AUDIT alcohol screening: scoring and interpretation

Summaries of the AUDIT reproduce a scoring table and stop. The instrument ships with a free WHO manual that answers what the tables leave open — how items 9 and 10 are scored, what a score should trigger, where the cutoff came from: AUDIT: Guidelines for Use in Primary Care, 2nd edition. There is also an official Spanish version published by WHO itself, under the same reference number.

Two things get lost: the AUDIT’s output is a recommended action, not a severity label, and its cutoff behaves very differently depending on what you are detecting. This guide covers screening, documentation and referral; managing consumption and withdrawal is medical territory.

What the AUDIT measures and what backs it

The manual describes it as the only screening test designed specifically for international use, validated on primary care patients in six countries, Mexico among them. Ten items, three domains: hazardous use, dependence symptoms, harmful use.

The original validation study (Saunders and colleagues, 1993) drew the items from a 150-item schedule given to 1,888 primary care patients: among those diagnosed with hazardous or harmful drinking, 92% scored 8 or more, and 94% of those with non-hazardous consumption scored below 8. The reference standard matters — hazardous or harmful drinking, not a DSM disorder.

Administering it

The AUDIT asks about the past year, takes two to four minutes, and works as an oral interview or a self-report questionnaire. The manual recommends universal rather than targeted screening: screen only those you already suspect and you miss most heavy drinkers. It fits an intake interview and an annual re-screen.

Scoring is not uniform: items 1 to 8 score 0 to 4; items 9 and 10 score 0, 2 or 4, with three options — “No”, “Yes, but not in the last year” and “Yes, during the last year”. Scoring all ten items 0 to 4, the commonest error in circulating copies, mis-scores everyone who answers “Yes, but not in the last year”. Total range: 0 to 40.
The skip-outs belong to the interview only. The oral format lets you jump to questions 9 and 10 if the person never drinks or if questions 2 and 3 total zero; the manual calls that logic likely too hard to follow on paper.
Adapt the standard drink to local serving sizes, not the scoring.

A note for anyone using the Spanish version: in the PDF WHO serves on IRIS, the interview form (Cuadro 4, p. 17) codes item 2’s fifth option as (3) where the English original has (4). Scored as printed, it under-scores the heaviest drinkers by a point; the self-report grid in Anexo B (Cuadro 10, p. 31) is correct. Its Cuadro 6 also mislabels the third row “Zona II” where its own text says Zone III, and its footnote asks for “el nivel máximo” where the English asks for the next highest level, and “y” where the English has “or”.

Interpreting the score: zones, not labels

The AUDIT does not return severity. It returns a risk level with an intervention attached. Box 6 of the manual:

Risk level	AUDIT score	Recommended intervention
Zone I	0–7	Alcohol education
Zone II	8–15	Simple advice
Zone III	16–19	Simple advice plus brief counseling and continued monitoring
Zone IV	20–40	Referral to specialist for diagnostic evaluation and treatment

The third column is the only one that answers the clinical question. The manual hedges its own bands: absent better research they are tentative, subject to clinical judgment. Its footnote, in the English, carries the most actionable rule in the document: provide the next highest level of intervention to patients who score 2 or more on questions 4, 5 and 6, or 4 on questions 9 or 10. A 9 with a 4 on question 9 is not an ordinary 9.

The cutoff is not a border

The manual recommends 8 or more as an indicator of hazardous and harmful use and possible dependence, adding that a cutoff of 10 gives greater specificity at the cost of sensitivity.

On sex and age, verbatim: “Since the effects of alcohol vary with average body weight and differences in metabolism, establishing the cut off point for all women and men over age 65 one point lower at a score of 7 will increase sensitivity for these population groups.” The English is ambiguous about whether “over age 65” attaches to men only; WHO’s official Spanish resolves it as all women, and men over 65. Either way it is a sensitivity-increasing suggestion, expressly subordinated to national standards and clinician judgment — not an alternative cutoff with accuracy data of its own. The closest empirical support is a comparison of emergency-department samples in Mexico (n = 1,417) and among Mexican-Americans (n = 586), where lowering cut points substantially improved performance for females across the four screeners compared.

Now the counterweight. The 2018 USPSTF review pooled seven studies (n = 8,852) using the recommended cutoff of higher than 8 in general adult populations: sensitivity 0.38 to 0.73, specificity 0.89 to 0.97. That does not contradict the 92% from 1993 — the target condition changed. At 8 the AUDIT detects dependence well and the broader band of unhealthy drinking that is not yet a disorder poorly. Do not move the cutoff on your own authority; know what you are screening for first.

What an AUDIT score does not tell you

The manual says it plainly: diagnosis is a necessary step after a high score, “since the instrument does not provide sufficient basis for establishing a management or treatment plan.” It should be done by a professional trained in diagnosing alcohol use disorders, through a structured interview such as the CIDI or SCAN.

It does not map onto DSM-5. The AUDIT was built on ICD-10 constructs; DSM-5 merged abuse and dependence into a single graded disorder. There is no AUDIT-score-to-DSM-5-severity conversion.
It is not a session-to-session progress measure. Its window is a year, so re-administering in two weeks measures nearly the same period twice. Unlike the PHQ-9, it is a screen plus periodic re-screen — worth settling when you decide how you will measure progress.
The evidence thins at the edges. The manual concedes that age has not been systematically analyzed, that one study found low sensitivity but high specificity above 65, and that evidence on women is “somewhat limited.”
AUDIT-C cutoffs are not WHO’s. The three-question version came out of a 1998 VA study that excluded female patients; the familiar thresholds of 3 for women and 4 for men come from that US research line, which the USPSTF adopts as standard. WHO publishes no AUDIT-C cutoff.

The Spanish-language evidence, and its limits

An official translation is not a validation. WHO’s Spanish version was produced by José Martínez-Raga, Bartolomé Pérez Gálvez, Miguel Castellano Gómez and María Dolores Temprado Albalat, with the imprint of the “Conselleria de Benestar Social,” in Spain. It is a Spain translation, not a Mexican adaptation, and WHO publishing it does not validate it in any Spanish-speaking population.

Reliability and factor structure are documented. Morales Quintero and colleagues (2019) tested the AUDIT in 1,932 young people from southern Mexico aged 14 to 30: Cronbach’s alpha of .804 and a confirmatory factor analysis supporting the three-factor model; they opened the paper noting that validation studies for that population did not exist. That is internal consistency and structure, not diagnostic accuracy — there was no reference interview.

Diagnostic accuracy exists, and it comes from emergency departments. Cremonte and colleagues (2010) compared screening instruments against a DSM-IV diagnosis of alcohol dependence obtained with the CIDI. In the Mexican sample (n = 526 current drinkers, Pachuca), the AUDIT at a cut point of 8 reached 92% sensitivity and 98% specificity, alpha .86. But the target is dependence and the setting an emergency department: positive predictive value in outpatient practice will differ, and that study used a locally translated version, not WHO’s.

And what we could not find. In PubMed-indexed literature we did not locate a diagnostic-accuracy validation of the AUDIT in a Mexican general-population or primary-care sample against a structured diagnostic interview. The closest, a 1998 chapter by Medina-Mora, Carreño and De la Fuente, reports “adequate levels of specificity and sensitivity” in Mexican general-population and general-practice samples, without naming the standard. The rest is emergency-department data, factor structure without a gold standard, and 80 outpatients with schizophrenia at Mexico City’s National Institute of Psychiatry, in which the best cut point was 4, not 8.

From screening to referral: where your scope ends

The manual uses stepped care: start at the lowest level of intervention the score suggests and move up if the person does not respond. Its brief-intervention elements are familiar — the result, the risks, an agreed goal — and it asks for something often skipped: feed back the negative result too.

The hard boundary is medical. Detoxification may be necessary, and special attention is owed to anyone reporting daily consumption of large amounts or scoring on the dependence questions. Asking how long since the person had an alcohol-free day, and whether they have had withdrawal symptoms, is assessment; managing withdrawal is not. That is where you refer — to the alcohol or substance-use service covering your client’s location, and their physician — and document it, and whether it was accepted.

Where the score lives in the note

Many jurisdictions regulate what a clinical record must contain, sometimes naming substance use explicitly. In Mexico, NOM-004-SSA3-2012 requires the progress note to describe the clinical picture “including, where applicable, abuse and dependence on tobacco, alcohol and other psychoactive substances,” and to record relevant results of previously requested studies — where a screening score belongs. What that means for an independent practice is in the guide to Mexico’s clinical record standard. Verify the rules where you practice; the discipline travels.

“AUDIT by interview: 17 (Zone III); 3 on question 5 and 2 on question 6; no alcohol-free days in three weeks. Intake baseline: 11. Hazardous use with dependence symptoms to be confirmed; the instrument does not establish a diagnosis. Reduction goal agreed; referred for medical evaluation for possible withdrawal; re-screen in three months.”

In a SOAP note the number goes in O, the zone in A, the action in P. A 17 with no zone and no referral explains nothing later.

Licensing: free to reproduce, not to commercialize

The document’s own notice, verbatim: “This document is not a formal publication of the World Health Organization (WHO), and all rights are reserved by the Organization. The document may, however, be freely reviewed, abstracted, reproduced, and translated, in part or in whole but not for sale or for use in conjunction with commercial purposes.”

Administering it with your patients is what WHO wrote and distributes the manual for. But it is not public domain, unlike the PHQ-9 and GAD-7: WHO reserves all rights and grants a permission conditioned on non-commercial use. Where that condition bites — a fee-charging practice, a paid product, an app — WHO does not gloss, so ask WHO rather than reason by analogy. The permission covers this edition: the second (2001) and its official Spanish version, not third-party adaptations. And it is not CC BY-NC-SA; WHO’s general licence postdates this document. Other instruments are compared in the guide to psychological test licensing.

Scoring scales with different rules and filing the number next to its zone and the action taken is easy with one client and heavy with thirty. That is where a platform helps: gesell.ai includes built-in validated scales with automatic scoring — PHQ-9 and GAD-7 among them — inside each client’s structured chart, so today’s score sits beside the previous ones, ready for your note. Always as decision support you evaluate: the scale contributes the data point; the clinical judgment and the referral remain yours.