Emergency department initiation of pharmacotherapy for alcohol use disorder: A retrospective cohort study

Alcohol misuse is a major public health issue in the United States. Alcohol use disorder (AUD), the most severe form of alcohol misuse, carries a prevalence of approximately 8%, and there are over 90,000 yearly alcohol-related deaths in the United States.¹ Persons with AUD utilize emergency departments (EDs) at disproportionately high rates,¹ and thus the ED represents a critical opportunity to provide intervention and treatment to this high-risk population.²

The ED is increasingly recognized as an optimal setting to initiate medications for AUD (mAUD).³ The U.S. Food and Drug Administration (FDA) has approved naltrexone, acamprosate, and disulfiram for the treatment of moderate or severe AUD, though the data to support their use are largely derived from non-ED, outpatient studies, and FDA approval does not mandate study across diverse clinical settings.⁴ A number of other medications, including gabapentin and topiramate, have demonstrated modest treatment benefit,³ though the strongest efficacy data are for naltrexone and acamprosate, which are considered first-line agents.^{3, 4}

Emerging evidence suggests mAUD initiation in the ED is feasible^{5, 6} and may be associated with improved patient outcomes, including reduced alcohol consumption and higher quality of life.⁵ The American Academy of Emergency Medicine recommends considering mAUD in all willing patients.³ However, few EDs have established screening and intervention protocols for AUD.⁷ Moreover, little research has been done to characterize current prescribing patterns in the ED setting. We therefore conducted a retrospective cohort study examining practice patterns for prescription of mAUD in our high-volume, academic ED. Our primary objective was to determine the prevalence of initiation of mAUD and the prescribing patterns for the medications used for this purpose.

This project was approved by the Washington University in St. Louis Institutional Review Board. We enrolled a convenience sample of encounters from July 2018, when our institution began using Epic as its electronic health record (EHR), to April 2023. All encounters involving patients ≥ 18 years old discharged from the Barnes Jewish Hospital ED during the study period with a diagnosis consistent with Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) moderate or severe AUD were eligible for inclusion. Encounter diagnoses were selected by physicians as part of routine care in the Epic EHR using the Intelligent Medical Objects (IMO) physician-friendly Interface Terminology. We leveraged the IMO mapping between the IMO Interface Terminology and the U.S. edition of Systematized Nomenclature of Medicine Clinical Terms (SNOMED CT) to identify eligible encounters.⁸ We chose the SNOMED CT concept “alcohol dependence” to connote moderate or severe AUD. All IMO terms that are mapped to one of the descendent concepts of “alcohol dependence” in the SNOMED-CT hierarchy were included.

Because opioid use precludes prescription of the opioid antagonist naltrexone, we also assessed for opioid use in our sample. We defined opioid use by (1) the presence of any diagnosis mapping to the SNOMED CT concepts of “opioid dependence,” “opioid abuse,” or “long-term current use of opiate-analgesic drug,” in the current encounter diagnoses, active problem list, or past medical history at the time of the included encounter or (2) any encounter diagnosis mapping to the previously referenced SNOMED CT concepts in the 6 months prior to the included encounter. Separately, we evaluated encounters for the presence of an opioid on the home medication list. The following variables were obtained from the EHR to characterize included encounters: full name, medical record number, date of birth, age, race, ethnicity, sex, insurance status, primary care provider (PCP), and home medications.

The primary outcome was the prevalence of initiation of the first-line agents naltrexone or acamprosate, including administration in the ED of extended-release intramuscular naltrexone (XR-NTX) or prescription upon discharge of XR-NTX, oral naltrexone, or acamprosate, among patients without a preexisting active prescription for these medications. Secondary outcomes included (1) the prevalence of ED prescription of any mAUD, including naltrexone, acamprosate, disulfiram, gabapentin, and topiramate to patients with or without a preexisting prescription and (2) evidence of social work involvement in the encounter, as documented by a consult order or a social work note. As multiple indications exist for most mAUD, encounter notes were manually reviewed to confirm the indications for prescriptions written in the ED.

We identified a total of 631 encounters, corresponding to 509 unique patients. Table 1 demonstrates demographics, health characteristics, and encounter details. The median age at encounter was 50 years, and 22.2% involved female patients. The majority of encounters involved Black (54.8%) or White (38.5%) patients, and only 2.1% involved Hispanic patients. The most common insurance was Medicaid (46.6%). A total of 52.9% of encounters involved patients with a documented PCP. With respect to opioids, 91 encounters (14.4%) involved patients with a diagnosis suggestive of opioid use, and 38 encounters (6.0%) involved patients with an opioid on their home medication list.

A total of 76 encounters (12.0%) involved patients already taking at least one medication that could be indicated for AUD treatment, 35 of which (5.5% of all encounters) were a first-line mAUD. Forty-five encounters (7.1%) involved a preexisting prescription for gabapentin, 24 (3.8%) for naltrexone, and 11 (1.8%) for acamprosate. There were two (0.3%) preexisting prescriptions for topiramate and none for disulfiram.

Oral naltrexone was initiated for AUD treatment during four encounters (0.6%) and refilled for AUD treatment four times (0.6%); acamprosate was initiated once (0.2%). In addition, one patient (0.2%) with a preexisting prescription for acamprosate was prescribed gabapentin as an adjunct mAUD, and two patients (0.3%) received gabapentin refills for treatment of neuropathy. There were no instances in which XR-NTX was administered or prescribed. Only 13.3% of encounters involved an order for a social work consult.

In summary, in this large retrospective study of more than 600 AUD-related ED visits, although 88% of encounters involved patients previously untreated for AUD, fewer than 1% were associated with initiation of a recommended first-line mAUD. Initiation rates were affected to some degree by patient characteristics; some of our patients were already on mAUD, and, in many of our patients, the first-line agent naltrexone was contraindicated by active opioid use. However, given the high mortality associated with untreated AUD, our results still highlight a critical missed opportunity in the ED to engage this vulnerable population in care.^{1, 2}

Importantly, while most ED directors support screening and intervention for AUD,⁷ justification for prescribing mAUD in the ED is largely extrapolated from data in support of this practice in outpatient settings⁹; questions remain as to whether the small existing body of literature on the efficacy of mAUD initiated in the ED is sufficient to justify widespread adoption of this practice. Namely, to our knowledge, only one small study provides direct evidence that initiation of naltrexone in the ED leads to a meaningful reduction in drinking,⁵ and there is no direct evidence to support ED initiation of acamprosate or gabapentin for AUD.¹⁰

In practice, EM physicians cite several barriers to initiating mAUD, including limited familiarity with prescribing, difficulty identifying appropriate candidates, a sense of futility, and stigma.¹¹ The absence of clear benefit for mAUD in the ED may also be an impediment to practice adoption.¹² However, though ED-specific evidence is limited, literature supports mAUD in other settings,⁴ and the overall risk of harm associated with first-line mAUD, when appropriately prescribed, is low.³ The recent ED guidelines published by Strayer and colleagues,³ therefore, favor initiation of mAUD. Unfortunately, the current guidelines do not adhere to a rigorous methodological construct such as the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) or Grading of Recommendations Assessment, Development, and Evaluation (GRADE) frameworks.^{13, 14} There exists a need for more rigorous ED-specific clinical guidelines on mAUD initiation, which may provide clearer guidance and help overcome some of the knowledge-related barriers to practice uptake. Further research is needed to identify whether mAUD initiation in the ED elicits a meaningful improvement in ED-specific outcomes.

In addition, although the retrospective, chart-based nature of this study cannot account for informal involvement of social workers in patient care, our results suggest low social work utilization. Importantly, previous literature suggests that, even when a formal ED algorithm for AUD treatment is implemented, most AUD patients fail to attend follow-up visits.⁶ To this end, increased or protocolized social work utilization may represent an additional area of improvement in providing optimal care for AUD patients in the ED.

There are several limitations to this study. First, the absence of a standardized screening protocol in our ED posed a challenge in developing an appropriate retrospective definition for AUD. We opted to define a cohort with moderate-to-severe AUD, as there is high agreement between common clinical diagnostic systems for this population; the DSM-5 definition of severe AUD demonstrates high concordance with the ICD-10 diagnosis of alcohol dependence, and most cases of moderate AUD are also classified as alcohol dependence by ICD-10.¹⁵ While this approach likely underestimated AUD encounters in our ED, it reduced the risk of capturing alcohol-related ED encounters that did not represent true AUD. Moreover, our approach highlights practice patterns for treatment of those patients in whom there is less question of therapeutic candidacy and treatment benefit. Second, our analyses of home medication lists are based on the information available in our EHR. Especially for patients who typically seek care outside of our hospital system, the home medication lists tied to ED visits may be incomplete or inaccurate. Consequently, preexisting home mAUD or home opioids may be underestimated. Finally, it is difficult to characterize medication indications retrospectively. Excepting acamprosate, multiple indications exist for the included medications. For prescriptions initiated in the ED, encounter notes were manually reviewed to confirm indications. However, it was beyond the scope of this letter to establish indications for preexisting home prescriptions. For example, several of the included encounters (7.1%) involved patients with preexisting prescriptions for gabapentin, which has multiple FDA-approved indications. Although gabapentin can be used off-label for AUD, most gabapentin prescriptions on home medication lists likely reflect treatment of comorbid conditions.

Our results highlight a missed opportunity to engage persons with AUD in care. In light of our results, next steps for our institution may include implementation of quality-improvement measures to establish a protocolized approach for targeted screening and pharmacologic treatment of AUD. For the field at large, further prospective trials are needed to elucidate the benefits of initiating mAUD in the ED, identify optimal candidates, optimize relevant protocols to overcome treatment barriers and characterize the long-term outcomes of this intervention, such as continuation and outpatient follow-up rates.

AUTHOR CONTRIBUTIONS

Karlee S. De Monnin contributed to conception and study design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, and critical revision of the manuscript for important intellectual content. Emily Terian contributed to study design, analysis and interpretation of data, drafting of the manuscript, and critical revision of the manuscript for important intellectual content. Julianne Yeary contributed to study design and critical revision of the manuscript for important intellectual content. Elizabeth Bathon contributed to study design and critical revision of the manuscript for important intellectual content. Phillip Asaro contributed to conception and study design, acquisition of data, analysis and interpretation of data, and critical revision of the manuscript for important intellectual content. Carrie M. Mintz contributed to conception and study design, analysis and interpretation of data, drafting of the manuscript, critical revision of the manuscript for important intellectual content, and acquisition of funding. Kevin Baumgartner contributed to conception and study design, analysis and interpretation of data, drafting of the manuscript, critical revision of the manuscript for important intellectual content, and study supervision.

FUNDING INFORMATION

Funded by grant K08 AA029714 (CMM).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.