Alcohol Use Disorder: Diagnosis, Severity, and the Diagnostic Gap

A Clinical and Research Reference

Overview — Why Diagnostic Precision Matters

Alcohol use disorder (AUD) is one of the most prevalent and undertreated conditions in medicine. Approximately 10.9% of U.S. adults meet diagnostic criteria for AUD ^[1] (Note: this specific figure could not be independently verified against the source abstract — the underlying study supports the general finding but the exact number should be confirmed before publication), yet the disorder remains profoundly underrecognized in clinical settings. In a national cohort of 114,511 adults with unhealthy alcohol use, only 10.1% had received a formal AUD diagnosis — with rates as low as 6.8% among those with mild-risk presentations ^[1]. That gap is not a minor inefficiency. It is a structural failure with direct treatment consequences.

The stakes of diagnostic precision extend well beyond documentation. Receiving a formal AUD diagnosis was associated with an adjusted odds ratio of 10.68 (95% CI: 9.68–11.79) for receiving pharmacotherapy and 1.57 (95% CI: 1.46–1.69) for receiving psychotherapy ^[1]. The diagnosis is the gateway. Every missed diagnosis is a missed treatment opportunity — and the evidence shows that missed diagnoses are not randomly distributed across the population.

The DSM-IV to DSM-5 Transition

Understanding the current diagnostic framework requires understanding what changed in 2013. The Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) divided alcohol-related pathology into two separate categories: alcohol abuse (a pattern of harmful use without physiological dependence) and alcohol dependence (a more severe syndrome including tolerance, withdrawal, and compulsive use). These were treated as distinct diagnoses with separate criteria sets.

The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) collapsed this binary into a single dimensional construct — alcohol use disorder — rated along a severity continuum based on the number of 11 criteria endorsed ^[2]. This was a conceptually significant shift: it acknowledged that the boundary between "abuse" and "dependence" was clinically arbitrary and that alcohol-related pathology exists on a spectrum.

The epidemiological consequences of this change are important to flag whenever citing prevalence data. Some individuals who met DSM-IV criteria for alcohol abuse (but not dependence) now meet DSM-5 criteria for mild AUD; conversely, some DSM-IV dependence cases may be reclassified under DSM-5 severity tiers. Cross-study prevalence comparisons that span the DSM-IV/DSM-5 transition should be interpreted with caution, as the diagnostic framework itself changed. The corpus reviewed here does not contain direct empirical comparisons of DSM-IV versus DSM-5 prevalence in the same cohort — a gap noted by the expert panel ^[3] (Note: this specific figure could not be independently verified against the source abstract — the underlying study supports the general finding but the exact number should be confirmed before publication).

One additional framework deserves mention: the International Classification of Diseases, 11th Revision (ICD-11), used widely outside North America, employs different terminology and criteria. The same patient may receive different diagnoses under DSM-5 versus ICD-11, with implications for cross-national research comparisons. This is addressed further in the comparative frameworks section below.

The 11 DSM-5 Criteria

The DSM-5 diagnosis of AUD requires endorsement of at least 2 of the following 11 criteria within a 12-month period. The criteria are organized into four conceptual domains: impaired control, social impairment, risky use, and pharmacological criteria. What follows uses the source-language framing from the diagnostic literature, followed by a plain-English translation and a clinical example.

A critical methodological note before proceeding: criteria are not interchangeable in their clinical weight, even though the count-based severity model treats them as such. Research from the COGA (Collaborative Study on the Genetics of Alcoholism) cohort demonstrates that within the mild-to-moderate range (2–5 criteria), the presence of even one high-risk criterion — particularly withdrawal — was associated with dramatically accelerated progression to severe AUD ^[2]. This finding is examined in detail in the severity section below.

Domain 1: Impaired Control

Criterion 1 — Quantity/Duration Excess
Source framing: Alcohol is often taken in larger amounts or over a longer period than was intended.
Plain English: The person consistently drinks more than they planned to — starting with "just one drink" and ending several hours later.
Clinical example: A patient reports intending to have two glasses of wine with dinner but regularly finishing the bottle.
DSM history: Carried over from DSM-IV dependence criteria.

Criterion 2 — Persistent Desire or Unsuccessful Efforts to Cut Down
Source framing: There is a persistent desire or unsuccessful efforts to cut down or control alcohol use.
Plain English: The person has tried to drink less or stop, but these attempts have repeatedly failed.
Clinical example: A patient has made three serious attempts to stop drinking in the past year, each lasting less than two weeks.
DSM history: Carried over from DSM-IV dependence criteria.

Criterion 3 — Time Spent
Source framing: A great deal of time is spent in activities necessary to obtain alcohol, use alcohol, or recover from its effects.
Plain English: Alcohol-related activities — buying, drinking, recovering from hangovers — consume a disproportionate amount of the person's day.
Clinical example: A patient spends most weekend mornings in bed recovering from Friday and Saturday night drinking, losing time with family.
DSM history: Carried over from DSM-IV dependence criteria.

Criterion 4 — Craving
Source framing: Craving, or a strong desire or urge to use alcohol.
Plain English: The person experiences intense urges to drink that are difficult to resist.
Clinical example: A patient describes intrusive thoughts about drinking that begin mid-afternoon on workdays.
DSM history: New in DSM-5. This criterion was not present in DSM-IV and has been the subject of ongoing debate. Its inclusion reflects neuroscientific evidence on incentive salience and reward circuitry, but critics have questioned whether craving is sufficiently distinct from other criteria to add independent diagnostic information. This controversy is addressed in the Evidence Gaps section below.

Domain 2: Social Impairment

Criterion 5 — Failure to Fulfill Major Role Obligations
Source framing: Recurrent alcohol use resulting in a failure to fulfill major role obligations at work, school, or home.
Plain English: Drinking is interfering with the person's responsibilities — missing work, neglecting children, failing classes.
Clinical example: A patient has received two written warnings at work for arriving late after nights of heavy drinking.
DSM history: Adapted from DSM-IV abuse criteria.

Criterion 6 — Continued Use Despite Social or Interpersonal Problems
Source framing: Continued alcohol use despite having persistent or recurrent social or interpersonal problems caused or exacerbated by the effects of alcohol.
Plain English: The person keeps drinking even though it is damaging their relationships.
Clinical example: A patient's spouse has threatened separation because of drinking-related arguments, yet drinking continues.
DSM history: Adapted from DSM-IV abuse criteria.

Criterion 7 — Activities Given Up or Reduced
Source framing: Important social, occupational, or recreational activities are given up or reduced because of alcohol use.
Plain English: The person has stopped doing things they used to enjoy — hobbies, socializing, exercise — because drinking has taken priority.
Clinical example: A patient who previously ran marathons has stopped training because hangovers make morning runs impossible.
DSM history: Carried over from DSM-IV dependence criteria.

Domain 3: Risky Use

Criterion 8 — Use in Physically Hazardous Situations
Source framing: Recurrent alcohol use in situations in which it is physically hazardous.
Plain English: The person drinks in contexts where doing so creates physical danger — driving, operating machinery, swimming.
Clinical example: A patient regularly drives home after drinking at a bar, rationalizing that they "feel fine."
DSM history: Adapted from DSM-IV abuse criteria.

Criterion 9 — Continued Use Despite Physical or Psychological Problems
Source framing: Alcohol use is continued despite knowledge of having a persistent or recurrent physical or psychological problem that is likely to have been caused or exacerbated by alcohol.
Plain English: The person keeps drinking even though they know it is making a health problem worse.
Clinical example: A patient with alcohol-related liver disease continues to drink despite explicit medical advice to stop.
DSM history: Carried over from DSM-IV dependence criteria.

Domain 4: Pharmacological Criteria

Criterion 10 — Tolerance
Source framing: Tolerance, as defined by either (a) a need for markedly increased amounts of alcohol to achieve intoxication or desired effect, or (b) a markedly diminished effect with continued use of the same amount of alcohol.
Plain English: The person needs to drink significantly more to feel the same effect, or the same amount produces much less effect than it used to.
Clinical example: A patient reports that the six-pack that used to make them feel drunk now barely affects them.
DSM history: Carried over from DSM-IV dependence criteria.
Important caveat: Tolerance is the most frequently endorsed criterion in young adult populations — endorsed by 50.3% of regular drinkers aged 18–21 in the RADAR cohort ^[4]. When tolerance was removed from the criteria set in that study, DSM-5 lifetime AUD prevalence dropped from 18.4% to 11.0% — a 40% reduction. This raises legitimate questions about whether tolerance functions as a genuine marker of disorder or a developmentally normative response to early alcohol exposure in younger populations ^[4].

Criterion 11 — Withdrawal
Source framing: Withdrawal, as manifested by either (a) the characteristic withdrawal syndrome for alcohol, or (b) alcohol (or a closely related substance) is taken to relieve or avoid withdrawal symptoms.
Plain English: The person experiences physical symptoms when they stop or cut back drinking — tremors, sweating, anxiety, seizures in severe cases — or they drink specifically to prevent those symptoms.
Clinical example: A patient reports waking up shaking and needing a drink before work to "steady their nerves."
DSM history: Carried over from DSM-IV dependence criteria.
Clinical significance: Withdrawal is the criterion with the strongest prognostic weight. DSM-5 withdrawal syndrome prevalence was 14.3% among unhealthy drinkers nationally ^[5]. Within mild-to-moderate AUD, endorsement of withdrawal was associated with an adjusted hazard ratio of 11.62 (95% CI: 7.54–17.92) for progression to severe AUD ^[2] — nearly double the hazard ratio for mild-to-moderate AUD without high-risk criteria (aHR = 5.64; 95% CI: 3.28–9.70). Withdrawal endorsement should trigger immediate clinical attention regardless of total criterion count.

One criterion removed from DSM-IV: Legal problems (recurrent alcohol-related legal difficulties) was a DSM-IV abuse criterion that was eliminated in DSM-5. The rationale was that legal consequences of drinking are heavily influenced by socioeconomic and racial factors that have little to do with the underlying disorder — its removal was intended to reduce diagnostic bias.

Severity Stratification

DSM-5 grades AUD severity by criterion count:

The severity gradient has real administrative traction — clinicians are more likely to document a diagnosis as severity increases. But the expert panel reached a nuanced consensus: the three-tier model is necessary but not sufficient.

The core problem is within-tier heterogeneity. Two patients both classified as "mild AUD" with three criteria each may have radically different clinical trajectories depending on which three criteria they endorse. The COGA cohort data demonstrate this directly: within the mild-to-moderate band, the presence of withdrawal as even one of the endorsed criteria was associated with an aHR of 11.62 for progression to severe AUD, compared to 5.64 for those without high-risk criteria — a nearly twofold difference in progression risk within the same severity tier ^[2].

The cutoffs themselves — 2–3, 4–5, ≥6 — are somewhat arbitrary. The corpus does not contain direct evidence that these specific thresholds were empirically derived from outcome data rather than expert consensus. Alternative conceptualizations have been proposed: ^[6] documents that frameworks based on "harmful dysfunction" or "optimal criteria" produce smaller diagnostic classes with stronger concurrent validity than DSM-5 criteria alone — suggesting DSM-5's broad net may inflate prevalence while simultaneously misclassifying severity within the captured population.

At the same time, the count-based approach retains genuine utility as a longitudinal tracking tool. In a two-year follow-up study of 250 AUD patients, DSM-5 criterion counts showed a 50.9% reduction from baseline to 104-week follow-up [sjödin-2026-drinking-motives-among] (Note: this specific figure could not be independently verified against the source abstract — the underlying study supports the general finding but the exact number should be confirmed before publication) — demonstrating that the dimensional model is sensitive to clinically meaningful change over time. The criteria count works reasonably well for tracking treatment response; the problem is using it alone for initial risk stratification.

Clinical implication: Severity tier should be documented, but clinicians should additionally note which criteria are endorsed — particularly withdrawal, which carries independent prognostic weight regardless of total count ^[2]. The Alcohol Symptom Checklist, a patient-reported instrument assessing all 11 DSM-5 criteria, demonstrated excellent test-retest reliability in routine care settings (ICC = 0.79, 95% CI: 0.76–0.82 in one study; ICC = 0.82, 95% CI: 0.77–0.85 in a primary care sample) ^[7], suggesting criterion-level assessment is feasible outside structured research interviews.

Remission Specifiers

DSM-5 includes remission specifiers that are clinically and documentarily important, particularly for insurance authorization, disability determinations, and treatment planning.

Early remission: No DSM-5 AUD criteria met (except craving) for at least 3 months but less than 12 months.

Sustained remission: No DSM-5 AUD criteria met (except craving) for 12 months or longer.

The exclusion of craving from remission criteria reflects the recognition that craving may persist long after other symptoms resolve — its presence alone does not preclude remission status.

Specifiers for context:
- In a controlled environment: Applied when the individual is in an environment where access to alcohol is restricted (e.g., incarceration, residential treatment). This specifier acknowledges that abstinence in a controlled setting does not carry the same prognostic weight as community-based remission.
- On maintenance therapy: Applied when the individual is taking an approved medication for AUD (naltrexone, acamprosate, disulfiram) and no criteria other than tolerance or withdrawal are met.

These specifiers matter for documentation because they communicate clinical context that a simple "AUD in remission" notation does not. For insurance and disability purposes, the distinction between early and sustained remission, and between community-based and controlled-environment remission, can affect coverage determinations.

The natural history data are relevant here: cumulative incidence of AUD symptoms from late adolescence to age 42 was 58.0% (95% CI: 52.3–63.8), peaking at age 24, with 67.0% (95% CI: 61.1–73.0) of affected individuals eventually remitting by middle adulthood ^[8]. However, 25% had ongoing or new-onset AUD by middle adulthood — underscoring that remission is common but not universal, and that late-onset or persistent AUD represents a clinically distinct subgroup.

DSM-5 vs. DSM-IV vs. ICD-11

The same patient can receive different diagnoses under different frameworks. This is not merely academic — it has direct implications for cross-study prevalence comparisons, insurance coding, and international research collaboration.

DSM-IV (pre-2013): Two separate diagnoses — alcohol abuse (1+ of 4 criteria: role failure, hazardous use, legal problems, social/interpersonal problems) and alcohol dependence (3+ of 7 criteria including tolerance, withdrawal, and loss of control). A patient could have alcohol abuse or dependence, but not both simultaneously. The legal problems criterion was included.

DSM-5 (2013–present): Single dimensional diagnosis with 11 criteria, severity graded by count. Legal problems removed; craving added. The abuse/dependence distinction eliminated. Some former DSM-IV abuse-only cases now qualify as mild AUD; some former dependence cases may be reclassified at different severity tiers.

ICD-11 (current): Uses the categories of "harmful alcohol use" (a pattern causing health damage without dependence) and "alcohol dependence" (a cluster of behavioral, cognitive, and physiological phenomena). The ICD-11 dependence concept aligns more closely with DSM-IV dependence than with DSM-5's dimensional model. For researchers citing global prevalence data or conducting cross-national studies, this framework difference is consequential.

For prevalence researchers: Studies using DSM-IV criteria and studies using DSM-5 criteria are not directly comparable. The RADAR cohort study illustrates this concretely: DSM-5 lifetime AUD prevalence in their young adult sample was 18.4%, but this figure is heavily influenced by the tolerance criterion ^[4]. Under ICD-11 criteria, the same cohort would likely show different prevalence estimates. The corpus reviewed here does not contain direct empirical comparisons of DSM-IV versus DSM-5 prevalence in the same population — a significant gap for epidemiological work ^[3] (Note: this specific figure could not be independently verified against the source abstract — the underlying study supports the general finding but the exact number should be confirmed before publication).

Screening Tools — AUDIT

The Alcohol Use Disorders Identification Test (AUDIT) is a 10-item self-report questionnaire developed by the World Health Organization (WHO) and remains the best-validated screening instrument for identifying AUD in adults. It is critical to state clearly at the outset: the AUDIT is a screening tool, not a diagnostic instrument. A positive AUDIT score identifies individuals who warrant further diagnostic assessment; it does not establish an AUD diagnosis.

Scoring: Each item is scored 0–4, for a maximum total of 40. Standard cutoffs:
- Score ≥8: Hazardous or harmful drinking; warrants further assessment
- Score ≥15: Likely alcohol dependence in most adult populations; warrants urgent clinical attention

Performance for identifying DSM-5 AUD: An AUDIT score ≥8 carries a positive likelihood ratio of 6.5 (95% CI: 3.9–11) for AUD ^[9]. Notably, performance differs by sex: the likelihood ratio is 6.9 in females and 3.8 in males (p = .003) ^[9]. This sex difference is clinically important — the AUDIT performs better as a screening tool for AUD in women than in men, which may reflect differences in how alcohol-related harms manifest or are reported across sexes.

The AUDIT covers three domains: alcohol consumption (items 1–3), drinking behavior and dependence (items 4–6), and alcohol-related harms (items 7–10). This breadth is both its strength and its limitation in busy clinical settings — 10 items takes longer to administer than briefer alternatives.

Screening Tools — AUDIT-C

The AUDIT-C is a three-item consumption-only version of the AUDIT, comprising the first three questions about frequency of drinking, typical quantity, and frequency of heavy episodic drinking. It is widely embedded in electronic health record (EHR) systems, including the Veterans Health Administration (VHA), because of its brevity.

Cutoffs (note: population-specific):
- ≥3 (women) or ≥4 (men): Identifies hazardous drinking in most primary care populations
- ≥9: Associated with likely dependence in some health system validation studies

Performance: In a study comparing AUDIT-C to the TAPS (Tobacco, Alcohol, Prescription medication, and other Substance use) tool for identifying AUD, the AUDIT-C showed an area under the curve (AUC) of 0.90 for both females and males, with sensitivity/specificity of 0.83/0.83 for females and 0.81/0.84 for males ^[10]. The TAPS tool, while acceptable (AUC ~0.82–0.84), performed meaningfully worse than AUDIT-C for alcohol-specific screening ^[10].

The critical limitation: Despite strong AUC performance for identifying AUD, the AUDIT-C is substantially less useful than the full AUDIT for AUD identification specifically — positive likelihood ratios of only 1.8 (males) and 2.0 (females) ^[9]. This is a meaningful distinction. The AUDIT-C is an efficient tool for flagging unhealthy alcohol use patterns; it is not an adequate substitute for the full AUDIT when the clinical question is specifically whether AUD is present.

The screening-to-diagnosis gap: The AUDIT-C's widespread EHR deployment has not translated into proportionate increases in AUD diagnosis rates. Among adults with unhealthy alcohol use identified via AUDIT-C screening, only 10.1% received a formal AUD diagnosis ^[1]. The AUDIT-C fires; the diagnostic follow-through rarely happens. This is the central implementation failure the evidence documents — and it is not a failure of the screening tool itself, but of the clinical workflow that should follow a positive screen.

Screening Tools — CAGE, T-ACE, TWEAK, and Others

CAGE

The CAGE questionnaire (Cut down, Annoyed, Guilty, Eye-opener) is a four-item instrument that has been widely used in clinical settings for decades. It is rapid — administrable in under a minute — and requires no scoring beyond counting "yes" responses.

Cutoff: ≥2 positive responses is the standard threshold for a positive screen.

Limitations: The CAGE was developed and validated primarily for identifying alcohol dependence, not hazardous drinking without dependence. In the DSM-5 era, where mild AUD (2–3 criteria) represents a clinically meaningful category, the CAGE's insensitivity to hazardous drinking without dependence features is a significant limitation. It is less appropriate than the AUDIT for primary care universal screening. The corpus reviewed here does not contain direct CAGE sensitivity/specificity data against DSM-5 criteria — clinicians should be cautious about applying older CAGE validation data to current diagnostic standards.

T-ACE and TWEAK

T-ACE (Tolerance, Annoyed, Cut down, Eye-opener) and TWEAK (Tolerance, Worried, Eye-opener, Amnesia, K/Cut down) were specifically developed and validated for use in pregnant populations, where standard AUDIT cutoffs may not perform adequately and where the clinical stakes of missed identification are particularly high.

Both instruments weight tolerance heavily — reflecting that tolerance questions may be more sensitive than consumption questions for identifying problematic drinking in pregnancy, where social desirability bias may suppress self-reported quantity. These tools are the preferred screening instruments in obstetric settings.

MAST

The Michigan Alcoholism Screening Test (MAST) and its shortened versions (SMAST, Brief MAST) were developed for outpatient clinical settings and have a longer history than the AUDIT. The MAST focuses heavily on consequences and loss of control. It is less commonly used in contemporary primary care settings, where the AUDIT has largely supplanted it, but may still appear in older literature and some specialty settings.

Choosing the Right Tool

The right screening instrument depends on the clinical context and population:
- Universal adult primary care screening: AUDIT or AUDIT-C
- When time is severely constrained: AUDIT-C, with follow-up full AUDIT for positives
- Pregnancy: T-ACE or TWEAK
- Adolescents: CRAFFT (Car, Relax, Alone, Forget, Friends, Trouble) — see Special Populations section
- When AUD (not just hazardous use) is the specific clinical question: Full AUDIT preferred over AUDIT-C ^[9]

SBIRT — Screening, Brief Intervention, Referral to Treatment

SBIRT (Screening, Brief Intervention, and Referral to Treatment) is a public health framework — not a single tool — that links universal screening to proportionate clinical response. The U.S. Preventive Services Task Force (USPSTF) endorses SBIRT for unhealthy alcohol use in adults in primary care settings.

The three components:
- Screening: Universal application of a validated instrument (AUDIT, AUDIT-C) to identify unhealthy alcohol use
- Brief Intervention: A structured, time-limited counseling conversation (typically 5–15 minutes) for patients who screen positive but do not meet criteria for AUD — focused on feedback, goal-setting, and motivational enhancement
- Referral to Treatment: For patients who screen positive and meet AUD criteria, referral to appropriate specialty care, medication-assisted treatment, or behavioral health services

The logic of SBIRT is that the screening step is only valuable if it triggers a proportionate clinical response. A positive AUDIT-C that generates no follow-up action has no clinical benefit. Yet the evidence documents precisely this failure: ^[11] notes that USPSTF-recommended screening "is not always performed consistently or correctly in primary care," and even when screening occurs, translation to diagnosis and management is incomplete. The ^[12] Lancet review identifies "insufficient systematic screening in primary health care" as a key driver of undertreatment.

Implementation realities: The corpus does not contain time-motion data on visit-level implementation barriers, and the panel flagged this as a genuine evidence gap. What the evidence does show is that the diagnosis — the step between screening and treatment — is the rate-limiting bottleneck. The aOR of 10.68 for medication receipt among diagnosed versus undiagnosed patients ^[1] means that SBIRT's referral-to-treatment component is largely inaccessible without the diagnostic step that SBIRT's brief intervention component is not designed to provide.

Screening in Special Populations

Pregnancy

T-ACE and TWEAK are the validated instruments for pregnant populations, where standard AUDIT cutoffs have not been adequately validated and where social desirability bias may suppress consumption reporting. Any alcohol use in pregnancy warrants clinical attention; the screening question is not about hazardous thresholds but about any use and associated risk factors.

Adolescents

The CRAFFT (Car, Relax, Alone, Forget, Friends, Trouble) is the standard validated screening tool for adolescents aged 12–21. Standard AUDIT cutoffs are not validated for this age group. The RADAR cohort data are relevant here: tolerance was endorsed by 50.3% of regular drinkers aged 18–21 ^[4], suggesting that DSM-5 criteria — particularly tolerance — may function differently in adolescent and young adult populations than in adults. Clinicians should be cautious about applying adult diagnostic thresholds to adolescent presentations.

Older Adults

Alcohol pharmacokinetics change with age: decreased lean body mass, reduced total body water, and altered hepatic metabolism mean that older adults achieve higher blood alcohol concentrations at equivalent doses compared to younger adults. Standard AUDIT cutoffs may underestimate risk in this population. Modified cutoffs — typically lower thresholds for "hazardous" drinking — are recommended for adults over 65. The corpus reviewed here does not contain specific sensitivity/specificity data for modified AUDIT cutoffs in older adults — clinicians should consult age-specific guidelines.

Veterans and Military Personnel

The VHA has deployed AUDIT-C as a universal EHR-embedded screening tool across its system. VHA data show that AUD survey-based prevalence was 10.1% while EHR-documented diagnosis rates were only 6.0% — with the largest gaps among patients aged 18–34 (prevalence 22.4% vs. diagnosis rate 6.9%) and Hispanic/Latinx patients (prevalence 17.7% vs. diagnosis rate 7.6%) ^[13]. Even in a system with mandatory universal screening, the screening-to-diagnosis gap persists — underscoring that the problem is not screening tool availability but clinical follow-through.

Racial and Ethnic Considerations

Screening tool performance varies by population, and clinicians should be cautious about applying validation data from one population to another. The diagnostic disparities documented in the corpus are not random: females, racial/ethnic minorities, residents of economically deprived areas, and privately insured patients were all significantly more likely to remain undiagnosed ^[1]. These disparities at the diagnostic stage are then amplified at the treatment stage, because diagnosis is itself the primary gateway to pharmacotherapy. Groups least likely to be diagnosed are therefore doubly disadvantaged in accessing medication-assisted treatment.

Differential Diagnosis — Heavy Drinking vs. AUD

The diagnostic threshold for AUD is 2 of 11 criteria within a 12-month period. Below that threshold, a person may engage in heavy or hazardous drinking without meeting criteria for a disorder. This distinction matters clinically, legally, and for treatment planning.

NIAAA low-risk drinking guidelines define hazardous drinking as:
- More than 4 drinks on any single day, or more than 14 drinks per week (men)
- More than 3 drinks on any single day, or more than 7 drinks per week (women)

Drinking above these thresholds is associated with increased risk of AUD and alcohol-related harms, but does not itself constitute a diagnosis. A person who regularly exceeds these limits but endorses fewer than 2 DSM-5 criteria does not meet criteria for AUD — though they may benefit from brief intervention under the SBIRT framework.

When screening positive does not mean AUD: A positive AUDIT-C (≥3 for women, ≥4 for men) identifies hazardous drinking patterns, not AUD. The AUDIT-C's positive likelihood ratio for AUD specifically is only 1.8–2.0 ^[9] — meaning a positive AUDIT-C modestly increases the probability of AUD but is far from diagnostic. The full AUDIT (score ≥8, LR 6.5) is substantially more informative for the specific question of AUD ^[9].

The diagnostic step: The Canadian guideline explicitly recommends that positive screens be followed by assessment to distinguish at-risk drinking from AUD, and then to categorize AUD severity using DSM-5 criteria ^[14]. The Spithoff/Kahan framework reinforces this: screen all patients yearly, then determine AUD severity for those who screen positive ^[15]. That second step — the diagnostic step — is where clinical practice most consistently fails.

Diagnostic Equity and Recognition Gaps

The underdiagnosis of AUD is not randomly distributed. The evidence documents systematic patterns of who gets missed.

By age: In the VHA cohort, patients aged 18–34 had an AUD survey-based prevalence of 22.4% but a clinical diagnosis rate of only 6.9% — the largest absolute gap of any demographic subgroup ^[13]. Young adults are simultaneously the highest-prevalence group and the least likely to be diagnosed.

By ethnicity: Hispanic/Latinx patients in the VHA cohort had a survey-based AUD prevalence of 17.7% but a diagnosis rate of only 7.6% ^[13]. In the All of Us cohort, racial/ethnic minorities were significantly more likely to remain undiagnosed ^[1].

By sex: Females were significantly more likely to be undiagnosed in the All of Us cohort ^[1], despite the AUDIT performing better in females than males (LR 6.9 vs. 3.8) ^[9]. This suggests the diagnostic gap for women is not a screening tool problem but a clinical recognition and follow-through problem.

By insurance status: Privately insured patients were more likely to be undiagnosed than publicly insured patients ^[1] — a counterintuitive finding that may reflect differences in clinical encounter structure, stigma, or provider assumptions about who is "at risk."

The treatment amplification effect: Because receiving a diagnosis is associated with aOR = 10.68 for medication receipt ^[1] (Note: this specific figure could not be independently verified against the source abstract — the underlying study supports the general finding but the exact number should be confirmed before publication), equity gaps at the diagnostic stage are ampl

Verified References

• ^[10] Adam, Angéline, Laska, Eugene, Schwartz, Robert P et al. (2025). "Identifying Alcohol Use Disorder and Problem Use in Adult Primary Care Patients: Comparison of the Tobacco, Alcohol, Prescription Medication and Other Substance (TAPS) Tool With the Alcohol Use Disorders Identification Test Consumption Items (AUDIT-C).". Subst Use Addctn J. DOI: 10.1177/29767342251326678 [abstract-verified: partial]
• ^[12] Carvalho, Andre F, Heilig, Markus, Perez, Augusto et al. (2019). "Alcohol use disorders.". Lancet. DOI: 10.1016/s0140-6736(19)31775-1 [abstract-verified: yes]
• ^[11] Grissom, Maureen O, Reed, Brian C, Starks, Steven M et al. (2024). "Addiction Medicine: Alcohol Use Disorder.". FP Essent. [abstract-verified: yes]
• ^[7] Hallgren, Kevin A, Matson, Theresa E, Oliver, Malia et al. (2022). "Practical assessment of DSM-5 alcohol use disorder criteria in routine care: High test-retest reliability of an Alcohol Symptom Checklist.". Alcohol Clin Exp Res. DOI: 10.1111/acer.14778 [abstract-verified: partial]
• ^[8] Kerr, Jessica A, Husin, Hanafi Mohamad, Leung, Janni et al. (2025). "The natural history of DSM-5 alcohol-use disorder from late adolescence to middle adulthood in Australia: a prospective cohort study.". Lancet Public Health. DOI: 10.1016/s2468-2667(25)00225-7 [abstract-verified: yes]
• ^[5] Livne, Ofir, Feinn, Richard, Knox, Justin et al. (2022). "Alcohol withdrawal in past-year drinkers with unhealthy alcohol use: Prevalence, characteristics, and correlates in a national epidemiologic survey.". Alcohol Clin Exp Res. DOI: 10.1111/acer.14781 [abstract-verified: yes]
• ^[2] Miller, Alex P, Kuo, Sally I-Chun, Johnson, Emma C et al. (2023). "Diagnostic Criteria for Identifying Individuals at High Risk of Progression From Mild or Moderate to Severe Alcohol Use Disorder.". JAMA Netw Open. DOI: 10.1001/jamanetworkopen.2023.37192 [abstract-verified: yes]
• ^[6] Scalco, Matthew D, Lorenzo-Luaces, Lorenzo, Evans, Miranda et al. (2022). "Conceptualization of Alcohol Use Disorder (AUD): Can Theoretical or Data Driven Approaches Improve the Construct Validity of AUD?". Res Child Adolesc Psychopathol. DOI: 10.1007/s10802-022-00965-7 [abstract-verified: partial]
• ^[4] Slade, Tim, Mewton, Louise, O'Dean, Siobhan et al. (2021). "DSM-5 and ICD-11 alcohol use disorder criteria in young adult regular drinkers: Lifetime prevalence and age of onset.". Drug Alcohol Depend. DOI: 10.1016/j.drugalcdep.2021.109184 [abstract-verified: yes]
• ^[15] Sheryl Spithoff, Meldon Kahan (2015). "Primary care management of alcohol use disorder and at-risk drinking: Part 1: screening and assessment.". Canadian family physician Medecin de famille canadien. [abstract-verified: partial]
• ^[13] Williams, Emily C, Fletcher, Olivia V, Frost, Madeline C et al. (2022). "Comparison of Substance Use Disorder Diagnosis Rates From Electronic Health Record Data With Substance Use Disorder Prevalence Rates Reported in Surveys Across Sociodemographic Groups in the Veterans Health Administration.". JAMA Netw Open. DOI: 10.1001/jamanetworkopen.2022.19651 [abstract-verified: yes]
• ^[14] Wood, Evan, Bright, Jessica, Hsu, Katrina et al. (2023). "Canadian guideline for the clinical management of high-risk drinking and alcohol use disorder.". CMAJ. DOI: 10.1503/cmaj.230715 [abstract-verified: yes]
• ^[9] Wood, Evan, Pan, Jeffrey, Cui, Zishan et al. (2024). "Does This Patient Have Alcohol Use Disorder?: The Rational Clinical Examination Systematic Review.". JAMA. DOI: 10.1001/jama.2024.3101 [abstract-verified: yes]
• ^[1] Yue, Yihua, Rothberg, Michael B, Back, Sudie E et al. (2026). "Rates of Diagnosis and Treatment for Alcohol Use Disorder Among All of Us Participants with Unhealthy Alcohol Use.". J Gen Intern Med. DOI: 10.1007/s11606-025-10089-5 [abstract-verified: yes]

Replacement Resolution Audit

Each REPLACE verdict from the adjudication pass was resolved by re-querying the indexed fulltext corpus and selecting the highest-scoring paper that the Level 3 verifier confirmed supports the claim.

• ^[16] → ^[9] (verifier: partial; score 0.66). Title: Cardiovascular Risk Profile on the Island of Santiago-Cabo Verde (PrevCardio.CV Study).
• ^[17] → ^[1] (verifier: partial; score 0.74). Title: Clinical characteristics and 6-month follow-up of adults with and without alcohol use disorder who self-harm.
• ^[17] → ^[18] (verifier: partial; score 0.69). Title: Implementing SBIRT (Screening, Brief Intervention and Referral to Treatment) in primary care: lessons learned from a mul
• ^[17] → ^[19] (verifier: partial; score 0.79). Title: Screening for high-risk drinking and alcohol use disorder: update of the 2023 national clinical practice guideline.
• ^[20] → ^[2] (verifier: partial; score 0.69). Title: The global burden of disease attributable to alcohol and drug use in 195 countries and territories, 1990-2016: a systema
• ^[16] → NO REPLACEMENT FOUND (considered 5 candidates; none verified)
• ^[21] → ^[10] (verifier: partial; score 0.80). Title: Problem Opioid Use Among US Military Veterans: Prevalence, Correlates, and Psychiatric Characteristics.
• ^[22] → ^[15] (verifier: partial; score 0.61). Title: Identifying and Treating Metabolic Dysfunction-Associated Steatotic Liver Disease Among At-Risk Veterans.