How common is gaming disorder? Large systematic reviews have estimated that roughly 2%–3% of people in mixed-population studies screen positive for gaming disorder when sampling quality is emphasized, while other pooled analyses report about 3%. Estimates are considerably higher in many adolescent-focused studies, but there is no single scientifically defensible worldwide percentage that applies to every age group and diagnostic system. The most important finding is that case definitions, survey instruments, sampling frames, and clinical verification change the answer. Stevens et al., 2021; Kim et al., 2022; Poon et al., 2026.
A prevalence estimate is meaningful only when its denominator, time frame, population, and method of identifying a case are known. A survey of adolescents who play games daily answers a different question from a national probability survey of all adults. A questionnaire threshold answers a different question from a structured clinical evaluation of impaired control, persistence despite harm, and significant functional impairment. The public-health task is to identify people with consequential problems while retaining an accurate picture of the much larger population who play for enjoyment, friendship, skill, creative expression, or competition.
The best-supported estimates and what they actually measure
Three percent in a broad international meta-analysis
The influential systematic review by Stevens, Dorstyn, Delfabbro, and King combined 53 studies involving 226,247 people across 17 countries and estimated a pooled gaming disorder prevalence of 3.05%, with a 95% confidence interval of 2.38%–3.91%. The authors did not present that number as a census of everyone in the world. Studies varied widely in how they recruited participants, applied diagnostic instruments, and defined positive cases. When their analysis emphasized more stringent sampling methods, the pooled estimate was 1.96%, accompanied by an exceptionally wide reported interval of 0.19%–17.12% in that subgroup. These are model-based research summaries, not independently confirmed clinical diagnoses. Stevens et al., 2021.
The same review found that the selected screening instrument accounted for a large share of the statistical differences among study estimates: the tool moderator explained approximately 77% of the modeled between-study variance. The review also associated higher reported prevalence with adolescent samples, smaller samples, and lower classification thresholds. This combination has a practical consequence. An apparent difference between two countries can emerge when one study measures several relatively broad signs of problematic gaming and the other requires a narrower, more impairment-centered pattern. Stevens et al., 2021.
A second major synthesis: 3.3%, then lower under stricter assumptions
Kim and colleagues synthesized 61 studies, including 227,665 participants from 29 countries, and calculated an overall pooled prevalence of 3.3% (95% CI 2.6%–4.0%). Restricting the analysis to 28 studies characterized as having representative samples lowered that estimate to 2.4% (95% CI 1.7%–3.2%). Applying a trim-and-fill procedure to address possible small-study or publication-related asymmetry produced an adjusted estimate of 1.4% (95% CI 0.9%–1.9%). These three figures are sensitivity analyses of the same research domain. Their divergence reveals how dependent a pooled estimate is on study quality and statistical assumptions. Kim et al., 2022.
The 1.4% trim-and-fill figure is not a measured gold-standard rate, because trim-and-fill assumes a model of missing studies and can behave unpredictably under strong heterogeneity. Likewise, the representative-sample subset is not automatically free of screening misclassification. The strongest interpretation is that multiple analytic choices place ordinary population estimates in the low single digits and that exact point estimates remain method-sensitive. Reading only the largest pooled percentage or only the smallest correction would give a distorted picture of the same evidence. Kim et al., 2022; Stevens et al., 2021.
Why adolescent reviews report higher numbers
Adolescent-focused reviews address a population with different exposure patterns, developmental circumstances, social expectations, and survey designs. Satapathy and colleagues combined 84 studies involving 641,763 adolescents and reported a pooled prevalence of 8.6% (95% CI 6.9%–10.8%). Heterogeneity was extreme, with I² reported as 100%. In such a setting, the pooled value describes the statistical center of highly dissimilar studies. It cannot establish that 8.6% of every adolescent population has been clinically diagnosed with ICD-11 gaming disorder. Satapathy et al., 2025.
A 2026 child-and-adolescent meta-analysis by Barboza and colleagues included 19 studies with 251,037 participants and reported pooled Internet Gaming Disorder prevalence of 10% (95% CI 6%–15%). Individual study results stretched from around 1% in some settings to more than 25% in others. The reported I² of 99.7% indicates exceptionally large between-study variation relative to expected sampling error. These studies were focused on children and adolescents, commonly used DSM-oriented questionnaires, and did not all deliver equivalent clinician-established diagnoses. Their reported proportion should be presented as an instrument- and study-dependent estimate for the populations investigated. Barboza et al., 2026.
The 8.6% and 10% figures cannot be treated as proof that gaming disorder has become more common since an earlier review reported 3.05%. The reviews differ in age inclusion, search dates, underlying studies, screening instruments, geographical representation, and diagnostic terminology. The same primary studies can also appear in more than one meta-analysis. To establish a real increase in prevalence, investigators would need repeated representative surveys using equivalent diagnostic constructs, similar sampling methods, comparable observation periods, and stable thresholds across time.
What the 2026 umbrella review adds
Poon and colleagues examined nine published meta-analyses and reported a median of 6.20% across their published prevalence estimates, with an interquartile range of 4.25%–8.30%. That statistic is the median of meta-analytic results; it is not a newly pooled individual-level rate, a weighted international population prevalence, or a probability that a randomly selected person has a diagnosis. An umbrella review can expose disagreements across research programs, but overlapping primary studies and changing definitions restrict its capacity to produce a single global denominator. Poon et al., 2026.
The scientific lesson across these reviews is stable even when their summary numbers differ. Gaming-related impairment is a genuine clinical and public-health concern, yet prevalence estimates are highly contingent on design. A measured percentage should travel with the characteristics of its population and its instrument, especially when it is quoted in headlines, educational material, health-policy briefs, or discussions about children.
What counts as a case of gaming disorder?
The ICD-11 clinical framework
The World Health Organization's ICD-11 defines gaming disorder through a persistent pattern involving impaired control over gaming, increased priority given to gaming over other interests and activities, and continuation or escalation of gaming despite negative consequences. The pattern must be sufficiently severe to cause significant impairment in personal, family, social, educational, occupational, or other important functioning. It would normally have been evident for at least 12 months, although the time requirement can be shortened in sufficiently severe presentations. Both online and offline video gaming are included. World Health Organization, 2020.
Under an ICD-11 reading, a person can play for many hours, pursue elite competition, or devote substantial free time to games while retaining effective control, relationships, responsibilities, and psychological well-being. Duration alone does not supply the required impairment and persistence criteria. Conversely, clinically consequential problems can occur without an unusually spectacular number of hours if gaming repeatedly displaces essential activities and the person cannot regulate the pattern. This distinction shapes the numerator in every prevalence calculation: counting intensive activity as illness would expand it beyond the clinical definition. World Health Organization, 2020.
DSM-5-TR Internet Gaming Disorder is a different classification context
In the American Psychiatric Association's DSM framework, Internet Gaming Disorder was introduced as a condition for further study rather than the same established disorder category as ICD-11 gaming disorder. Its research criteria organize nine proposed symptom domains over approximately a 12-month period, commonly evaluated with a five-of-nine symptom threshold and attention to clinically significant impairment or distress. ICD-11 uses a more compact impairment-centered clinical pattern. Researchers may label studies with Internet Gaming Disorder, gaming disorder, video-game addiction, or problematic gaming even when their tools and thresholds are not interchangeable. Wang and Cheng, 2020; King et al., 2020.
Because the diagnostic architectures differ, a study applying the DSM-oriented Internet Gaming Disorder Scale–Short Form and another applying the ICD-11-oriented Gaming Disorder Test can identify overlapping but different groups. Differences in endorsed items, threshold rules, recall periods, translation, and questionnaire administration remain relevant even when both instruments exhibit adequate reliability. Wang and Cheng found useful psychometric performance in a sample of 544 adult gamers, alongside differences in measurement properties and explanatory associations. Good reliability means responses can cohere; it does not establish that the same proportion would receive the same clinical diagnosis. Wang and Cheng, 2020.
Problematic gaming, a positive screen, and confirmed disorder
Problematic gaming is often used for a broad range of behaviors and associated difficulties that can fall below a diagnostic threshold. A positive screening result means that answers exceed an instrument's defined cutoff. Clinical diagnosis requires an integrated assessment of symptom meaning, actual functional consequences, persistence, context, and alternative explanations. These three categories occupy different positions along a measurement process and should be reported separately. A screening prevalence can be useful for identifying people who may benefit from further assessment, but it is not identical to the prevalence of confirmed disease. King et al., 2020; World Health Organization, 2020.
A teenager experiencing a temporary surge in gaming during school vacation, an adult who plays with friends to manage loneliness, and a person whose gaming persistently displaces sleep, school, family obligations, and self-care may endorse some of the same survey statements for different reasons. A careful assessment asks what changed, what was sacrificed, how much control the individual retains, whether harm persists despite attempts to change, and whether symptoms are better explained by distress or another condition. Measurement decisions are therefore clinical decisions translated into epidemiological procedures.
Why prevalence estimates differ so much
The denominator: all residents, all gamers, or frequent gamers
Prevalence is usually calculated by dividing cases by people in the eligible population. If researchers survey only people who play games, the denominator excludes nongamers; if they recruit from online gaming communities, it may disproportionately represent enthusiastic or highly engaged players. Population surveys that include nongamers can produce a lower all-resident proportion even when the conditional prevalence among active gamers is unchanged. A paper reporting 5% of current gamers and another reporting 2% of all residents may describe compatible patterns rather than a conflict.
Eligibility rules also affect the denominator. Excluding people who have not played recently can remove people whose symptoms improved after stopping gaming, while including everyone regardless of any past gaming may reduce the observed prevalence of current gaming-related impairment. Studies need to identify whether they estimate current disorder, lifetime experiences, symptoms in the previous year, or symptoms in the preceding few weeks. Without that temporal definition, comparing prevalence numbers across studies is an exercise in comparing unlike quantities.
Sampling and self-selection
National probability samples, school-based surveys, convenience samples, gaming-forum recruitment, treatment clinics, and online panel surveys differ in who can be included and how likely they are to participate. A clinic cohort intentionally contains many symptomatic people and cannot estimate national population prevalence without a suitable sampling model. A voluntary online survey about gaming problems may attract respondents with especially strong concerns or unusual gaming exposure. Probability samples can support more credible population inference, although nonresponse, survey weighting, and exclusions still require scrutiny. Stevens et al., 2021; Kim et al., 2022.
Representative sampling is more than achieving a large number. A survey of 100,000 members of one gaming forum may have greater selection bias than a properly weighted random sample of several thousand residents. Large sample size narrows random uncertainty under an appropriate design; it does not repair a flawed recruitment frame, biased nonresponse, or ambiguous case definition. This helps explain why large contemporary surveys can report more precise but still incompatible estimates.
Measurement instruments and cutoff scores
Gaming disorder instruments vary in their items, response formats, reference periods, and criteria for classifying a case. Some ask whether gaming feels important, rewarding, or difficult to interrupt; others directly examine serious consequences and persistence despite harm. Researchers can obtain a higher prevalence by setting a low cutoff or counting any symptom endorsement, even when the underlying behavior of respondents is unchanged. The large tool-related heterogeneity found by Stevens and colleagues makes questionnaire choice an essential reporting item, not a minor methodological footnote. Stevens et al., 2021.
King and colleagues reviewed 32 screening and assessment tools used in 320 studies representing 462,249 participants. They identified inconsistent coverage of DSM and ICD criteria, especially continued gaming despite harm, and found no single unequivocally superior instrument across all psychometric and conceptual considerations. Several tools had stronger supporting evidence than others, but changing instruments without calibration obstructs comparison over time or between countries. King et al., 2020.
Sensitivity, specificity, and the base-rate problem
A scale with reasonably high sensitivity can still produce substantial false-positive classifications when the true condition is uncommon. Consider a hypothetical population of 10,000 people in which 2% genuinely meet carefully assessed criteria. A screen with 90% sensitivity and 90% specificity would flag 180 of the 200 true cases, but also 980 of the 9,800 people without the disorder, generating 1,160 positives. The apparent screen-positive prevalence would be 11.6%, and only about 15.5% of the positive screens would represent true cases under those assumptions. This calculation is an illustration of base rates, not a performance claim about any named gaming-disorder questionnaire.
Specificity becomes especially important when survey results are interpreted as clinical diagnoses. A small percentage of false positives among the much larger group without disorder can dominate the positive category. At the other end, a highly specific threshold can miss people with meaningful subthreshold harm. The appropriate cutoff depends on whether the intended use is broad prevention, clinical triage, research surveillance, or definitive diagnostic classification. Transparent prevalence work reports the purpose and error characteristics of the instrument, not merely its threshold score.
Recall periods and persistence
A question asking about experiences during the previous week can capture intense but temporary circumstances; a one-year window captures a different pattern; a lifetime question captures still another. The ICD-11 framework ordinarily expects a sustained course and significant impairment. A survey that never verifies duration can identify people with transient problems as though they have the persistent pattern required for clinical classification. Conversely, asking only about the past month may fail to characterize fluctuating difficulties in someone with a longer course. World Health Organization, 2020.
Translation and cross-cultural validity
Gaming is embedded in local language, family expectations, schooling structures, leisure alternatives, and access to digital technology. Terms such as withdrawal, control, neglect, excessive time, and conflict may acquire different connotations across translations. A scale that measures one latent construct in one country can operate differently elsewhere even when individual items look linguistically similar. Cross-cultural comparisons need evidence of measurement invariance, equivalent thresholds, and contextual interpretation, especially when regional prevalence contrasts are large. Wang and Cheng, 2020; Poon et al., 2026.
Differences in how cultures value family obligations and educational achievement can also affect reports of impairment. Such context must be evaluated rather than automatically used to dismiss distress. A pattern that genuinely damages sleep, relationships, school performance, or personal safety deserves consideration; the investigator must establish the relationship to gaming and the appropriateness of the instrument in the respondent's cultural setting.
The major gaming disorder scales and their interpretation
IGDS9-SF and DSM-oriented measures
The nine-item Internet Gaming Disorder Scale–Short Form, or IGDS9-SF, operationalizes the DSM Internet Gaming Disorder research framework. A study can use its total score, a prespecified symptom threshold, or a particular response rule. Those choices do not necessarily produce the same number of cases. An IGDS9-SF score is a questionnaire result. When papers use it to claim prevalence, readers should check the scale version, language, the response cutoff, and whether clinical impairment was independently assessed. Its widespread use helps enable comparisons only when procedures are harmonized. King et al., 2020; Wang and Cheng, 2020.
Gaming Disorder Test and ICD-11-oriented instruments
The Gaming Disorder Test, or GDT, was designed to reflect the ICD-11 clinical concept. It focuses on the core symptoms and consequences that anchor that classification. An ICD-11-oriented questionnaire may improve conceptual alignment with the diagnosis, but clinical validation, cutoff selection, population representativeness, and differential assessment remain necessary. A psychometric comparison found that the GDT and IGDS9-SF were both useful within its adult-gamer sample, while exhibiting differences in certain validity and invariance findings. Neither should be treated as an automatic diagnostic interview. Wang and Cheng, 2020.
Older addiction and problematic-use scales
Some epidemiological studies apply instruments originally designed to capture problematic internet use, generalized video-game involvement, or broad addiction-like symptoms rather than the particular ICD-11 disorder. Such tools may provide information about problematic engagement and associated harm, but folding their positive screens into a pool labeled gaming disorder increases construct heterogeneity. The systematic review of assessment tools documented that diagnostic criteria were not consistently covered across instruments. Each synthesis should disclose whether included tools measure an established clinical construct, a DSM research condition, or a wider behavioral-risk category. King et al., 2020.
What clinical interviews add
A structured interview can ask for concrete examples, verify the time course, separate ongoing functional loss from parental disapproval or temporary conflict, and examine alternative causes of distress. Informant reports, school or work functioning, sleep records, and clinical history may clarify discrepancies between endorsed symptoms and daily functioning. Interviews cost more than online surveys and cannot always be delivered to very large samples, but two-stage designs can combine broad screening with diagnostic assessment in a probability subsample. This produces stronger information about the gap between positive-screen prevalence and clinically confirmed prevalence.
How age changes the prevalence question
Children and adolescents
Child and adolescent estimates deserve their own denominators and developmental interpretation. Young people may experience stronger external constraints on access, parental rule changes, fluctuating social opportunities, and rapid shifts in gaming priorities during school transitions. Research samples often come from classrooms or online gaming communities, where selection affects who is represented. Higher estimates in adolescent meta-analyses can arise partly from real developmental vulnerabilities and partly from differences in tools, contexts, and cutoffs. The Satapathy and Barboza syntheses both show substantial variation that prevents applying a single pooled percentage to a particular school district or family. Satapathy et al., 2025; Barboza et al., 2026.
Parents and clinicians should interpret gaming behaviors by their consequences and trajectory. A teenager who stays up playing occasionally may need a conversation about sleep, routines, and responsibilities; a teenager who persistently cannot control gaming while grades, relationships, basic needs, or mental health deteriorate needs more thorough assessment. A reported national or international prevalence number does not diagnose the child in front of the clinician or establish that a particular genre, device, or game caused the difficulty.
Adults and older adults
Adult gaming spans students, employees, parents, professional players, and older adults. Studies focusing only on adolescents cannot be generalized to these groups without age-specific evidence. Adults may experience impairment in work, finances, parenting, or relationships, while retirees may have different patterns of leisure and social connection. Adult research can also miss nontraditional gamers when recruitment follows esports communities or popular platforms. Both Stevens and Kim identified age composition as a relevant source of prevalence variation. Stevens et al., 2021; Kim et al., 2022.
Sex, gender, and the limitations of subgroup estimates
Several meta-analyses report higher positive-screen rates among male respondents than female respondents. Stevens and colleagues estimated a male-to-female ratio of approximately 2.5 to 1 across their included studies. Kim and colleagues reported pooled male and female subgroup estimates of 8.5% and 3.5% under their own analytic specifications. These two summaries should not be merged as though they refer to the same denominators or identical measurement processes. Stevens et al., 2021; Kim et al., 2022.
Sex-related patterns can reflect differing gaming participation, game genres, cultural expectations, symptom expression, recruitment choices, and the measurement tools selected. The literature has historically emphasized male-dominated gaming settings, which can affect the ability to recognize difficulties among girls and women. Some surveys report binary sex categories without investigating gender diversity. Better epidemiology measures exposure to gaming as well as associated impairment, reports subgroup sample sizes, and tests whether scales behave similarly across populations rather than assuming a fixed biological difference.
Geographic differences: real variation and methodological noise
Reviews have described differences across regions, but a country is not a diagnostic instrument and geography alone is not an explanation. Economic conditions, access to broadband, school demands, cultural expectations, game markets, local regulation, and clinical awareness may contribute to differences in exposure and harm. The same comparisons are confounded when studies recruit different ages and use different questionnaires. Barboza and colleagues found especially broad geographical contrasts within their youth-focused evidence, alongside I² of 99.7%. At that degree of heterogeneity, apparent international rankings require substantial caution. Barboza et al., 2026.
A meaningful cross-country comparison would apply the same diagnostic algorithm, translations validated for equivalent meaning, comparable sampling frames, identical age bands, similar dates of data collection, and standardized methods for confirming impairment. Without these safeguards, a higher numerical estimate may indicate greater risk, easier detection, a lower score cutoff, or an unusually selected sample. Disentangling these explanations requires more than putting pooled point estimates into a league table.
Has gaming disorder become more common over time?
Gaming exposure and forms of digital participation have changed considerably, but the direction of disorder prevalence cannot be inferred from changing screen time or the release dates of studies. A review published in 2026 may include different countries, age bands, and diagnostic tools from a review published in 2021. Statistically significant year-related patterns within observational meta-regressions remain vulnerable to composition effects. They can generate a hypothesis about change without demonstrating an increase in a stable clinical condition. Satapathy et al., 2025; Poon et al., 2026.
The stronger design is repeated surveillance within comparable populations, with constant measures and diagnostic thresholds, stable inclusion criteria, appropriately weighted response rates, and sensitivity analyses for platform and demographic change. Researchers should distinguish an increase in public awareness and screening from an increase in genuine incident disorder. Reports also need to separate lifetime exposure, current symptoms, and onset of new cases, because these are different epidemiological quantities.
Prevalence, incidence, and persistence are different
A prevalence survey is a snapshot
Point prevalence counts people who meet a defined case rule at a given time. Period prevalence counts qualifying people across a specified interval. Lifetime prevalence asks whether criteria have been met at any earlier point. These definitions can generate very different percentages from the same population. Even the word current needs a stated recall period. Clinical studies, school surveys, and international meta-analyses sometimes use the term prevalence without equally transparent time specifications, which makes direct comparison hazardous.
Incidence measures new cases
Incidence asks how many people develop a new qualifying condition during a follow-up period among those initially at risk. Cross-sectional prevalence research cannot determine incidence because the timing of symptom onset is ordinarily unavailable. An increase in screen-positive prevalence could reflect more new cases, longer average duration, reduced recovery, or changes in recruitment or classification. Inferring rising incidence from a larger cross-sectional percentage is therefore a category error.
Longitudinal stability cannot be read as permanent illness
Sun, Ho, Kuss, and Cross synthesized longitudinal research on problematic gaming and gaming disorder. Their analyses reported that approximately 43%–45% of baseline threshold-positive participants remained above threshold at one-year follow-up, and approximately 34%–38% remained above threshold at two years, within the relevant study subsets. This literature consisted predominantly of adolescent groups and combined related but differently defined outcomes. The results demonstrate that symptom status can change and that dimensional severity may persist even when categorical screen status changes. They do not imply that every person falling below a cutoff has fully recovered, or that an individual probability of remaining clinically ill is exactly 43%. Sun et al., 2025.
Longitudinal assessments are also susceptible to attrition, developmental change, fluctuating school obligations, and the reliability of repeated self-report. A person close to a scale cutoff can cross it after a small score change without a comparably large change in well-being. Clinicians should attend to real functioning and the individual's broader context across repeated evaluations; public-health researchers should report both continuous severity and diagnostic-category transitions when their data allow.
Risk factors and correlates should not be treated as causal prevalence explanations
Depression, anxiety, ADHD traits, social isolation, sleep disruption, impulsivity, and family stress can co-occur with gaming-related difficulties, but the prevalence of these characteristics does not establish the prevalence of gaming disorder. Most prevalence studies are cross-sectional, so correlations cannot determine whether symptoms preceded gaming problems, resulted from them, or reflect shared contextual factors. A person may use games to cope with distress, while prolonged gaming-related disruption may also intensify distress. Testing direction requires longitudinal and intervention evidence that measures both conditions over time.
Reward learning, social reinforcement, motivations for competence and belonging, frustration, escape from stress, and design features of particular platforms can influence gaming patterns. Their existence does not mean that ordinary engagement is pathological or that a particular feature has been proven to cause a clinical diagnosis. To connect a mechanism to population prevalence, investigators need exposure measurement, a well-specified causal model, appropriate control for confounding, and validated case ascertainment. Mechanistic plausibility is a reason to investigate an effect, not a prevalence estimate.
Special consideration is needed for neurodevelopmental conditions. Focused interests, intense engagement, difficulties with transitions, or emotion-regulation challenges may influence questionnaire responses in people with ADHD or autism. Those features alone do not satisfy ICD-11 gaming disorder criteria. Clinical assessment should distinguish enduring individual traits, the actual benefits of a chosen activity, environmental demands, and impaired control accompanied by consequential harm. This avoids inflating case counts by mapping unrelated traits onto addiction-like wording.
What high heterogeneity means in a meta-analysis
A confidence interval and a prediction interval answer different questions
A confidence interval around a pooled proportion expresses uncertainty about the summary estimate under a statistical model. It does not describe the likely range of prevalence in every individual country or new study. A prediction interval, when estimated appropriately, instead considers how widely a future comparable study's true result could plausibly vary under the model. When study populations and instruments are heterogeneous, even a narrow confidence interval around the mean can coexist with a very broad distribution of underlying rates.
The 2026 Barboza review illustrates this distinction: its pooled estimate was 10% with a 95% confidence interval of 6%–15%, but the extreme heterogeneity and a much broader reported prediction range underscored that different study settings could yield very different rates. The scientifically appropriate message is variation requiring explanation, rather than precision about a universal 10% figure. Barboza et al., 2026.
I-squared is a warning signal, not an explanation
The I² statistic summarizes the fraction of observed variation in an analysis that is attributed to between-study heterogeneity under its model, rather than sampling error. Extremely high I² does not reveal which causes are responsible. Age, region, selection, measurement thresholds, and calendar period can all matter, and several may be correlated. Meta-regression estimates are observational relationships between study-level features and pooled results, vulnerable to ecological bias and unmeasured confounding. They do not demonstrate that changing one variable would causally change a person's risk.
Why averages of averages require special care
An umbrella review of meta-analyses can bring together systematic reviews that share dozens of the same primary studies. Counting every meta-analysis as independent evidence would effectively count the same participants more than once. Likewise, taking a median of published meta-analytic percentages produces a descriptive statistic about the literature, not a representative population statistic. Poon and colleagues' 6.20% umbrella-review median therefore complements the underlying meta-analyses as a map of research dispersion; it cannot replace age-specific, method-specific population estimates. Poon et al., 2026.
Interpreting rates in real-world decisions
For families and individuals
A prevalence estimate describes a group, while assessment considers one person's pattern of life. Useful questions concern control over gaming, recurring harmful consequences, whether gaming repeatedly outranks essential commitments, persistence of the pattern, and changes in sleep, mood, learning, relationships, and work. The goal is to identify genuine impairment and provide appropriate support. Heavy gaming alone is a poor diagnostic shortcut, and a positive online quiz is a prompt for context-sensitive evaluation rather than a self-contained clinical verdict. World Health Organization, 2020; King et al., 2020.
For clinicians and schools
Screening can identify potential needs, but its positive predictive value depends on the prevalence of disorder in the screened population and on instrument performance. School screening programs should plan confirmatory pathways and minimize stigma, particularly when interpreting symptoms in children. Clinical interviews should assess time course, impairment, comorbidity, protective factors, and alternative explanations. Schools and healthcare providers can address sleep, peer relationships, distress, and family communication even when a full gaming disorder diagnosis is not established.
For researchers and policymakers
Public-health reporting should state the exact target population, participation rate, recruitment method, time frame, diagnostic reference, instrument, cutoff, and whether interviews verified screen positives. Studies should supply both the numerator and denominator, subgroup counts, confidence intervals, weighting procedures, and sensitivity analyses. Countries considering service investment can use prevalence evidence alongside severity, treatment access, functioning, and burden of harm; a single headline percentage cannot substitute for needs assessment.
A practical checklist for reading any gaming disorder prevalence study
First, identify who could enter the sample: the general population, people who play games, students, clinical patients, or an online gaming community. Second, check the age range, country, calendar year, response rate, and weighting. Third, determine whether the study measured ICD-11 gaming disorder, DSM-oriented Internet Gaming Disorder, or a broader concept of problematic gaming. Fourth, inspect the exact questionnaire or clinical interview and its cutoff. Finally, ask whether impairment and persistence were directly established, whether positive screens received diagnostic confirmation, and how large the between-study uncertainty was when results were pooled.
These checks determine whether a percentage can answer a particular question. A school survey of adolescent questionnaire positives can inform adolescent screening policy. It cannot alone establish the proportion of all US adults who satisfy ICD-11 diagnostic criteria. A large cross-country meta-analysis can summarize a literature base. It cannot reliably rank national prevalence when included surveys do not ask equivalent questions. Good interpretation begins with the original study's operational definition.
Frequently asked questions
What percentage of the population has gaming disorder?
There is no single universally valid percentage. Broad international meta-analyses found approximately 3.05% and 3.3% using their respective study sets, with estimates around 1.96% and 2.4% in more strictly sampled subsets. Adolescent-focused reviews have reported considerably higher pooled screen-positive proportions. Each number refers to its own sampling and measurement assumptions, not a universal rate of clinician-confirmed ICD-11 diagnoses. Stevens et al., 2021; Kim et al., 2022; Satapathy et al., 2025.
Does one in ten teenagers have a gaming disorder?
One 2026 systematic review reported 10% across its selected child-and-adolescent studies. Extreme heterogeneity (I² 99.7%), variable instruments, and population differences make this a statistical review result rather than a universal statement about all teenagers. A different large adolescent review reported 8.6%, also with extreme heterogeneity. Neither number means that one in ten teenagers in a particular school has received or should receive a clinical diagnosis. Barboza et al., 2026; Satapathy et al., 2025.
Is the DSM-5-TR Internet Gaming Disorder category the same as ICD-11 gaming disorder?
They are related research and clinical frameworks with different symptom organization and status. ICD-11 includes gaming disorder as a recognized diagnosis with a pattern centered on loss of control, prioritization, continuation despite harm, and significant impairment. The DSM framework places Internet Gaming Disorder among conditions for further study and describes more numerous research criteria. Studies operationalizing these frameworks may classify different respondents. World Health Organization, 2020; Wang and Cheng, 2020.
Can someone game for many hours without having gaming disorder?
Yes. Time spent playing is contextual information, not a diagnostic criterion sufficient by itself. A clinical concern arises when control is impaired and gaming persistently takes priority over important activities despite substantial negative consequences, causing significant functional impairment. The relevant questions concern behavior, impact, and course. Intensive recreation, professional practice, and temporary periods of increased play require evaluation within their actual circumstances. World Health Organization, 2020.
Does a positive gaming-addiction questionnaire mean a diagnosis?
No. A positive questionnaire result means the responses met an instrument-specific rule. Screening can be useful for triage and research, but questionnaires differ in content and cutoffs, and they can classify people whose symptoms do not satisfy clinical criteria. An informed clinical assessment examines duration, functional impairment, context, distress, and differential explanations. King et al., 2020.
Are gaming disorder rates rising?
Some study-level analyses have described calendar-time associations, but comparing pooled percentages from publications that examine different populations and use different instruments does not demonstrate a genuine change in population incidence or prevalence. Reliable trend estimation needs comparable repeated population samples and stable methods. Rapid changes in gaming technology do not supply a measured clinical trend on their own. Satapathy et al., 2025; Poon et al., 2026.
Do most screen-positive teenagers stay symptomatic?
Longitudinal evidence indicates variable stability rather than a fixed inevitable course. A 2025 review reported that approximately 43%–45% remained above a problem-gaming or gaming-disorder threshold after one year and about 34%–38% after two years in the relevant study sets. These figures include diverse definitions and predominantly young participants. Movement across a cutoff is not identical to verified clinical recovery or progression. Sun et al., 2025.
Scientific conclusion
Gaming disorder prevalence is a property of a clearly defined population assessed by a clearly defined method. International systematic reviews consistently identify a meaningful minority with gaming-related symptoms and impairment, but their pooled rates shift substantially when researchers change age bands, recruitment methods, instruments, diagnostic frameworks, and case thresholds. The best interpretation gives priority to representative samples, valid measurement of impairment, explicit diagnostic classification, and transparent uncertainty. This allows prevalence research to inform effective prevention and clinical care while preserving the distinction between healthy gaming, problematic patterns, screening results, and established disorder. Stevens et al., 2021; King et al., 2020; Poon et al., 2026.
Related Articles
Gaming Disorder: Symptoms, Diagnosis, and Treatment — The clinical framework, core symptoms, ICD-11 diagnostic concepts, and treatment context.
Gaming Disorder Screening: Questionnaires and Clinical Assessment — The purpose and limitations of screening scores, assessment tools, and clinical interviews.
Heavy Gaming vs Gaming Disorder: Warning Signs and Functional Impairment — Why hours of play and strong enthusiasm need to be interpreted alongside control and actual harm.
Internet Gaming Disorder in DSM-5-TR vs Gaming Disorder in ICD-11 — The classification frameworks that shape research case definitions and estimates.
Behavioral Addictions: Recognized Diagnoses and Scientific Debate — The wider scientific context of recognized disorders and proposed behavioral addictions.
References
Barboza, J. J., Bonilla Asalde, C., Rivera-Lozada, O., et al. (2026). Global prevalence of internet gaming disorder in children and adolescents: A systematic review and meta-analysis. Child and Adolescent Psychiatry and Mental Health, 20, 90. https://doi.org/10.1186/s13034-026-01083-8
Kim, H. S., et al. (2022). Prevalence of gaming disorder: A meta-analysis. Addictive Behaviors, 126, 107183. https://doi.org/10.1016/j.addbeh.2021.107183
King, D. L., Chamberlain, S. R., Carragher, N., et al. (2020). Screening and assessment tools for gaming disorder: A comprehensive systematic review. Clinical Psychology Review, 77, 101831. https://doi.org/10.1016/j.cpr.2020.101831
Poon, E. C. S., Soh, X. C., Poh, T. J. H., Tan, A. C. S., and Hartanto, A. (2026). Global prevalence of Internet Gaming Disorder: An umbrella review of meta-analytic evidence and implications for child and adolescent psychiatry. Brain Sciences, 16(7), 728. https://doi.org/10.3390/brainsci16070728
Satapathy, P., Khatib, M. N., Balaraman, A. K., et al. (2025). Burden of gaming disorder among adolescents: A systemic review and meta-analysis. Public Health in Practice, 9, 100565. https://doi.org/10.1016/j.puhip.2024.100565
Stevens, M. W. R., Dorstyn, D., Delfabbro, P. H., and King, D. L. (2021). Global prevalence of gaming disorder: A systematic review and meta-analysis. Australian and New Zealand Journal of Psychiatry, 55(6), 553–568. https://doi.org/10.1177/0004867420962851
Sun, A.-P., Ho, C.-H., Kuss, D. J., and Cross, C. L. (2025). The temporal stability of problematic gaming and gaming disorder: A systematic review and meta-analysis. Addictive Behaviors Reports, 21, 100592. https://doi.org/10.1016/j.abrep.2025.100592
Wang, H.-Y., and Cheng, C. (2020). Psychometric evaluation and comparison of two gaming disorder measures derived from the DSM-5 and ICD-11 frameworks. Frontiers in Psychiatry, 11, 577366. https://doi.org/10.3389/fpsyt.2020.577366
World Health Organization. (2020). Addictive behaviours: Gaming disorder. https://www.who.int/news-room/questions-and-answers/item/addictive-behaviours-gaming-disorder
World Health Organization. (n.d.). Gaming disorder: Frequently asked questions. https://www.who.int/standards/classifications/frequently-asked-questions/gaming-disorder