The AUDIT-C is the three-item consumption subset of the World Health Organization's ten-item AUDIT. It asks how often a person drinks, how much they typically drink on a drinking day, and how often they drink heavily on a single occasion. Each item scores 0 to 4, giving a total from 0 to 12. It measures alcohol consumption - frequency, typical quantity and heavy episodic drinking - and nothing else.
That restriction is deliberate, and it is also the instrument's defining limitation. The seven AUDIT questions the AUDIT-C omits cover dependence symptoms (impaired control, increased salience, morning drinking) and alcohol-related harm (guilt, blackouts, injuries, others' concern). The consequence shows up cleanly in the original validation data: the AUDIT-C was marginally better than the full AUDIT at detecting heavy drinking (AUROC 0.891 versus 0.881, p = 0.03) and significantly worse at detecting active alcohol abuse or dependence (0.786 versus 0.811, p < 0.001). It is a consumption screen that performs well as a first filter, not a substitute for the full instrument when the question is whether a disorder is present.
02 - Origin & purpose
Where it comes from.
Bush, Kivlahan, McDonell, Fihn and Bradley published the AUDIT-C in the Archives of Internal Medicine in 1998, as part of the Ambulatory Care Quality Improvement Project. Working with 243 male general medicine patients at a US Veterans Affairs centre, they asked whether the AUDIT's first three consumption questions could carry most of the parent instrument's screening performance at under a third of the length. They could. The paper's purpose was pragmatic: to make routine alcohol screening short enough to survive contact with a busy clinic.
The original study excluded women - the authors noted explicitly that alcohol screening questionnaires function differently by sex and that they had too few female participants. That gap was filled by Bradley and colleagues in 2003 with 393 female VA patients, and then in 2007 with 1,319 men and women in academic family practice, which established the sex-specific thresholds now in general use. The AUDIT-C has since become the default screening instrument across the US Veterans Health Administration and is one of the two tools named by the US Preventive Services Task Force in its Grade B recommendation for unhealthy alcohol use screening in adults.
03 - Scoring & cutoffs
How scoring works.
Sum the three items for a total between 0 and 12. The threshold is sex-specific: 4 or more is a positive screen in men, 3 or more in women. At those thresholds, Bradley and colleagues (2007) reported sensitivity 0.86 and specificity 0.89 in men, and sensitivity 0.73 and specificity 0.91 in women, against a reference standard of risky drinking and/or DSM-IV alcohol use disorder.
A positive screen means "ask more", not "diagnose". The appropriate next step is the full ten-item AUDIT, or a structured assessment of dependence symptoms and harm, since the AUDIT-C by construction asks about neither.
Three qualifications are worth carrying. First, the WHO's familiar threshold of 8 applies to the ten-item AUDIT, not to the AUDIT-C; there are no WHO-issued AUDIT-C cutoffs, and the 4/3 thresholds are derived from US primary care and veteran populations. Second, European validation samples have generally found a higher optimal threshold of around 5, and studies in community-dwelling older adults suggest 5 for men and 4 for women. Third, in pregnancy the conventional approach is that any score above 0 warrants discussion, since no consumption threshold is established as safe.
Score
Severity
Interpretation
0–2
Lower risk
Below the screening threshold for both men and women.
3–3
Positive in women
At or above the threshold for women (≥3). Below the threshold for men. Assess further in women.
4–7
Positive
At or above the threshold for both men (≥4) and women. Indicates hazardous drinking; assess further with the full AUDIT.
8–12
Strongly positive
High likelihood of hazardous or harmful drinking; specificity 97% for heavy drinking at ≥8 in the original validation. Assess for dependence and alcohol-related harm.
04 - Validation evidence
How well it performs.
Bush and colleagues' original sample of 243 male veterans had a high base rate - 35% heavy drinking, 21% active abuse or dependence, 41% either - which makes the accuracy figures generous relative to general practice. Across thresholds, the sensitivity-specificity trade-off is steep: for heavy drinking, a cutoff of 3 gave sensitivity 98% and specificity 57%, a cutoff of 4 gave 91% and 70%, and a cutoff of 5 gave 73% and 88%. The CAGE, tested in the same population, achieved an AUROC of only 0.717 and identified just 56% of patients with heavy drinking or active abuse or dependence.
The strongest accuracy evidence comes from Bradley and colleagues (2007) in 392 men and 927 women in academic family practice, where alcohol misuse prevalence was 33% and 19% respectively. The AUDIT-C produced an AUROC of 0.94 (95% CI 0.91-0.96) in men and 0.90 (0.87-0.93) in women, performing as well as the full AUDIT and significantly better than self-reported risky drinking, AUDIT question 3 alone, or the augmented CAGE (all p < 0.001). In the earlier female-only validation, the AUDIT-C returned an AUROC of 0.91, ahead of the full AUDIT's 0.87 in the same sample.
Reliability evidence is unusually strong for a screening instrument, because it comes from routine care rather than a research setting. Simon and colleagues (2024) examined more than 18,000 Kaiser Permanente Washington primary care adults screened twice between one and 21 days apart: the intraclass correlation was 0.93 for screens completed twice through the patient portal, 0.81 for two in-clinic screens, and 0.83 for mixed modalities. Reliability was somewhat lower among American Indian or Alaska Native and multiracial patients.
0.94 / 0.90
AUROC FOR ALCOHOL MISUSE, MEN / WOMEN (BRADLEY 2007)
0.86 / 0.89
SENSITIVITY / SPECIFICITY AT ≥4 IN MEN (BRADLEY 2007)
0.73 / 0.91
SENSITIVITY / SPECIFICITY AT ≥3 IN WOMEN (BRADLEY 2007)
ICC 0.93
TEST-RETEST, PORTAL SCREENS 1-21 DAYS APART, >18,000 ADULTS (SIMON 2024)
05 - How it compares
How it compares to the alternatives.
Instrument
Items
Time
When to reach for it
AUDIT-C
3
~1 min
Range 0-12. Rapid first-line screen for hazardous and heavy drinking; universal screening at scale. Positive at ≥4 in men, ≥3 in women.
Range 0-40. When you need dependence symptoms and harm as well as consumption; severity grading; the natural follow-on from a positive AUDIT-C. ≥8 (≥7 in women and adults over 65).
Range 0-10. Drug problem severity in patients already known to use; does not cover alcohol. ≥3.
Single-item alcohol screen (NIAAA)
1
~15 s
Count response. The minimum-burden option, endorsed alongside the AUDIT-C by the USPSTF. ≥1 heavy drinking day in the past year (sensitivity 82%, specificity 79% for any unhealthy use).
Range 0-32. Cannabis misuse screening; the AUDIT analogue for cannabis. ≥8.
06 - When to use it
Right tool, wrong tool.
Reach for it when
-You are screening an entire caseload rather than investigating a suspicion - one to two minutes makes universal screening viable.
-Intake, annual review or self-completed pre-appointment forms.
-You want to detect hazardous consumption early, before dependence or harm has developed.
-You need a first filter that routes cleanly onto the full AUDIT without repeating questions.
-Psychiatric settings where alcohol use is a common and under-asked comorbidity.
Reach for something else when
-The question is whether alcohol use disorder is present. The AUDIT-C omits every dependence and harm item; use the full AUDIT or a diagnostic assessment.
-You need to measure change during treatment. No minimal important difference has been established.
-The patient is under 18. The USPSTF issued an insufficient evidence statement for adolescents aged 12 to 17; the CRAFFT is the instrument generally used in that age group.
-You are working outside the country the thresholds were derived in without checking local standard-drink definitions - see limitations.
-The patient has already disclosed problematic drinking. Screening adds nothing; move to assessment.
Test-retest reliability is the AUDIT-C's best-documented measurement property, with intraclass correlations between 0.81 and 0.93 depending on administration mode in a very large routine-care sample. Self-completion through a patient portal was more reliable than in-clinic administration, which is consistent with the long-standing finding that people report drinking more accurately when the setting feels private.
No published Cronbach's alpha exists for the AUDIT-C as a three-item scale, and no standard error of measurement has been reported. We have deliberately not calculated one here, because doing so would require assumptions the literature does not support. Nor is there an established minimal clinically important difference: the instrument was developed and validated as a screen, and its responsiveness to change has not been demonstrated.
One finding bears directly on how much precision to expect. Delaney and colleagues (2014), analysing two US national samples, found that up to 21% of patients produced AUDIT-C results inconsistent with the drinking they had reported on the same three questions - screening positive while describing consumption below recommended limits, or screening negative while describing consumption above them. The score is a useful trigger for conversation. It is not a measurement of intake, and small differences between two administrations should not be interpreted clinically.
08 - Limitations
What it cannot tell you.
Lower sensitivity in women. The original validation excluded women entirely, and at the recommended thresholds sensitivity is 0.73 in women against 0.86 in men. Question 3's six-drink threshold also sits above the four-drink binge definition used for women in the United States, which is why sex-specific variants of the item exist.
Performance varies by race and ethnicity. Frank and colleagues (2008) found comparable AUROCs across White, African-American and Hispanic patients but significant differences in sensitivity at the recommended cut points - among women, 85% in Hispanic patients against 67% in African-American and 70% in White patients; among men, 95% in White against 76% in African-American patients. Specificities did not differ significantly.
Standard drinks are not standard. The AUDIT was built around a 10 g standard drink; a US drink is 14 g and a Japanese one nearly 20 g. Scores are therefore not directly comparable across countries, and the US Alcohol Use Disorders Identification Test (USAUDIT) was developed partly in response.
Reduced sensitivity in older adults. The WHO manual notes low sensitivity above age 65, and validation work in community-dwelling older adults suggests higher thresholds than the standard 4 and 3.
It does not measure dependence or harm. This is by design, but it means a low score does not exclude a serious alcohol problem in someone whose consumption pattern is atypical.
Self-report. Under-reporting is expected where privacy is limited or where disclosure carries perceived consequences.
Not a diagnostic instrument. The WHO is explicit on this point for the AUDIT family: a diagnosis requires structured interview or specialist assessment.
Recall period is the past year, so recent change in drinking may not be reflected.
[9]Babor TF, Higgins-Biddle JC, Saunders JB, Monteiro MG AUDIT: The Alcohol Use Disorders Identification Test - Guidelines for Use in Primary Care, 2nd ed. WHO/MSD/MSB/01.6a (2001) ↩
Related scales
Instruments clinicians often use alongside this one
See how Aisel removes friction where it costs most. A 20-minute walkthrough tailored to your clinic.
We value your privacy
We use cookies to analyse site usage and improve your experience. Analytics and embedded media (e.g. YouTube) only load if you accept. Read our cookie policy.