Journal of Clinical Medicine, volume 8, issue 10, article 1730, 18 October 2019
- Open access
- 3,377 Polish gamers
- Online survey
- Conditional inference tree
- MIMIC model
- Poland
- Open access
Free full text: the open access article (CC BY 4.0), re-typeset as a clean PDF with the same content as the published version. Please cite the published version.
- internet gaming disorder
- video games
- gaming addiction
- problematic gaming
- behavioural addictions
- gaming disorder
- diagnostic criteria
- DSM-5
- ICD-11
- withdrawal
- loss of control
- preoccupation
- negative consequences
- clinical weighting of criteria
- diagnosis
- Internet Gaming Disorder Scale–Short-Form (IGDS9-SF)
- Polish IGDS9-SF
- conditional inference tree (Ctree)
- machine learning
- decision tree
- MIMIC model
- confirmatory factor analysis
- online survey
- Poland
Not all nine criteria for Internet gaming disorder carry the same weight: withdrawal and loss of control matter most. In a decision-tree analysis of 3,222 Polish gamers, those who very often felt withdrawal and loss of control had a 78% probability of disordered gaming, which questions the WHO’s choice to leave withdrawal out of gaming disorder.
Key findings
- Withdrawal is the first and strongest split. In a conditional inference tree on the nine IGD criteria (N = 3,222), very frequent withdrawal symptoms divided gamers first: 32.46% of those reporting withdrawal ‘very often’ were disordered gamers, against 0.19% of the others (Pontes et al., 2019).
- Withdrawal plus loss of control marks most disordered gamers. Gamers who very often experienced both withdrawal and loss of control had a 77.77% probability of disordered gaming; this ‘impaired self-control’ pathway held 21 of the 31 disordered gamers (Pontes et al., 2019).
- Four criteria carried the diagnostic information. Withdrawal, loss of control, negative consequences and preoccupation were the criteria that defined five gamer subtypes, from ‘healthy’ (95% of gamers) to ‘preoccupied’, ‘low risk’, ‘harmful’ and ‘impaired self-control’ (Pontes et al., 2019).
- The Polish IGDS9-SF works well. The Polish Internet Gaming Disorder Scale–Short-Form was unidimensional (CFI = 0.95, RMSEA = 0.04) and reliable (α = 0.82), and scores were higher in younger gamers, men and those who played more (Pontes et al., 2019).
At a glance
- Design
- Cross-sectional online survey (Qualtrics) of gamers recruited through the three largest Polish gaming forums (gry-online.pl, gamesboard.pl, gamesfanatic.pl) and their social media channels; no incentives
- Participants
- 3,377 gamers who had played in the past year (82.7% men; mean age 20, SD 4.3, range 12–49); 3,222 retained after excluding missing data
- Measure
- Polish Internet Gaming Disorder Scale–Short-Form (IGDS9-SF; nine DSM-5 criteria, 1 = never to 5 = very often), back-translated and piloted with 52 gamers; a criterion counted as endorsed when answered ‘very often’
- Disordered gaming
- Five or more of nine criteria endorsed: 31 gamers (0.96%)
- Analysis
- MIMIC model with age, gender and weekly gaming time (FIML, 5,000 bootstrap samples) and a conditional inference tree (partykit, minimum split 50 cases), in R
- Model fit
- χ²(51) = 407.48, CFI = 0.95, TLI = 0.94, RMSEA = 0.04, SRMR = 0.02; Cronbach’s α = 0.82, composite reliability = 0.86
Figure 2. Conditional inference tree on the nine IGD criteria
Probability of disordered gaming in each gamer subtype
Tables
Table 1. The nine IGD criteria as measured with the IGDS9-SF, their clinical criteria and inclusion in ICD-11 gaming disorder
| Item | IGDS9-SF item wording | Clinical criterion | In ICD-11? |
|---|---|---|---|
| 1 | Do you feel preoccupied with your gaming behavior? (Some examples: Do you think about previous gaming activity or anticipate the next gaming session? Do you think gaming has become the dominant activity in your daily life?) | Preoccupation | No |
| 2 | Do you feel more irritability, anxiety or even sadness when you try to either reduce or stop your gaming activity? | Withdrawal | No |
| 3 | Do you feel the need to spend increasing amount of time engaged gaming in order to achieve satisfaction or pleasure? | Tolerance | No |
| 4 | Do you systematically fail when trying to control or cease your gaming activity? | Loss of control | Yes |
| 5 | Have you lost interests in previous hobbies and other entertainment activities as a result of your engagement with the game? | Giving up other activities | Yes |
| 6 | Have you continued your gaming activity despite knowing it was causing problems between you and other people? | Continuation | Yes |
| 7 | Have you deceived any of your family members, therapists or others because the amount of your gaming activity? | Deception | No |
| 8 | Do you play in order to temporarily escape or relieve a negative mood (e.g., helplessness, guilt, anxiety)? | Escape | No |
| 9 | Have you jeopardized or lost an important relationship, job or an educational or career opportunity because of your gaming activity? | Negative consequences | Yes |
Table 2. Summary of the Conditional Inference Tree-derived rules alongside each IGD criteria endorsement pathway
| Rule | First split | Second split | Third split | Gamer subtype |
|---|---|---|---|---|
| 1 | ‘withdrawal’ ≤ 4; 0.19% (95% CI 0.03–0.34) | ‘preoccupation’ ≤ 4; 0% (95% CI 0–0), n = 3061 | ‘Healthy’ | |
| 2 | ‘withdrawal’ ≤ 4; 0.19% (95% CI 0.03–0.34) | ‘preoccupation’ = 5; 7.14% (95% CI 1.63–12.65), n = 84 | ‘Preoccupied’ | |
| 3 | ‘withdrawal’ = 5; 32.46% (95% CI 22.01–42.92) | ‘loss of control’ ≤ 4; 8.00% (95% CI 0.48–15.51) | ‘negative consequences’ ≤ 3; 0% (95% CI 0–0), n = 35 | ‘Low Risk’ |
| 4 | ‘withdrawal’ = 5; 32.46% (95% CI 22.01–42.92) | ‘loss of control’ ≤ 4; 8.00% (95% CI 0.48–15.51) | ‘negative consequences’ > 3; 26.66% (95% CI 4.28–49.04), n = 15 | ‘Harmful’ |
| 5 | ‘withdrawal’ = 5; 32.46% (95% CI 22.01–42.92) | ‘loss of control’ = 5; 77.77% (95% CI 62.09–93.45), n = 27 | ‘Impaired Self-Control’ |
Answers to each criterion: 1 = ‘Never’, 2 = ‘Rarely’, 3 = ‘Sometimes’, 4 = ‘Often’, 5 = ‘Very Often’. Endorsement was operationalised as ‘Very Often’. Percentages are the share of disordered gamers in each node. The article’s table title gives N = 3377; the node sizes add up to 3,222, the analysed sample.
The question
The DSM-5 lists nine criteria for Internet gaming disorder and requires five, treating all criteria as equal, while the WHO’s gaming disorder leaves several of them out, including withdrawal. We asked whether some criteria carry more diagnostic weight than others, using a machine-learning decision tree on a large sample of Polish gamers.
Implications
Practical implications
Clinicians and the authors of diagnostic manuals may give withdrawal and loss of control particular weight when assessing disordered gaming. Because withdrawal was central to identifying disordered gamers, the WHO criteria for gaming disorder, which omit it, may be laxer than the DSM-5 criteria and could hinder efficient diagnosis.
Theoretical implications
The nine IGD criteria differ in clinical weight, supporting the distinction between core and peripheral criteria, and treating each criterion as a continuous rather than binary symptom reveals distinct pathways and gamer subtypes.
Limitations and future research
- The sample was a community-based, non-probability sample of gamers, not a clinical sample.
- The cross-sectional design does not allow causal conclusions.
- Disordered gaming was classified psychometrically; a clinical gold standard such as structured interviews is needed.
Questions this paper answers
Which symptoms best identify gaming disorder?
In 3,222 Polish gamers, withdrawal and loss of control were the most informative: gamers experiencing both very often had a 78% probability of disordered gaming (Pontes et al., 2019).
Are all Internet gaming disorder criteria equally important?
No. A decision-tree analysis found that withdrawal, loss of control, negative consequences and preoccupation carried the diagnostic information, while the other five criteria did not shape the tree (Pontes et al., 2019).
Why does withdrawal matter for the WHO definition of gaming disorder?
The WHO’s ICD-11 gaming disorder does not include withdrawal, yet withdrawal was the first split in the tree, so the authors argue that the WHO criteria may be laxer than the DSM-5 criteria.
What is a conditional inference tree?
A machine-learning method that repeatedly splits a sample on the predictor most strongly associated with the outcome, producing if–then rules and subgroups.
Key concepts
- Internet gaming disorder (IGD)
- A DSM-5 condition for further study, defined by at least five of nine criteria within 12 months.
- Conditional inference tree (Ctree)
- A non-parametric decision-tree method that uses permutation tests to choose splits, avoiding selection bias.
- Withdrawal
- Irritability, anxiety or sadness when trying to reduce or stop gaming.
- MIMIC model
- A factor model with covariates (multiple indicators, multiple causes), used here to test the scale and its links with age, gender and gaming time.
Open science and materials
- Full text (PDF) Open access article re-typeset as a clean PDF, including the tables and figures
- Published version Journal of Clinical Medicine (open access)
- Conference abstract Earlier version presented at the 5th International Conference on Behavioral Addictions (ICBA 2018, Cologne): Pontes, Schivinski, Brzozowska-Woś & Griffiths, Journal of Behavioral Addictions, 7(Suppl. 1), 130, abstract OP-147
Study materials
- Scale items
- The nine IGDS9-SF items, the DSM-5 criterion each measures and whether it is part of ICD-11 gaming disorder are in Table 1 below.
How to cite
Pontes, H. M., Schivinski, B., Brzozowska-Woś, M., & Stavropoulos, V. (2019). Laxer clinical criteria for gaming disorder may hinder future efforts to devise an efficient diagnostic approach: A tree-based model study. Journal of Clinical Medicine, 8(10), Article 1730. https://doi.org/10.3390/jcm8101730
Open access under a Creative Commons CC BY 4.0 licence.