One in Nine Hospitalized for Alcohol Use Disorder Leaves With a Prescription

The medication is generic, inexpensive and already sitting in the hospital pharmacy. It reduces heavy drinking in the weeks after discharge, which is exactly when the risk of relapse and readmission runs highest. And in a review of 4,232 hospitalizations published Oct. 22 in the Journal of General Internal Medicine, it went home with roughly one patient in nine.

Neela Nataraj and John T. Ratelle of Mayo Clinic in Rochester, Minnesota, and colleagues reviewed every adult discharged with an alcohol use disorder diagnosis from two hospitals inside a large academic medical center between 2018 and 2023. In 467 of those hospitalizations — 11 percent — the patient left with a new prescription for acamprosate, disulfiram or naltrexone in its oral or long-acting injectable form.

What the team did next is the reason the paper is worth reading. They modelled which patients got the prescription, expecting to find a clinical profile: the sicker patients, the ones with withdrawal seizures, the ones already familiar with treatment. Instead, almost nothing about the patient predicted anything. What predicted prescribing was the shape of the hospital stay.

The Cohort Behind the Eleven Percent

The two hospitals shared one important characteristic: neither had a dedicated inpatient addiction medicine consult service. That single detail frames the entire result. The study describes what happens to hospitalized drinkers in the ordinary architecture of American hospital medicine, where an alcohol use disorder is coded, managed for withdrawal, and then released with a referral sheet.

The cohort behind the number is not small or marginal. Across 2,347 patients and those 4,232 admissions, alcohol-related diagnoses appeared often enough that the authors describe the condition as pervasive among hospitalized adults. The outcome was new initiation — a first prescription written at discharge — and the analysis used generalized estimating equations to avoid counting the same patient's repeated admissions as independent observations.

That design choice matters for how the headline reads. Eleven percent is the rate of starting medication, not the share of patients who left medicated. Anyone already taking naltrexone or acamprosate on admission falls outside the numerator, and patients whose AUD was never documented at all fall outside the denominator, since the cohort was assembled from diagnosis codes rather than screening. The real treatment gap is wider than the paper's main figure, not narrower.

The One Patient Characteristic That Mattered

Of all the patient-level variables the researchers tested, exactly one survived adjustment: whether alcohol use disorder was the primary reason for admission rather than a secondary note on a chart. Those patients had 1.25 times the odds of a new prescription, with a confidence interval running from 1.02 to 1.54.

That is a modest effect. It is also telling. When AUD is the admitting diagnosis, the hospital's attention is on alcohol by construction — the withdrawal protocol gets ordered, the social work note gets written, the question of what happens next gets asked out loud. When AUD appears as the fourth item on a problem list behind cellulitis and heart failure, the same treatment decision never gets made, even though the discharge risk is identical.

Everything else the authors measured about the person — age, sex, comorbidity, prior treatment history, insurance — dropped out of the model. This is a study about a prescribing habit, and habits belong to organisations, not to patients.

Why Three Days Beat Eight

Length of stay produced the study's most counterintuitive curve. Relative to stays of two days or less, admissions lasting three to four days carried 1.70 times the odds of a discharge prescription (95% CI 1.33–2.17). Stays of five to seven days dropped to 1.47 (1.08–2.02). Stays of eight days or more fell to 1.37, and the confidence interval there crossed one (0.98–1.91).

If the mechanism were simply "more time in the building means more chances to prescribe," the odds should climb with every additional day. They do not. They peak in the middle and then decline, which suggests something more specific: the three-to-four-day admission is where alcohol withdrawal management finishes and discharge planning begins, and where a clinical team still has a window to act on the problem it was admitted to treat. The eight-day admission is usually long for another reason — a fall, an infection, a surgical complication — and by the time that patient leaves, the alcohol question has been buried under the reason for the extended stay.

What the Adjusted Model Actually Implies

The authors do not push this interpretation, but the pattern is consistent with a well-documented feature of hospital care: the discharge prescription is the product of a pathway, not a decision. Pathways get triggered by admitting diagnoses. They do not get triggered by problem-list entries. A patient admitted for alcohol withdrawal walks through a door marked "withdrawal protocol," and inside that door someone has thought about medication. A patient admitted for pneumonia who also drinks heavily walks through a different door.

The Service That Discharged You Mattered More Than Your Diagnosis

Two organisational variables moved the number more than any patient factor. Consultation by care management raised the odds of a new prescription by 1.37 (1.02–1.85). Discharge from a general medicine teaching service raised them by 1.41 (1.09–1.83), and discharge from a family medicine service raised them by 2.07 (1.52–2.81), each measured against a hospital internal medicine non-teaching team.

The family medicine result, at more than double the reference group, is the most interesting number in the paper. Family medicine teams in the United States are built around continuity: the resident or attending discharging the patient often expects to see that patient again in clinic, sometimes within a week. A hospitalist team has no such expectation. It hands the patient to a primary care system it may never interact with, and the prescription becomes a message in a bottle.

The interpretation is not proven by this dataset — the authors describe associations, and a retrospective chart review cannot distinguish a culture of prescribing from a case mix that happens to favour it. But the direction fits the rest of the findings. Care management consultation brings a person whose job is to plan what happens after discharge. Family medicine brings a clinic that will actually see the patient. Both are substitutes for the thing the two study hospitals did not have: an addiction medicine consult service whose entire function is to make this decision for every patient with the diagnosis, on every team, regardless of which door the patient came through.

Where Medication Access Already Stalls

The 11 percent figure lands on top of a national picture that has been flat for years. A cross-sectional study published in January in JAMA Network Open tracked every county in the United States from 2017 to 2023 and found that the share with at least one substance use disorder treatment facility offering acamprosate, disulfiram or naltrexone rose from 34.1 percent in 2017 to 43.9 percent in 2021 — and then stopped rising. Counties adjacent to metropolitan areas sat 22.4 percentage points below metropolitan counties in the probability of having any such facility, and remote rural counties 23.6 points below.

That study measured availability at facilities. The Mayo cohort measures a decision made at the moment of discharge, which is upstream of any facility question and cheaper to fix. Nobody has to build a clinic to write a prescription that the patient's own pharmacy already stocks.

What This Means for Treatment

For patients and families, the practical implication is that asking is not optional. Someone being discharged after an alcohol-related admission should ask directly whether a medication for alcohol use disorder is appropriate, and if the answer is no, why. The evidence for naltrexone and acamprosate is strong enough that a "no" deserves a reason, and the reason is frequently that no one on the team thought about it.

For hospitals, the paper's own conclusion is structural rather than educational. The authors point at service configuration and consultative support as targets — which in practice means either building an addiction consult service or, far more cheaply, hard-wiring the question into discharge order sets so that the default becomes a documented decision instead of an omission.

Where the Evidence Goes Next

Three limits bound what this study can claim. It is retrospective and observational, so every odds ratio describes association. It draws on a single academic medical center, and one without an inpatient addiction consult service, which means the 11 percent describes a specific kind of hospital rather than a national average — plausibly worse than centers with dedicated services, and plausibly better than community hospitals with fewer resources and thinner discharge planning. And the cohort was built from coded diagnoses, which undercount.

What survives all three limits is the shape of the finding: initiation of medication-assisted treatment after an alcohol-related hospitalization depends more on how the hospital is organised than on who the patient is. That is an uncomfortable finding for a health system that prefers to explain treatment gaps through patient complexity. It is also the more hopeful one, because service lines and order sets can be changed by a committee, and patients cannot.