CAGE Alcohol Screening Calculator
CAGE Alcohol Screening Calculator
Score Ewing’s four CAGE questions out of 4. At the usual cut-off of two, the reviewed average sensitivity is 0.71 and specificity 0.90 — and at a prevalence of one in ten that makes most positives false.
CAGE score
4 lifetime items, 0 to 4Cut down: yes (1). Annoyed: no (0). Guilty: no (0). Eye-opener: yes (1)
Scoring
Range 0 to 4 · conventional cut-off 2 or more
- every item is a lifetime question
- all four ask “ever”. A person in sustained remission for a decade can score 4, and a person drinking forty units a week for the first time this year can score 0. The instrument identifies a history and not a current intake, which is why consumption-based screening exists alongside it
- 0.71 and 0.90
- Dhalla and Kopec’s 2007 review of three reliability and 16 validity studies gives these as the average sensitivity and specificity, with test-retest reliability of 0.80 to 0.95. The settings in which those averages held are medical and surgical inpatients, ambulatory medical patients and psychiatric inpatients; performance in primary care “has been varied”
- the predictive value collapses with prevalence
- 75.3 per cent at a prevalence of 30 per cent, 44.1 per cent at 10 per cent. Sensitivity and specificity do not move at all between those two lines: the only thing that changed is who is being screened. This is the most useful single fact about any screening instrument and it is almost never printed beside one
- where it does not work
- the review’s own words: “it has not performed well in white women, prenatal women, and college students”, and “it is not an appropriate screening test for less severe forms of drinking”. The original validation was done with men
Worked example
Cut down: yes (1). Annoyed: no (0). Guilty: no (0). Eye-opener: yes (1)
1 + 0 + 0 + 1 = 2 points
2 reaches the conventional cut-off of two or more
Apply Dhalla and Kopec's 0.71 and 0.90 to 1,000 people at a prevalence of 10 per cent: 100 with the disorder give 71 true positives, and 900 without give 90 false positives. 71 of 161 positives is a positive predictive value of 44.1 per cent
At a prevalence of 30 per cent the same two figures give 213 true and 70 false positives — 75.3 per cent. The instrument did not change; the population did
And note what this total cannot distinguish: both of these points could have been earned thirty years ago, because both questions ask "ever"
Two wordings of the four items, both in circulation
| Letter | NCETA (Flinders University) wording | European Union Drugs Agency wording |
|---|---|---|
| C | Have you ever felt you ought to Cut down on your drinking? | Have you felt the need to Cut down on your drinking? |
| A | Have people Annoyed you by criticising your drinking? | Do you feel Annoyed by people complaining about your drinking? |
| G | Have you ever felt bad or Guilty about your drinking? | Do you ever feel Guilty about your drinking? |
| E | Have you ever had a drink first thing in the morning to steady your nerves or to get rid of a hangover? | Do you ever drink an Eye-opener in the morning to relieve the shakes? |
What a positive CAGE is worth at four prevalences
| Prevalence of alcohol abuse or dependence | Positive predictive value | Negative predictive value |
|---|---|---|
| 30 in 100 | 75.3 per cent | 87.9 per cent |
| 10 in 100 | 44.1 per cent | 96.5 per cent |
| Sensitivity used | 0.71 (Dhalla and Kopec 2007 average) | |
| Specificity used | 0.90 (same review) |
Four lifetime questions, and what two of them are worth
Ewing’s CAGE asks four questions whose initials spell its name: cutting down, being annoyed, feeling guilty, and the morning eye-opener. Each yes scores one point, the range is 0 to 4, and the conventional reading is that two or more is a positive screen. It has survived since 1984 because it takes about thirty seconds and because it is free of any proprietor — the National Centre for Education and Training on Addiction records it as being in the public domain, and the European Union Drugs Agency records its access status as open.
The property that decides how a total should be read is that all four items ask about a lifetime. Someone in sustained remission for twenty years can still answer yes four times; someone drinking forty units a week for the first time this year can answer no four times. The CAGE therefore identifies a history of alcohol-related harm, not a current intake, and Dhalla and Kopec’s 2007 review of three reliability and 16 validity studies says the second half of that directly: it “is not an appropriate screening test for less severe forms of drinking”. A reader who wants to know how much someone is drinking should count the drinks.
That review gives the figures this page works with: test-retest reliability 0.80 to 0.95, average sensitivity 0.71 and average specificity 0.90, in medical and surgical inpatients, ambulatory medical patients and psychiatric inpatients. Performance in primary care “has been varied”, and the instrument “has not performed well in white women, prenatal women, and college students” — a limitation that follows from its origin, since the original validation was done with men. Those two averages are what the predictive-value table is built from, and the arithmetic is the reason to bother: at a prevalence of 30 per cent a positive CAGE carries a positive predictive value of 75 per cent, and at 10 per cent it carries 44 per cent. The instrument is identical in both lines. More than half of the positives in the second line do not have the disorder, and no feature of the test reveals which.
Attribution, as the free-use grant asks. The CAGE is Ewing’s, from Detecting alcoholism: the CAGE questionnaire, JAMA 1984. The National Centre for Education and Training on Addiction records it as “Available in the public domain. May be downloaded and used with appropriate acknowledgement”, and the European Union Drugs Agency records its access status as “Open access”.
A total on this page is a number, not a diagnosis. These instruments quantify what a person reports, or what an observer records at one moment; a diagnosis rests on a clinical assessment that no questionnaire total stands in for. A low total does not exclude what the instrument screens for. No sensitivity quoted here is 1, every figure was measured in a published cohort rather than in the person in front of you, and someone who endorses nothing may still have the disorder. This page renders no dose, no medication, no treatment regimen and no disposition. It computes the published total and prints the published thresholds with the body that published each one; what follows from the number is a clinical decision this page does not make. Every cut-off, sensitivity, specificity and severity band here is printed with the cohort it was measured in, or with a statement that no source read for this page attaches one — because an instrument’s accuracy is a property of the population it was measured in and not of the instrument. If you are reading this about yourself and you are in distress or thinking about harming yourself, please contact your local emergency number or a crisis line now rather than treating a number as an answer: 999 or Samaritans on 116 123 in the UK and Ireland, 988 in the United States and Canada, or your local emergency service elsewhere.
Frequently asked questions
What is a positive CAGE score?
Two or more of the four items answered yes is the conventional cut-off, and it is the one the National Centre for Education and Training on Addiction prints. Some primary-care protocols read a single yes as worth following up; no source read for this page attaches a sensitivity and specificity pair to a cut-off of one, so this page does not print one.
How accurate is the CAGE questionnaire?
Dhalla and Kopec’s 2007 review gives an average sensitivity of 0.71 and an average specificity of 0.90 across 16 validity studies, in medical and surgical inpatients, ambulatory medical patients and psychiatric inpatients. Those are averages across studies rather than a single cohort’s figures, and performance in primary care was varied.
Does the CAGE measure how much someone drinks?
No. All four items are lifetime questions and none asks about quantity or frequency. A person can score 4 having not drunk for two decades, and score 0 while drinking far above every national guideline. A total on this page is a number, not a diagnosis. These instruments quantify what a person reports, or what an observer records at one moment; a diagnosis rests on a clinical assessment that no questionnaire total stands in for.
Why does a positive CAGE mean less in a general population?
Because the positive predictive value depends on how common the disorder is among the people being screened. On the review’s own 0.71 and 0.90, a positive carries a 75.3 per cent positive predictive value at a prevalence of 30 per cent and 44.1 per cent at 10 per cent. Nothing about the instrument changed between those two numbers.
Is the CAGE valid in women?
Less so. Dhalla and Kopec report that it “has not performed well in white women, prenatal women, and college students”, and the National Centre for Education and Training on Addiction notes both that poor sensitivity has been found in women and that the original validation was done with men.
Related calculators
References
- Fischer J, Roche A, Duraisingam V. CAGE — tool overview. National Centre for Education and Training on Addiction (NCETA), Flinders University, 2021. Source of the four items in the wording used here, the 0-to-4 scoring with a cut-off of two or more, the Dhalla and Kopec average sensitivity of 0.71 and specificity of 0.90 with the settings they came from, the statement that poor sensitivity has been found in women and that the original validation was done with men, and the licence position: “Available in the public domain. May be downloaded and used with appropriate acknowledgement.”
- Dhalla S, Kopec JA. The CAGE questionnaire for alcohol misuse: a review of reliability and validity studies. Clin Invest Med. 2007. Abstract read: test-retest reliability 0.80 to 0.95; average sensitivity 0.71 and average specificity 0.90; three articles on reliability and 16 on validity, generally Level II evidence; “a valid tool for detecting alcohol abuse and dependence in medical and surgical inpatients”; performance in primary care “has been varied”; “it has not performed well in white women, prenatal women, and college students”; and “it is not an appropriate screening test for less severe forms of drinking”.
- European Union Drugs Agency, drugs library entry “CAGE Alcohol questionnaire”. Records the instrument’s access status as “Open access” and reproduces the four items in a second, different wording — “Have you felt the need to Cut down on your drinking?”, “Do you feel Annoyed by people complaining about your drinking?”, “Do you ever feel Guilty about your drinking?”, “Do you ever drink an Eye-opener in the morning to relive the shakes?” (the misprint for “relieve” is the page’s). Cites Ewing JA. Detecting alcoholism: the CAGE questionnaire. JAMA. 1984;252(14):1905–1907.
- CamCOPS task documentation, CAGE Questionnaire, University of Cambridge Department of Psychiatry. Records the intellectual-property position as “Believed to be in the public domain (based on Mayfield et al. 1974, Ewing 1984, O’Brien 2008), particularly Ewing (1984), the original author, who encourages widespread use”, and quotes Ewing’s own closing sentence. Also records the pre-publication history: Ewing J, Rouse BA. Identifying the hidden alcoholic. 29th International Congress on Alcohol and Drug Dependence, Sydney, 3 February 1970.
Not medical advice. For healthcare professionals and education. Reference intervals vary by laboratory and assay — always use your own laboratory's. Never base a dose or a treatment decision on this page alone. Full disclaimer at calcengines.com/disclaimer/