 : Banking without banks: exploring the disruptive effects of converging technologies that will shape the future of banking. J Secur Oper Custody 2014; 7:69–80Google Scholar
19 Schnorf S, Sedley A, Ortlieb M, et al: A comparison of six sample providers regarding online privacy benchmarks, in the proceedings of SOUPS Workshop on Privacy Personas and Segmentation, 2014Google Scholar
25 Duggan M: Who Plays Video Games and Identifies as a “Gamer.” Washington, DC, Pew Research Center, 2015. http://www.pewinternet.org/2015/12/15/who-plays-video-games-and-identifies-as-a-gamer/Google Scholar
26 : Women, socialization and video games. Women High Educ 2011; 20:8–9Google Scholar
27 Maitland S, Adams GR: Replication and Generalizability of the Problem Gambling Severity Index: Are Results Consistent and Comparable Across Studies? Final Rep Ont Probl Gambl Res Cent 2007Google Scholar
28 Guelfi J-D: DSM-4-TR. Elsevier Masson, 2004Google Scholar
29 : Applications of Item Response Theory to Practical Testing Problems. New York, Routledge, 1980Google Scholar
32 Morey RD, Romeijn J-W, Rouder JN: The philosophy of Bayes factors and the quantification of statistical evidence. http://pcl.missouri.edu/sites/default/files/paper.pdfGoogle Scholar
33 JASP Team: JASP, Version 0.7.5.6Google Scholar
36 : Working towards an international consensus on criteria for assessing internet gaming disorder: a critical commentary on Petry et al. (2014). Addiction 2016; 111:167–175Crossref, Medline, Google Scholar
38 Too Much of a Good Thing. http://taiwantoday.tw/ct.asp?xItem=241186&ctNode=2235&mp=9Google Scholar
40 United States Department of Commerce: Bureau of the Census: Census of Population and Housing, 2010 [United States]: National Summary File of Redistricting Data. http://www.icpsr.umich.edu/ICPSR/studies/33442/version/1Google Scholar