About the CAGE Questionnaire for Alcohol Use
CAGE is a four-question screen for problem drinking — Cutting down, Annoyance at criticism, Guilt, and Eye-openers — each scored yes (1) or no (0), for a total of 0 to 4. Two or more positive answers is the conventional threshold for concern, but a large diagnostic meta-analysis found its accuracy at that threshold varies a great deal by setting: pooled sensitivity was 0.87 in hospital inpatients but only 0.60 in ambulatory patients. CAGE was built to detect lifetime alcohol dependence, not current hazardous drinking, and a head-to-head study found every patient CAGE identified was also identified by AUDIT — CAGE is the narrower of the two tools, not a faster version of the same one.
Formula
CAGE = count of 'yes' answers across the four questions (0–4)- each question
- Scored 1 for yes, 0 for no. No weighting between items.
- There is no built-in time frame — the questions ask 'have you ever', so a positive answer can reflect a distant episode rather than current drinking.
- Two or more positive answers is the conventional cutoff, but the questionnaire's own diagnostic accuracy at that cutoff differs substantially between inpatient and ambulatory settings.
Interpreting the result
A score of 0 or 1 does not rule out a drinking problem — the original derivation work and later reviews are explicit that CAGE is not sensitive enough at a strict 4-item or even a 1-item cutoff to be reassuring on its own, and a single positive answer still warrants a conversation. A score of 2 or more is conventionally treated as screen-positive and should prompt a full alcohol history covering quantity, pattern and withdrawal symptoms, not a diagnosis by itself. Because CAGE's accuracy is setting-dependent — considerably better in hospital inpatients than in ambulatory or primary care patients in the pooled evidence — a negative CAGE in an outpatient setting is weaker reassurance than the same result on a hospital ward.
| Score | Band | What it means | Action |
|---|---|---|---|
| 0 | No positive responses | Below the conventional threshold, but does not exclude hazardous drinking | Consider AUDIT if current consumption, rather than lifetime dependence, is the question |
| 1 | One positive response | Below the conventional 2-point threshold | Still warrants a conversation about drinking — CAGE misses hazardous use that has not yet caused conflict |
| 2–4 | Screen positive | Conventional threshold for concern; pooled sensitivity for alcohol abuse/dependence ranged from 0.60 (ambulatory) to 0.87 (inpatient) at this cutoff | Take a full alcohol history, assess for organ damage and withdrawal risk, and offer or refer for treatment |
Scroll the table sideways for every column.
What the CAGE needs (4 inputs)
- Cut down
- Have you ever felt you should cut down on your drinking? Yes = 1 point.
- Annoyed
- Have people annoyed you by criticising your drinking? Yes = 1 point.
- Guilty
- Have you ever felt bad or guilty about your drinking? Yes = 1 point.
- Eye-opener
- Have you ever had a drink first thing in the morning to steady your nerves or get rid of a hangover? Yes = 1 point.
What it returns
- CAGE score (0–4)
- Count of yes answers.
- Screen result
- Below threshold, or screen positive (2 or more), as conventionally applied.
How it is calculated
CAGE was constructed from clinical interview questions found to discriminate patients with alcoholism from those without it in a psychiatric inpatient population, then popularised as a four-letter mnemonic for use in general clinical practice. It works as a lifetime-history screen rather than a current-consumption measure: each item probes a consequence of drinking that tends to appear only once dependence has developed (attempts to cut down, others' criticism, guilt, morning drinking to relieve withdrawal), which is why it performs well at detecting dependence but poorly at detecting hazardous drinking that has not yet produced these consequences.
Facts & figures
| Setting | Pooled sensitivity |
|---|---|
| Inpatients | 0.87 |
| Primary care patients | 0.71 |
| Ambulatory patients | 0.60 |
Scroll the table sideways for every column.
From a meta-analysis of 10 studies against DSM criteria as the reference standard (pooled AUC 0.87, 95% CI 0.85–0.89; LR+ 3.44, LR− 0.18). The authors concluded the CAGE is of limited diagnostic value for screening purposes at its recommended ≥2 cutoff — accuracy depends heavily on where and in whom it is used.
| Tool | Identified as hazardous/harmful drinkers |
|---|---|
| AUDIT | 36% |
| CAGE | 22% |
Scroll the table sideways for every column.
Every patient identified by CAGE was also identified by AUDIT — CAGE identified a subset, not a different group, of the patients AUDIT flagged.
Evidence
Derivation — Mayfield, McLeod and Hall
1974 · n = 366366 patients admitted to a psychiatric service over one year, of whom 39% had alcoholism, used to test which brief interview questions discriminated alcoholic from non-alcoholic patients.
The four questions were not sensitive as an alcoholism detector if a full 4-item positive response was required as the threshold; requiring only 2 or 3 positive items instead gave a workable rapid screening technique for large groups — the basis for the ≥2 cutoff used ever since.
Popularisation and clinical dissemination — Ewing
1984A widely cited clinical review that named the CAGE acronym (Cut down, Annoyed, Guilty, Eye-opener) from the earlier interview questions and described their use in clinical and research settings, making CAGE the standard bedside mnemonic it remains today.
Established CAGE as the standard four-question clinical screen; the paper itself is a review and dissemination piece rather than a new validation cohort.
Diagnostic meta-analysis across clinical populations — Aertgeerts et al.
2004Meta-analysis of 10 studies (from 35 identified in a systematic search) using DSM criteria as the reference standard, pooling CAGE diagnostic accuracy across inpatient, primary care and ambulatory populations.
At the conventional ≥2 cutoff, pooled sensitivity was 0.87 in inpatients, 0.71 in primary care patients and 0.60 in ambulatory patients; pooled AUC was 0.87 (95% CI 0.85–0.89), with LR+ 3.44 and LR− 0.18. The authors concluded CAGE has limited diagnostic value for screening at this cutoff, given how much accuracy varies by setting.
Comparison with AUDIT in general medical admissions — McCusker et al.
2002 · n = 103Prospective study of 103 general medical inpatients, administering both CAGE and AUDIT to compare which patients each questionnaire identified as hazardous or harmful drinkers.
AUDIT identified 36% of patients as hazardous or harmful drinkers versus 22% identified by CAGE, and every CAGE-positive patient was also AUDIT-positive — confirming CAGE identifies a narrower, more dependence-focused subset of the drinkers AUDIT detects.
How it compares
CAGE vs AUDIT
Use AUDIT when the question is current hazardous or harmful drinking; CAGE is the narrower, dependence-focused tool and, in a head-to-head comparison, identified a subset of the patients AUDIT flagged, never a group of its own.
CAGE's four questions probe lifetime consequences of drinking, which is why it performs relatively well at detecting established dependence but poorly at detecting hazardous use that has not yet caused conflict, guilt, or a felt need to cut down. AUDIT includes quantity and frequency items, which is what lets it catch earlier-stage drinkers before they accumulate CAGE-detectable consequences. In 103 general medical admissions, AUDIT identified 36% as hazardous or harmful drinkers against 22% for CAGE, and every CAGE-positive patient was also AUDIT-positive.
CAGE vs AUDIT-C (three consumption questions)
AUDIT-C takes about the same time as CAGE and asks the question CAGE cannot — how much, how often, and how often heavily.
Both are four-or-fewer-item instruments that fit into a consultation, so the argument for CAGE on grounds of brevity does not really hold. The difference is what they measure: AUDIT-C is three consumption items scored 0 to 12, and it outperformed the full AUDIT for detecting heavy drinking in its derivation study; CAGE contains no consumption item at all. Where the clinical question is whether this patient's drinking could be damaging their liver now, AUDIT-C answers it and CAGE does not. CAGE retains a role as a very rapid screen for lifetime dependence, and the two can reasonably be used together.
Pearls & pitfalls
- A negative CAGE does not rule out hazardous drinking. The tool was built to detect dependence, and a comparison study found every drinker it did flag was also flagged by AUDIT — it never identified a case AUDIT missed.
- Diagnostic accuracy at the standard ≥2 cutoff is setting-dependent: pooled sensitivity was 0.87 in inpatients but only 0.60 in ambulatory patients in the largest meta-analysis. Weigh a negative result accordingly.
- The questions have no time frame ('have you ever'), so a positive answer can reflect a resolved episode from years ago rather than current drinking.
- A single positive answer, short of the 2-point threshold, still warrants a conversation — the original derivation found the full 4-item threshold too insensitive to use.
- If the clinical question is current hazardous or harmful consumption rather than lifetime dependence, use AUDIT (or AUDIT-C) instead — it is what a comparison study found identifies patients CAGE does not.
Critical actions
- At a score of 2 or more, take a full alcohol history covering quantity, pattern and withdrawal symptoms — a positive CAGE is a starting point, not a diagnosis.
- Assess for alcohol-related organ damage — liver, pancreas, cardiac and neurological — once the screen is positive.
- Offer or refer for treatment rather than stopping at the number; CAGE identifies who to ask more, not what to do next on its own.
- Use AUDIT alongside or instead of CAGE when the question is current hazardous drinking rather than lifetime dependence.
- Do not treat a score of 0 or 1 as reassurance in an ambulatory setting, where pooled sensitivity is lowest.
Why this score exists
The four CAGE questions were not written to measure how much someone drinks; they were selected because they discriminated alcoholic from non-alcoholic psychiatric inpatients in the original derivation work, and every item asks about a consequence of drinking — an attempt to cut down, others' criticism, guilt, needing a drink to function in the morning — that tends to appear only once dependence is established. Ewing's later review turned those questions into the CAGE mnemonic for everyday clinical use, but the underlying design goal never changed: this is a dependence screen, and the comparison literature confirms it behaves like one, missing hazardous drinkers who have not yet accumulated these consequences.
About the creator
Originator of the CAGE questionnaire
Devised the four CAGE questions and coined the acronym; his 1984 JAMA paper is the reference most often cited for the instrument itself.
First author, 1974 validation study
Published the validation study cited here, which tested the four questions against clinical diagnosis of alcoholism.
Limitations
- Built and best validated for detecting alcohol dependence, not for the broader and more common category of hazardous or harmful drinking.
- Pooled diagnostic accuracy at the standard ≥2 cutoff varies considerably by clinical setting — meaningfully better in inpatients than in ambulatory or primary care populations.
- No time frame on any item, so it cannot distinguish a current problem from a past one that has resolved.
- A large diagnostic meta-analysis concluded the tool has limited screening value at its own recommended cutoff, which is a more cautious read than CAGE's everyday reputation as a quick, reliable screen suggests.
- Does not quantify consumption, so it cannot inform brief-intervention conversations about drink counts the way AUDIT or AUDIT-C can.
If you are the patient
CAGE is four short yes-or-no questions your clinician might ask about your drinking: whether you have ever felt you should cut down, whether people have criticised your drinking, whether you have felt guilty about it, and whether you have ever needed a drink first thing in the morning. Two or more 'yes' answers is usually taken as a signal to talk about your drinking in more detail — it is a conversation-starter, not a diagnosis. It is worth knowing that this particular screen is better at picking up drinking that has already become a dependence than at picking up drinking that is risky but has not yet caused these kinds of problems — for that, your clinician may use a different, more detailed questionnaire (AUDIT) instead of or alongside CAGE.
Frequently asked questions
What does CAGE stand for?#
Cut down, Annoyed, Guilty, Eye-opener — the four questions asking whether you've felt you should cut down on drinking, been annoyed by criticism of your drinking, felt guilty about it, or had a morning drink to steady your nerves or treat a hangover.
What CAGE score indicates a problem?#
Two or more positive answers is the conventional threshold for concern, though a single positive answer still warrants further discussion, and the tool's own derivation work found a strict 4-item threshold too insensitive to use.
How accurate is the CAGE questionnaire?#
It varies substantially by setting. A meta-analysis of 10 studies found pooled sensitivity at the ≥2 cutoff was 0.87 in hospital inpatients but only 0.60 in ambulatory patients, with an overall pooled AUC of 0.87. The same analysis concluded CAGE has limited diagnostic value for screening at its recommended cutoff.
Is CAGE or AUDIT better for alcohol screening?#
It depends on what you're screening for. CAGE detects lifetime alcohol dependence; AUDIT also catches current hazardous and harmful drinking that has not yet caused the consequences CAGE asks about. A direct comparison in 103 general medical patients found AUDIT identified 36% as hazardous/harmful drinkers versus 22% for CAGE, and every CAGE-positive patient was also AUDIT-positive.
Can you have a drinking problem with a CAGE score of 0 or 1?#
Yes. CAGE was built to detect dependence with a history of consequences, not current hazardous drinking, and it has been shown to miss drinkers that AUDIT identifies. A low CAGE score should not be treated as ruling out a drinking problem, especially in an ambulatory setting where the tool's sensitivity is lowest.
Does the CAGE questionnaire ask how much someone drinks?#
No, and that is its defining limitation. All four items ask about the consequences and control of drinking — cutting down, annoyance at criticism, guilt, and eye-openers. None asks about quantity or frequency, so a patient drinking 60 units a week without guilt or failed attempts to stop can score zero. AUDIT was commissioned specifically to close that gap and its first three items measure consumption directly.
What time period does CAGE cover?#
Lifetime, not the last year. The questions are conventionally asked as 'have you ever', which means a positive answer can reflect a problem resolved a decade ago and a negative answer does not establish that current drinking is safe. That makes CAGE poorly suited to monitoring change over time — an AUDIT score can be repeated and compared between visits in a way a CAGE score cannot.
Is CAGE appropriate in pregnancy or in older adults?#
It performs less well in both. In pregnancy the concern is that any alcohol matters, and instruments such as T-ACE and TWEAK were developed because CAGE misses drinking levels that are relevant in that setting. In older adults, physiological changes mean harm occurs at lower intakes than the CAGE items typically capture, and the SMAST-G was designed for that group. In a gastroenterology clinic the practical response to both is to use a consumption-based instrument rather than relying on CAGE alone.
References
Original / primary reference
Validation and comparison studies
- Aertgeerts B, Buntinx F, Kester A. The value of the CAGE in screening for alcohol abuse and alcohol dependence in general clinical populations: a diagnostic meta-analysis. J Clin Epidemiol. 2004;57(1):30-39.
- McCusker MT, Basquille J, Khwaja M, Murray-Lyon IM, Catalan J. Hazardous and harmful drinking: a comparison of the AUDIT and CAGE screening questionnaires. QJM. 2002;95(9):591-595.