Hospital Algorithm Screened 9,140 Records to Start 23 on Buprenorphine

An automated screening algorithm at two Bronx teaching hospitals flagged 9,140 patient records over two years, but only 23 hospitalized patients started buprenorphine as a result — a 5 percent initiation rate that the researchers present as a caution for health systems counting on electronic health records to close the addiction treatment gap.

The findings appear in a brief report published October 8 in Addiction Science & Clinical Practice, an open-access journal, by a team from Albert Einstein College of Medicine and Montefiore Medical Center led by first author Abigail Haber and corresponding author Aaron D. Fox. The paper reports screening data from a randomized trial of in-hospital buprenorphine initiation strategies registered on ClinicalTrials.gov as NCT 05118204 and funded by the National Institutes of Health under award RM1DA055437.

A funnel that narrows at every stage

Recruitment ran from October 2022 to October 2024 at two urban teaching hospitals that share the same Epic electronic health record system. The research team built an algorithm that produced a daily list of hospitalized patients with likely opioid misuse or opioid use disorder, drawing on 12 drug-related terms in clinical notes, positive urine toxicology, relevant diagnosis codes, prior prescriptions for addiction medication, and documentation of withdrawal treatment.

Over the two-year window the algorithm generated 9,140 records. Staff excluded 8,534 of them, or 93 percent, during a preliminary manual chart review. The most common reason was that the record did not actually show opioid misuse, accounting for 43 percent of the exclusions — a misidentification rate the authors attribute to the algorithm's inability to distinguish a real mention from a negated or qualified one.

Another 21 percent of excluded patients were already receiving medication for opioid use disorder before admission, and 16 percent failed study-specific criteria such as documented low blood oxygen.

From 434 eligible patients to 23 prescriptions

After manual review and a second pass by an addiction specialist, 434 patients remained potentially eligible. Staff successfully approached 201 of them. Of those, 14 percent declined any addiction services and 48 percent declined buprenorphine specifically, according to the report.

Forty-three patients agreed to be assessed. Twenty-nine were found eligible for the trial, and 23 enrolled and started buprenorphine. Each of the 434 patients who could have benefited represents a hospitalization that ended without treatment, the authors note.

Stigma and logistics, not just software

The paper's conclusion cuts both ways. On the technical side, the authors suggest that better algorithm precision — including natural language processing and artificial intelligence tools that can separate affirmed from denied drug use — could sharply cut the manual review burden that consumed staff time.

On the human side, they are blunt. "Patient concerns, and potential stigma regarding BUP treatment, could be ongoing challenges to BUP uptake," the authors write. They also point to the practical difficulty of enrolling acutely ill patients who may be discharged before an addiction consultation happens, a constraint that surfaces repeatedly in hospital-based addiction care.

Buprenorphine is a partial opioid agonist used to treat both opioid use disorder and chronic pain, and it carries a lower overdose risk than full agonist opioids. The authors note that prescriptions for the medication have plateaued in the United States even as opioid use disorder prevalence has risen.

Why hospital initiation matters

Clinical guidelines increasingly push hospitals to start the medication before discharge, because patients who leave without treatment face elevated risks of overdose and readmission. The Bronx experience suggests that technology alone will not deliver that change.

The report was approved by the Einstein Institutional Review Board, and the authors declare no competing interests beyond one co-author's royalties for an online chapter on primary care for patients with opioid use disorder. People seeking medication for opioid use disorder can find treatment programs and opioid rehab providers across the country.