Problem gambling means gambling that is causing harm to you or to the people around you. It is not, despite how often it is used, a medical diagnosis. Doctors diagnose something narrower called gambling disorder, and researchers measure something broader with a questionnaire called the PGSI. Understanding what is problem gambling means understanding that these three things are not the same, and that mixing them up is how newspapers end up printing numbers that are wrong by a factor of six.
This page explains each one in plain language: what the PGSI score means, what the DSM-5 and ICD-11 actually require for a diagnosis, and why two respected British surveys give very different answers to the same question.
Written from published clinical sources. This article has not been reviewed by a clinician — see our editorial policy. Nothing here is medical advice.
“Problem gambling” is not a diagnosis
This surprises people. The phrase appears in government reports, in charity names and in newspaper headlines, but it does not appear as a diagnosis in either of the two manuals doctors use.
It is a public health and research term. The UK Government’s own evidence review glossary defines problem gambling as gambling “to an extent that compromises, disrupts or damages family, personal or recreational pursuits”. That is a description of a life going wrong. It is not a clinical threshold.
The same glossary defines harmful gambling more widely still, as “any frequency of gambling that results in people experiencing harm”. Most gambling harm in a population happens to people who would never meet a clinical threshold. There are simply far more of them.
The PGSI: a screen, not a diagnosis
The Problem Gambling Severity Index comes from the Canadian Problem Gambling Index, developed by Ferris and Wynne and published in 2001. It is the most widely used gambling measure in the world.
It asks nine questions about the past 12 months. Each answer scores from 0 to 3: never is 0, sometimes is 1, most of the time is 2, almost always is 3. That gives a total between 0 and 27. The Gambling Commission sets out the scoring in full.
| Score | Usual label | What it means |
|---|---|---|
| 0 | No reported problems | May gamble, even heavily, but reports none of the nine symptoms |
| 1–2 | Low risk | Some symptoms reported, few or no consequences |
| 3–7 | Moderate risk | More symptoms and some negative consequences |
| 8–27 | Problem gambling | Symptoms plus loss of control and identifiable harm |
The Gambling Commission has moved away from labelling people and towards describing the gambling. A person is not “a problem gambler”; they have a PGSI score of 8 or more.
What the PGSI does not tell you
This is the part that gets misreported constantly. A PGSI score of 8 is not a diagnosis of addiction. The regulator that publishes most of Britain’s gambling statistics says so directly in its guidance on using those statistics. Under the heading “How the GSGB should not be used”, the first item is: “As a measure of addiction to gambling.”
The same guidance says: “The PGSI should not be confused with a measure of gambling addiction.” So when you see a headline converting a percentage into a number of “gambling addicts”, that headline is doing something the people who collected the data have explicitly asked nobody to do.
A screening questionnaire is a way of sorting a large population quickly. A diagnosis needs a clinician, a conversation and time.
Gambling disorder in the DSM-5
The DSM is the American Psychiatric Association’s diagnostic manual. Its current edition is the text revision, DSM-5-TR, published in 2022; the criteria for gambling were set in DSM-5 in 2013.
That 2013 edition made a change that mattered. Gambling was moved out of the impulse-control chapter and placed in the chapter on substance-related and addictive disorders. The APA explains in its own summary of changes that this was done because gambling activates the brain’s reward system in ways that resemble drugs of abuse. It was the first time a behaviour, rather than a substance, was classed as an addictive disorder.
A diagnosis requires four or more of nine criteria within a 12-month period, together with clinically significant distress or impairment. The APA’s patient-facing page confirms the four-in-a-year threshold. Broadly, the nine cover:
- Needing to gamble more money to get the same excitement.
- Feeling restless or irritable when trying to cut down.
- Repeated failed attempts to control, cut back or stop.
- Being preoccupied with gambling.
- Gambling when feeling distressed.
- Returning another day to win back losses, known as chasing.
- Lying to hide how much you gamble.
- Risking or losing a relationship, job or opportunity because of gambling.
- Relying on others for money to deal with gambling debts.
Severity is then graded by how many criteria are met: mild is four to five, moderate is six to seven, and severe is eight to nine. A US federal comparison table sets out these bands alongside the older DSM-IV version.
What changed from DSM-IV
Three things, and they all matter when you read older research.
The name changed. DSM-IV called it “pathological gambling”. DSM-5 calls it “gambling disorder”, which is less loaded.
The threshold dropped. DSM-IV had ten criteria and needed five. DSM-5 has nine and needs four.
And one criterion was deleted: the DSM-IV item about committing illegal acts such as forgery, fraud or theft to finance gambling. Crime is a consequence of gambling harm, not a symptom of the disorder, and keeping it as a criterion mixed the two up. Because of these changes, prevalence figures based on DSM-IV and DSM-5 are not directly comparable.
Gambling disorder in the ICD-11
The ICD is the World Health Organization’s classification, used by health systems worldwide. In ICD-11, gambling disorder has the code 6C50 and sits among “disorders due to addictive behaviours”.
ICD-11 does not count criteria. It describes a pattern. The WHO’s own fact sheet on gambling sets out the three diagnostic requirements: impaired control over gambling; increasing priority given to gambling, so that it takes precedence over other interests and daily activities; and continuation of gambling despite negative consequences. The behaviour must also be severe enough to cause real distress or damage to the person’s life.
ICD-11 does one thing the DSM does not. It codes where the gambling happens:
- 6C50.0 — gambling disorder, predominantly offline.
- 6C50.1 — gambling disorder, predominantly online.
- 6C50.Z — gambling disorder, unspecified.
That distinction exists because online gambling is continuously available, faster, and harder for anyone else to see. The codes can be looked up in the WHO ICD-11 browser.
The three compared
| PGSI | DSM-5 | ICD-11 | |
|---|---|---|---|
| What it is | Screening questionnaire | Diagnostic manual | Global health classification |
| Who uses it | Researchers, regulators, surveys | Clinicians, mainly in the US | Health systems worldwide |
| How it works | Score 0–27 across 9 items | 4 or more of 9 criteria in 12 months | 3 required features, described not counted |
| Name used | Problem gambling (score 8+) | Gambling disorder | Gambling disorder (6C50) |
| Severity grading | Four bands by score | Mild, moderate, severe by criteria count | No criteria-count bands |
| Online vs offline | Not distinguished | Not coded | Coded separately |
| Can it diagnose? | No | Yes | Yes |
Why two British surveys say 2.4% and 0.4%
This is the clearest example of why the detail matters, and it trips up almost everyone who writes about gambling.
The Gambling Commission’s Gambling Survey for Great Britain, Annual Report 2025, published on 16 July 2026, found that 2.4% of adults had a PGSI score of 8 or more (95% confidence interval 1.9 to 2.8). Among people who had gambled in the past 12 months the figure was 4.0%.
The Adult Psychiatric Morbidity Survey 2023/24, published on 27 November 2025, used the same nine-question PGSI and found 0.4% of adults scored 8 or more.
Same instrument, same country, roughly the same period, six times the difference. Neither survey is wrong. The gap is caused by how the two were run.
| GSGB 2025 | APMS 2023/24 | |
|---|---|---|
| PGSI 8+ | 2.4% | 0.4% |
| Published | 16 July 2026 | 27 November 2025 |
| Recruitment | Push-to-web, with a postal option | Interviewer-recruited, in-home |
| How PGSI was answered | Self-completed | Self-completed |
| Population | Adults 18+, Great Britain | Adults 16+, England |
| Sample | 20,775 | 6,912 |
Note the row that most write-ups get wrong. It is often said that the difference is that one survey was filled in privately and the other was read out by an interviewer. That is not so. In both surveys the gambling questions were answered by the respondent privately. In APMS the gambling module sits in the self-completion part of the interview, precisely because the survey covers sensitive subjects.
The real difference is in who ends up taking part. An interviewer knocking on a randomly chosen door gets a broad cross-section. A survey that invites people online tends to attract people with an interest in the subject, and people who gamble a lot have more reason to respond to a gambling survey. That self-selection pushes the estimate up.
This is not our theory. The independent review of the GSGB by Professor Patrick Sturgis of the LSE called the survey’s development “exemplary in all respects” while warning that policymakers should be “mindful to the fact there is a non-negligible risk that they substantially over-state the true level of gambling and gambling harm in the population”. The Office for Statistics Regulation review, published on 22 May 2025, criticised the Commission for leaving that warning out of the statistical releases themselves.
The practical rule is simple. Never put a figure from one survey next to a figure from another and call the difference a trend. The Gambling Commission’s guidance says its statistics should not be used “to provide direct comparisons with results from other gambling or health surveys”. Compare GSGB with GSGB, and APMS with APMS.
Other screens you may come across
The PGSI is the most common measure, but it is not the only one, and older studies often used something else. Knowing which was used tells you how much weight a figure can carry.
- DSM-IV screens. Many surveys before roughly 2015 used a questionnaire based on the old DSM-IV criteria. Because DSM-IV used ten criteria and a threshold of five, those results cannot be lined up with DSM-5 results.
- SOGS. The South Oaks Gambling Screen dates from 1987 and is still used, especially in Asia. Some versions ask about your whole life rather than the past year, which produces much higher numbers that are not comparable with past-year measures.
- Short screens. Very brief tools exist for busy settings such as a GP surgery, where the aim is only to decide whether a longer conversation is needed.
None of these is a diagnosis either. They all do the same job: sort quickly, then refer. The point of a screen is to start a conversation, not to end one.
Whenever you read a gambling statistic, four details tell you whether you can trust the comparison you are about to make: the survey’s name, the year, the instrument used, and how people were recruited and answered. If any of the four is missing, treat the number with care.
The words people use, and why they changed
Language in this field has shifted noticeably over the past decade, and the shift is deliberate.
| Older phrase | Preferred now | Why |
|---|---|---|
| Problem gambler | Person experiencing gambling harm | Describes the situation, not the person’s identity |
| Pathological gambling | Gambling disorder | The DSM renamed it in 2013; less stigmatising |
| Gambling addict | Person with gambling disorder | “Addict” is not a clinical term and follows people around |
| Responsible gambling | Safer gambling, or gambling harm reduction | Puts less of the burden on the individual |
This is not word policing for its own sake. The old language put the blame squarely on one person’s character, which made asking for help harder and let the design of the products off the hook. Research on the wider harms framework, and the statistical reviews that have followed, have pushed the field towards describing what is happening rather than labelling who it is happening to.
You will still see the old terms everywhere, including in the names of long-established charities and in the name of the PGSI itself. That is history, not disagreement.
Why any of this matters to you
If you are reading this because you are worried about your own gambling, the labels are largely beside the point. You do not need a score of 8. You do not need four criteria. You do not need anyone to call it a disorder.
Harm is not a switch that flips at a threshold. Most of the damage gambling does in a country happens to people well below any clinical line, as the World Health Organization notes. Someone with a PGSI score of 3 who is lying to their partner about money is being harmed, whatever the questionnaire calls them.
The thresholds exist so that services can be planned and research can be compared. They were never meant to be a test you have to fail before you are allowed to ask for help.
If you recognise yourself in this
A few things are worth knowing. Gambling disorder is treatable, and the treatments have decent evidence behind them. Self-exclusion schemes work better than willpower alone, and our guide to self-exclusion explains how to register in your country. Our article on how gambling addiction develops covers the signs and the treatment options, and our step-by-step guide to getting help sets out what to do first.
You can also talk to someone free and in confidence today. In the United Kingdom the National Gambling Helpline is 0808 8020 133, open 24 hours a day. In the United States the National Problem Gambling Helpline is 1-800-MY-RESET (1-800-697-3738), also 24 hours. In Australia, Gambling Help Online is 1800 858 858. Numbers and opening hours for 25 countries are in our global directory of gambling helplines.
If you want to understand the wider framework these terms sit inside, start with our introduction to responsible gambling.
Last reviewed: 23 September 2026. Sources checked on this date.