Implementing Programs to Initiate Buprenorphine for Opioid Use Disorder Treatment in High Need, Low Resource Emergency Departments: A Nonrandomized Controlled Trial

Abstract

Objectives

We hypothesized implementation Facilitation (IF) would enable us to rapidly and effectively implement ED-initiated buprenorphine (BUP) programs in rural and urban settings with high-need, limited resources, and dissimilar staffing structures.

Methods

This multicenter implementation study employed IF using a participatory action research approach to develop, introduce, and refine site-specific clinical protocols for ED-initiated BUP and referral in three EDs not previously initiating BUP. We assessed feasibility, acceptability, and effectiveness by triangulating mixed-methods formative evaluation data (focus groups/interviews and pre-/post- surveys involving staff, patients, and stakeholders), patients’ medical records, and 30-day outcomes from a purposive sample of 40 BUP-receiving patient-participants who met research eligibility criteria (English-speaking, medically stable, locator information, non-prisoners). We estimated the primary implementation outcome (proportion receiving ED-initiated BUP amongst candidates) and main secondary outcome (30-day treatment engagement) using Bayesian methods.

Results

Within 3 months of initiating IF activities, each site implemented BUP programs. During the 6-month programmatic evaluation, there were 134 ED-BUP candidates amongst 2,522 encounters involving opioid use. 52 (41.6%) practitioners initiated BUP to 112 (85.1%; 95%CI 79.7%–90.4%) unique patients. Amongst 40 enrolled patient-participants, 49.0% (35.6%–62.5%) were engaged in addiction treatment 30 days later (confirmed); 26 (68.4%) reported attending one or more treatment visits; there was a four-fold decrease in self-reported overdose events (OR 4.03; 95%CI 1.27–12.75). ED clinician readiness increased by a median of 5.02 (95%CI: 3.56–6.47) from 1.92/10 to 6.95/10 (n(pre)=80, n(post)=83).

Conclusions

IF enabled us to effectively implement ED-based BUP programs across heterogeneous ED settings rapidly, which was associated with promising implementation and exploratory patient-level outcomes.

Introduction

The opioid epidemic continues to decimate small and large communities with overdose death rates accelerating to an all-time high since the emergence of COVID-19.¹ Emergency department (ED) patients who survived an overdose were shown to have a 5% one-year mortality, which was reduced by 59% among those who received buprenorphine (BUP) or methadone in that timeframe.² However, only one-third received these medications for opioid use disorder (MOUD, OUD).² As the only source of care for many individuals with OUD and the point of healthcare entry for patients experiencing overdose, the ED is a critical venue to initiate interventions. D’Onofrio and colleagues’ randomized trial of 329 patients with OUD demonstrated the feasibility, safety, efficacy, and cost-effectiveness of initiating BUP in the Yale ED.^3,4 Importantly, that study was conducted in an urban, academic ED with robust addiction resources where the Principal Investigator (PI) served as chair and champion, the co-PI provided ongoing BUP treatment in the on-site primary care clinic, and study staff identified participants, scheduled appointments within 72 hours, and dispensed enough take-home medication to last until the appointment. The most effective methods for implementing these interventions across a variety of settings and less resourced conditions are unknown.

Despite that landmark study and considerable efforts to incorporate this life-saving treatment into standard ED practice, adoption remains limited.^5–7 Absent active external facilitation, the persistence of these norms reflects stigma and therapeutic pessimism as well as stringent prescribing regulations, diversion concerns, and inadequate training and referral options for MOUD.⁸ These implementation challenges are more pronounced where the majority of patients receive care – in community EDs, rural areas, critical access hospitals, and other settings without the resources found in large, academic research centers. This study, NIDA CTN-0079 (ED-CONNECT), was conceived to address the critical public health need to identify implementation strategies capable for producing changes in clinical practice that are practical and effective across heterogenous ED contexts and to accomplish this in an accelerated timeline for rapid dissemination. We hypothesized that by employing a participatory action research approach and Implementation Facilitation (IF) strategies, ED-initiated BUP program implementation in rural and urban settings with high-need, limited resources, and dissimilar staffing structures would be feasible and acceptable and accomplished rapidly, within one year.

Methods

Design

In this three-site implementation science study, we employed a systems-level IF intervention and a participatory action research approach to develop, implement, iteratively refine, and evaluate clinical programs for the initiation of BUP and referral for ED patients with untreated OUD in characteristically distinct EDs without pre-existing BUP programs. To inform site-specific clinical protocols and implementation strategies and to assess feasibility, acceptability, and effectiveness at the program and patient-levels, we triangulated mixed-methods data from the IF formative evaluation involving staff, patients, and other stakeholders with medical records of ED patients and a purposive sample of 40 patients who received ED-initiated BUP who met additional research eligibility criteria. The 6-month programmatic evaluation began upon the approval of sites’ clinical protocols and spanned September 2018–April 2019, which occurred 3 months following IRB approval and initial site visits for IF activities. The study was designed in accordance with the Standards for Reporting Implementing Studies (StaRI).⁹ Detailed study methods, including description of the formative evaluation involving patients, ED and community clinicians and staff, and other key informants, are published separately¹⁰. The study was IRB-approved with additional protections granted through a Certificate of Confidentiality and was monitored by an independent Data and Safety Monitoring Board.

Sites:

The study was conducted at Catholic Medical Center (CMC), Manchester, NH, Valley Regional Healthcare (VRH), Claremont, NH, and Bellevue Hospital Center (BHC), New York, NY. Rather than through the more traditional process whereby a motivated, local champion investigator applies to be a participating site, these sites were approached for participation because of their unique and complementary implementation challenges. Together, they represent heterogeneous settings, including community, critical access, and public EDs across rural to urban population densities with varying treatment resources. Site characteristics are shown in Table 1. All sites had limited ED resources for managing their socio-economically vulnerable, high-need OUD patient populations, including one without ED-based social work coverage and extremely limited outpatient MOUD referral options and another without an ED-based champion. Each had differing ED staffing structures, including one that was changing locum tenens staffing agencies during site selection. Further, each site used different electronic medical record (EMR) platforms and announced plans to change them during the study. Lastly, none offered ED-initiated BUP prior to the study; in fact, one ED had an active policy prohibiting it. Neither NH site had X-waivered clinicians; at BHC, the handful of waivered ED clinicians did not use it in their practice.

Table 1.

Site Characteristics

	Valley Regional Healthcare (VRH)	Catholic Medical Center (CMC)	Bellevue Hospital Center (BHC)
ED Volume	~11,000 patient-visits/year	~35,000 patient-visits/year	~100,000 patient-visits/year
Setting	Rural	Urban with suburban and rural catchment zone	Urban
Institution	Critical access community hospital	Community hospital	Public, academically-affiliated, tertiary care and trauma center
ED Space	No crowding	Significant overcrowding	Overcrowding common
ED patients OUD diagnosis (12-months, pre-study)	Abuse: 01 Dependence: 01	Abuse: 575 Dependence: 417	Abuse: 923 Dependence: 1472
ED OUD treatment infrastructure and practice (pre-study)	- No ED-initiation of BUP; non-agonist medications given for withdrawal - BUP not in ED Pyxis - No ED providers with an X-waiver - No Addiction consultants - No referral options known or provided to patients.	- No ED-initiation of BUP; non-agonist medications given for withdrawal - BUP not in ED Pyxis - Policy precluded BUP in ED - No ED providers with an X-waiver - No Addiction consults in ED - No referrals provided. Patients directed to fire departments for resources. Waitlists for treatment programs were often several months.	- No ED-initiation of BUP; methadone given for withdrawal - BUP not in ED Pyxis - 4 ED providers with an X-waiver, but didn’t initiate BUP in ED - No Addiction consults in ED (excluded) - Patients admitted to inpatient detoxification or given a resource list. - ED not permitted to schedule patients in any of BHC outpatient addition centers, which includes an opioid treatment program (OTP), dual-diagnosis clinic, and office-based program. Waitlist of 3+ weeks. - Vast addiction resources in NYC without centralized system to help navigate
Referral options (identified during study)	- A single outpatient addiction practice in the county with once weekly new patient intakes. Patients were required to attend group therapy, have negative drug screens, and transport self.	- A single office-based provider was established (in response to study-identified needs) and accepted weekday walk-ins. - One inpatient detoxification and sober living (did not initiate BUP). - Lengthy, indeterminant waitlist at most other facilities.	- Bellevue’s outpatient programs remained minimally accessible (described above). - External linkages were established with 3 treatment programs, often providing appointments within 0-3 days - BHC “detox” began continuing BUP initiated in the ED (vs. detoxification with methadone)
ED Staffing	Single coverage via core staff of 4 plus 5 per diem locums: (6 attendings, 3 APPs). Non-EM trained ED director served as champion. - No ED-based social work - No in-hospital addiction or psychiatric specialty coverage	100% Locums staffing with agency change during study with new director and new incoming semi-permanent physicians (11) and permanent APPs (6). - Limited social work support - Clinical pharmacist - External Pain/Addiction consultant for inpatients only	- Attending-supervised, resident-delivered care with 120 department-wide attending faculty (80 worked at least one shift at BHC during study), 60 EM residents plus off-service residents; 16 APPs. - Prolonged hiring freeze with nurse-to-patient ratios often >1:20, social work coverage gaps, and loss of clerks/staff - Volunteer health coaches conducted substance use screening and interventions. - Peers hired during study -Addiction consults for inpatients only
Space	No crowding	Significant overcrowding	Overcrowding common
Unique Features	Economically fragile hospital and community with constrained material and personnel resources; concerns expressed that a ED-based BUP program may further stress these. EMR changed during study twice. Extremely limited and high-barrier treatment options with once weekly intakes. Plan was for ED providers to obtain X-waivers to prescribe bridge treatment.	Many significant changes occurring in the ED, including the staff/director turnover and new EMR – resulting in resistance further change and diminished trust outside of ED. Locums agency did not encourage BUP in ED and hospital policy prohibited BUP in the ED. Non-ED hospital leader served as champion and supported hire of outpatient BUP provider to take referrals.	Local expertise and city resources exist but less accessible in ED. Large understaffed and fragmented health system and underdeveloped referral network. Attending and cross-department resident providers working at multiple clinical sites. Antiquated EMR with new additions (e.g., templates) temporally suspended ahead of planned EMR transition. Plan was to establish linkages for referrals and encourage ED providers to obtain x-waivers.

¹Absence of any patients receiving any opioid-related diagnoses reflects incomplete coding, which may also speak to the lack of emphasis given to OUD in the ED as well as personnel/resource limitations. ED director estimated at least 2-3 ED patients with OUD per week.

Populations:

IF formative evaluation participants:

Web-based surveys to assess site characteristics and readiness to initiate BUP were distributed by email link to active ED nurses, physicians, advance practice providers (APPs), social workers, and administrators both prior to the study IF intervention and following the programmatic evaluation period. ED directors produced the recipient list and departmental data. Focus groups and qualitative interviews with ED and community staff, patients, and other stakeholders are reported separately.

Patient population:

The EMRs for adult ED patients presenting to the ED sites during the programmatic evaluation period were examined to identify patients with non-medical opioid use and evaluate implementation via a waiver of consent. Detailed review was conducted for all patients triaged at VRH and CMC, and, at BHC, because of its high patient volumes, for all patients who had either a positive clinical screen for any substance use, an opioid-related diagnosis at past or current visits, a chief complaint related to substance use or one of several other pre-specified complaints, or had been administered or prescribed MOUD or naloxone during the ED visit. Operationally, patients classified as “screening positive” for non-medical opioid use included those who were formally screened (clinically) or had other indicators of opioid use/OUD identified during visits within the past 12 months (e.g., opioid-related diagnosis).

Patient-participant population:

We enrolled a purposive sample of Patient-Participants who received ED-initiated BUP via administration and/or prescription and who provided written informed consent to participate in baseline and 30-day research visits to assess patient-level outcomes. Patient-participants were required to speak English, have two methods of contact, and be medically-psychiatrically stable; they could not be prisoners, engaged in MOUD, pain management, or a substance use intervention study. Our sampling strategy ensured patient-participant enrollment would be relatively even over time and across sites with recruitment shifts including day, evening, and weekend hours. To minimize the risk of study activities influencing clinical care, research staff did not interact with patients or clinical staff for recruitment until after the ED visit was completed. Coordinators attempted to contact all potentially eligible patients who presented outside of recruiting hours for whom adequate contact information existed.

Interventions:

The systems-level research intervention consisted of a bundle of interconnected IF activities to support programmatic development and implementation, which are based on a manualized program developed by Kirchner and colleagues and are summarized in Table 2¹¹ IF activities began following regulatory approvals with the distribution of web-based surveys followed by initial site visits from study investigators (RM, KH) in July 2018 to conduct focus groups, qualitative interviews, and identify potential local champions. Serving as external facilitators, RM and KH made a total of five in-person visits to each site to engage with ED and community treatment program staff, patients, and other stakeholders, conduct academic detailing and trainings, and collaborate with local clinicians to develop, customize, and refine program components and resources. They led 24 learning collaborative calls.¹⁰

Table 2:

Implementation Facilitation Roles and Activities

Activity	Definition
Formative Evaluation	Quantitative and qualitative methods used to identify evidence, context, and facilitation-related factors to inform site-specific ED-initiated BUP with referral program implementation, to monitor progress, and to assess program acceptability and impact.
Advising on ED-initiated BUP Clinical Protocol Development and Tailoring to Site	External facilitators provided framework for clinical sites on the development of a clinical protocol for nonmedical opioid use screening, assessment, and BUP treatment initiation with referral. External facilitators collaborated with site personnel to tailor and iteratively refine individual components of each site’s clinical protocol and implementation strategy based on local resources, existing workflows and personnel roles, and other contextual factors to optimize program feasibility, acceptability, and impact.
Assistance Developing Referral Plan	External facilitators collaborated with ED stakeholders to identify community OUD treatment providers and create practical workflows for referral.
Stakeholder Engagement	Engagement at the administrative, clinician, community, and patient levels began with and was informed by the initial formative evaluation. Goals included identifying local champions and aligning interests to support normalization of OUD care in the ED.
Clinician Education and Academic Detailing	All ED providers were offered educational sessions on OUD and BUP training as well as logistical support on practical issues, such as efficient use of the EMR.
Performance Monitoring and Feedback	External facilitators worked with ED leaders to incorporate clinician performance related to BUP-initiation and facilitated referral into the department’s standard quality improvement and feedback practices.
Learning Collaborative	ED stakeholders were invited to a weekly Learning Collaborative to promote shared learning regarding issues promoting and hindering implementation of addiction treatment.

Site-specific clinical protocols and implementation strategies were refined throughout the entirety of the study to improve programmatic feasibility, acceptability, and effectiveness using the Rapid Assessment Process.¹² This form of participatory action research is an intensive, team-based approach, involving cycles of gathering information, planning and implementing changes, and collecting feedback to inform subsequent revisions. Study staff maintained IF logs in which field notes and summaries from planned IF activities (e.g., interviews/focus groups, academic detailing) as well as ad-hoc clinical staff feedback and other observations were recorded allowing “real-time” review at least weekly. This process allowed us to explore and test modifications to how, when, and by whom clinical actions are performed across these disparate ED settings and to disseminate information expeditiously.

The clinical intervention (i.e., BUP initiation and referral) was delivered entirely by and at the discretion of ED clinicians, rather than as a research procedure. Clinical protocols incorporated critical actions similar to those previously tested by D’Onofrio et al. (described below) and were customized to sites’ local context, resources, and workflows.³

Measures/Outcomes

The overarching primary outcome was successful programmatic implementation at each site, requiring clinical programs to be developed, approved by both study external facilitators and site directors, and introduced at each site within the study timeframe. Approval required clinical protocols to incorporate the following critical actions: identifying patients with non-medical opioid use, assessing candidacy and withdrawal status for BUP initiation, obtaining a pregnancy test amongst women of child-bearing potential, administering and/or prescribing BUP when appropriate, and arranging referral for ongoing MOUD. Unless required for referral, liver enzyme and urine toxicology testing were deemed unnecessary. Further, sites needed to establish a feasible means of completing each protocol action and facilitate documentation of each in the EMR. For example, at VRH, a plan for providing BUP prescriptions was necessary as the sole referral center only accepted new patients one day per week. Whereas, at CMC, a plan for receiving ongoing treatment within 24 hours was necessary as expecting locums-staffed physicians to be x-waivered was unrealistic. In the absence of preliminary data at the sites and in the literature as well as uncertainty about if and when sites would introduce protocols, we did not set “Go-Live” dates or incorporate specific values for provider adoption or patient reach in our definition of successful programmatic implementation. Rather, this milestone marked the beginning of each site’s programmatic evaluation period, during which feasibility, acceptability, and effectiveness was described further through process measures, including the primary implementation outcome, and, secondarily, through patient-level outcomes to explore impact and acceptability. Collectively, these data, contextualized by qualitative data generated through the IF formative evaluation, provide a rich description the programmatic implementation from which conclusions about success can be drawn.

The Primary Implementation Outcome, abstracted from the EMR to evaluate implementation, is the proportion of unique patients receiving ED-initiated BUP amongst those determined to be candidates for it. Other implementation outcomes include provider adoption (number of unique practitioners who initiated BUP per EMR abstraction) and clinicians’ readiness for initiating BUP (measured before and after the study using a 10-point Likert scale). Abstracted EMR data were used to assess implementation fidelity to critical actions and to support quality improvement and overall study goals. We adhered to all methodological elements for chart reviews in the guidelines published by Annals,¹³ and have appended our operating procedures, abstraction forms, and details related to the training and monitoring. [Appendix-MRA]. Abstraction was performed by a second independent research coordinator for approximately 15% random days to monitor and improve data quality. Classification as a BUP candidate required EMR documentation of eligibility and willingness to receive ED-initiated BUP in accordance with sites’ clinical protocols or that candidacy could be reasonably inferred by documented clinical actions. Specifically, clinical protocols listed reasons for non-candidacy as recent methadone use, medically instability, allergy, unwilling to receive BUP in the ED; the CMC protocol also included patients who were not in withdrawal during their ED visit who could be seen on the same day/next morning for treatment. Abstraction forms included the pre-specified reasons for non-candidacy as well as free-text-fields..

The main secondary outcome, evaluating program effectiveness, is the proportion of enrolled patient-participants objectively confirmed to be engaged in formal addiction treatment on the 30^th day after their index ED visit. Study staff contacted treatment providers directly confirm dates of attendance, types of treatment, and specifically, whether on Day 30 they continued to be active patients with ongoing medication prescriptions/dispensations (i.e., had not stopped attending). Additional secondary outcomes include patient-reported outcomes used in the original Yale study and subsequent trials in progress;^14,15 these include related to changes in drug use (via the 30-day Timeline followback drug use inventory method),¹⁶ acceptability and treatment satisfaction scales, quality of life and health (EuroQol 5-Dimensions Health State Index Score-3L. [Eq-5D] and Patient Health Questionaire-9 [PHQ-9]),^17,18 risk behaviors (e.g., overdose events, injection drug use), and other process measures assessed at baseline and the 30-day research visit. Urine toxicology testing for opioids, including BUP and fentanyl, was performed.

Analyses

The statistical analytic plan, including precision calculations, can be found in [Appendix-Stats]. Analysis of the primary implementation outcome, (probability of candidates receiving ED-initiated BUP), was descriptive in nature. Bayesian estimates were used for the site-level probability values with 95% CIs due to anticipated sparse data and to account for a new BUP formulation that was anticipated to become available during the course of the study but that wasn’t released. Prevalence and bias-adjusted kappa statistics were generated to assess agreement between independent coordinators that abstracted data related to the (i) candidacy for- and (ii) provision of ED-initiated BUP. Engagement in treatment on Day-30 estimates followed the same Bayesian methods. Change from baseline for binary outcomes was presented as odds ratios with corresponding 95% CIs and p-values based on marginal mixed logistic regressions using PROC GLIMMIX in SAS 9.4 for paired dependent observations. For continuous outcomes, paired t-tests were used. Pre/Post clinician readiness was assessed via Wilcoxon rank sum test with continuity correction for non-paired samples.

As none of the sites had existing ED-initiated BUP programs or reliable data quantitating the target population, collective expert opinion largely informed estimates for precision calculations. We assumed that a total of 120-180 BUP candidates would be identified for the primary outcome measure, amongst whom 40% would receive ED-initiated BUP and 82.5% would consent to study participation – yielding 42-60 participants for the secondary treatment engagement outcome. Our simulations found that it would be unlikely for 95% CIs of the primary outcome proportion to be wider than 0.4. The upper limits for engagement in treatment and other secondary outcomes were derived from outcomes observed under the ideal conditions of the Yale study.³ Participants lost to follow-up were counted as not engaged.¹⁰

Qualitative data derived from the formative evaluation and other IF activities were synthesized using the described Rapid Assessment Process to inform the systems-level intervention in real-time. Formal thematic analyses will be conducted and reported separately with initial manuscripts in-press.¹⁹

Results

Implementation Feasibility:

Within 3 months of study approval and initial site visits by external facilitators, each site successfully developed and introduced protocol-based clinical programs for ED-initiation of BUP and referral. In the following 6-months, the protocol was adopted by 52 of 125 (41.6%) unique physicians/APPs across the 3 sites (BHC 38/99(38.4%), CMC 9/17(52.9%), VRH 5/9(55.6%)) who administered and/or prescribed BUP to 112 unique patients. Each site’s clinical protocol contained the pre-specified critical clinical actions and was customized according to their disparate contexts, in particular the differing availability of X-waivered ED practitioners and timely access to ongoing addiction treatment. Further, implementation required changing existing hospital policy restrictions, adding BUP to formularies, developing patient- and provider-facing resources, and navigating referrals for ongoing treatment.

Populations:

During the 6-month programmatic evaluation period, 66,340 patient-visits (includes repeat encounters) occurred, among whom research coordinators identified 2,522 (3.8%) who screened positive for past/present non-medical opioid use and classified 158 as ED-BUP candidates (134 unique). Of the 112 unique patients who received ED-initiated BUP, 40 were enrolled as consenting patient-participants to evaluate 30-day outcomes (BHC:13, VRH:15 CMC:12). 38 (95%) completed the Day-30 research visit (one lost, one incarcerated); 3 visits were completed remotely without biological specimens collected. (Figure 1 Flow diagram). The majority of patients classified as not being BUP candidates had insufficient documentation to determine candidacy, including a sizeable but unquantifiable number for whom the reason for non-candidacy did not exist as one of the pre-specified options on the abstraction forms. In these cases, free-text comments on abstraction forms included that patients preferred and/or received methadone, were admitted or referred for same-day treatment prior to experiencing opioid withdrawal, were in police custody (prescriptions not possible), or indicated reasons why patients were suspected (but not confirmed) to be engaged in treatment already. The primary reason for non-enrollment of BUP recipients was insufficient contact information in the EMR for recruitment, n=37 (33.0%); n=3 (2.7%) refused participation.

Figure 1: Patient and Patient-Participant Flow Diagram.

¹ Category includes both when candidacy could not be confirmed due to incomplete or ambiguous EMR documentation and when reasons for non-candidacy were other than the listed pre-specified options. Per free-text comments, other reasons included methadone use/preference, admitted/transferred for treatment, prisoner status, and ambiguous documentation about current treatment status.

Baseline characteristics for the 134 unique candidates to receive ED-initiated BUP and for the 112 that received it were abstracted from the EMR (Table 3a). BUP candidates had a mean age of 37.0±10.8 and were predominantly male (76.1%), white (66.4%), Medicaid beneficiaries (62.5%) and discharged (76.9%). Enrolled patient-participants (n=40) were similar, and 12.5% were employed, 27.5% were married/partnered, 25.0% had not completed high school/GED, and 95.0% met DSM-5 criteria for severe OUD. (Table 3b)

Table 3a:

Characteristics of BUP Candidates and Recipients of ED-initiated BUP

	Bellevue Hospital Center		Catholic Medical Center		Valley Regional Healthcare		All Sites
	BUP Candidate N (%)	BUP Received N (%)	BUP Candidate N (%)	BUP Received N (%)	BUP Candidate N (%)	BUP Received N (%)	BUP Candidate N (%)	BUP Received N (%)
N	62	43	40	39	32	30	134	112
Sex: Male	52 (83.9)	35 (81.4)	28 (70.0)	28 (71.8)	22 (68.8)	20 (66.7)	102 (76.1)	83 (74.1)
Age (Mean (SD))	42.8 (11.3)	42.0 (10.9)	31.5 (8.1)	31.4 (8.1)	32.9 (7.0)	32.7 (6.8)	37.0 (10.8)	35.8 (10.2)
Race
Black	22 (35.5)	12 (27.9)	2 (5.0)	2 (5.1)	1 (3.1)	1 (3.3)	25 (18.7)	15 (13.4)
White	22 (35.5)	18 (41.9)	37 (92.5)	36 (92.3)	30 (93.8)	28 (93.3)	89 (66.4)	82 (73.2)
Other	15 (24.2)	12 (27.9)	1 (2.5)	1 (2.6)	0 (0)	0 (0)	16 (11.9)	13 (11.6)
Unknown	3 (4.8)	1 (2.3)	0 (0)	0 (0)	1 (3.1)	1 (3.3)	4 (3.0)	2 (1.8)
Insurance Status
None/Missing	19 (30.6)	11 (25.6)	6 (15.0)	6 (15.4)	11 (34.4)	10 (33.3)	36 (26.9)	27 (24.1)
Medicaid	35 (56.5)	26 (60.5)	29 (72.5)	28 (71.8)	17 (53.1)	16 (53.3)	81 (60.4)	70 (62.5)
Medicare	4 (6.5)	3 (7.0)	4 (10.0)	4 (10.3)	2 (6.3)	2 (6.7)	10 (7.5)	9 (8.0)
Commercial	4 (6.5)	3 (7.0)	1 (2.5)	1 (2.6)	2 (6.3)	2 (6.7)	7 (5.2)	6 (5.4)
Disposition
Discharged	38 (61.2)	32 (74.5)	37 (92.5)	36 (92.3)	28 (87.5)	27 (90.0)	103 (76.9)	95 (84.8)
Admitted	4 (6.5)	4 (9.3)	3 (7.5)	3 (7.7)	4 (12.5)	3 (10.0)	11 (8.2)	10 (8.9)
Inpatient “detox” (SUD treatment)	20 (32.3)	7 (16.3)	0 (0)	0 (0)	0 (0)	0 (0)	20 (14.9)	7 (6.3)

Table 3b:

Characteristics of Patient-Participants

	Bellevue Hospital Center N (%)	Catholic Medical Center N (%)	Valley Regional Healthcare N (%)	All Sites N (%)
N	13	12	15	40
Male	12 (92.3)	9 (75.0)	9 (60.0)	30 (75.0)
Age (Mean (SD))	43.1 (11.3)	31.8 (8.0)	32.4 (6.6)	35.7 (10.0)
Ethnicity
Hispanic or Latino	5 (38.5)	1 (8.3)	0 (0)	6 (15.0)
Race
Black or African American	4 (30.8)	0 (0)	0 (0)	4 (10.0)
White	6 (46.2)	11 (91.7)	15 (100.0)	32 (80.0)
Other (multiracial [1], Asian [1], American Indian or Alaskan Native [1])	3 (23.1)	1 (8.3)	0 (0)	4 (10.0)
Education completed
Less than high school diploma	3 (23.1)	3 (25.0)	4 (26.7)	10 (25.0)
High school graduate or GED	8 (61.6)	6 (50.0)	8 (53.3)	22 (55.0)
Some college, Associate’s degree, or technical degree	2 (15.4)	3 (25.0)	3 (20.0)	8 (20.0)
Bachelor’s degree and/or higher education	0 (0)	0 (0)	0 (0)	0 (0)
Marital status
Married	1 (7.7)	2 (16.7)	0 (0)	3 (7.5)
Divorced or separated	3 (23.1)	2 (16.7))	7 (46.7)	12 (30.0)
Never married	8 (61.5)	6 (50.0)	3 (20.0)	17 (42.5)
Living with partner	1 (7.7)	2 (16.7)	5 (33.3)	8 (20.0)
Employment
Working now	0 (0)	2 (16.7)	3 (20.0)	5 (12.5)
Unemployed or on leave or other	12 (92.3)	7 (58.3)	10 (66.7)	29 (72.5)
Retired	1 (7.7)	0 (0)	0 (0)	1 (2.5)
Disabled permanently or temporarily	0 (0)	3 (25.0)	2 (13.3)	5 (12.5)
Severity of OUD
None	0 (0)	0 (0)	0 (0)	0 (0)
Mild	1 (7.7)	0 (0)	0 (0)	1 (2.5)
Moderate	0 (0)	0 (0)	1 (6.7)	1 (2.5)
Severe	12 (92.3)	12 (100.0)	14 (93.3)	38 (95.0)
Disposition
Discharge	6 (46.2)	11 (91.7)	11 (73.3)	28 (70.0)
Admitted to inpatient	2 (15.4)	1 (8.3)	3 (20.0)	6 (15.0)
Inpatient “detox” SUD treatment	5 (38.5)	0 (0)	0 (0)	5 (12.5)
Left against medical advice	0 (0)	0 (0)	1 (6.7)	1 (2.5)
Insurance Status1
Commercial, Employer or Union	2 (15.4))	1 (8.3)	3 (20.1)	6 (15.0)
Medicare	1 (7.7)	1 (8.3)	1 (6.7)	3 (7.5)
Medicaid	11 (84.6)	6 (50.0)	10 (66.7)	27 (67.5)
No Insurance or self-pay	1 (7.7)	4 (33.3)	5 (33.3)	9 (22.5)

¹None or multiple insurance statuses can be endorsed by participants.

Primary Outcome (Implementation):

112 of 134 unique candidates received ED-initiated BUP. The percentage of candidates who received BUP was estimated using the Bayesian method to be 85.1% (95%CI: 79.7%–90.4%), (Table 4). Differences in subgroups for unique patients were not formally tested; however, in post-hoc analyses, candidates identifying as white were more likely to receive BUP compared to those identifying as non-white by a difference of 22.90% (95%CI 6.86—39.94). Inter-rater agreement for the primary outcome chart abstraction was “substantial”; specifically, prevalence and bias-adjusted kappas were 0.63 (95% CI 0.54–0.71) with 81.4% overall agreement and 0.71 (95%CI: 0.35–1.00) with 85.7% overall agreement for the (i) candidacy for and (ii) provision of ED-initiated BUP, respectively.

Table 4:

Primary Outcome (Implementation) and Main Secondary Outcome (Effectiveness)

	Bellevue Hospital Center N (%)	Catholic Medical Center N (%)	Valley Regional Healthcare N (%)	Total N (%)
Amongst ED Patients
ED patients eligible for and willing to receive ED-initiated BUP (i.e., ED-BUP candidates)	62	40	32	134
ED-BUP candidates who received ED-initiated BUP	43 (69.4)	39 (97.5)	30 (93.8)	112 (83.6)
Primary Outcome: Bayesian estimate of percent and 95% CI of ED-BUP candidates who received ED-initiated BUP				85.1% (79.7% - 90.4%)
Amongst Participants
Enrolled participants who received ED-initiated BUP	13	12	15	40
Enrolled participants engaged in treatment 30 days post-index ED visit	7 (53.8)	8 (66.7)	4 (26.7)	19 (47.5)
Main Secondary Outcome: Bayesian estimate of percent and 95% CI of participants engaged in treatment 30 days post-index ED visit				49.0% (35.6% - 62.5%)

Patient-level Outcomes (Effectiveness):

Amongst the 40 participants who received ED-initiated BUP, 49.0% (95%CI: 35.6%–62.5%) were estimated using Bayesian methods to be actively engaged in formal addition treatment on the 30^th day following the ED index visit (objectively confirmed). Site and overall statistics are shown in Table 4. Twenty-six of the 38 study completers (68.4%) reported attending OUD treatment at least once after their index ED visit. Several outcomes related to opioid use and associated consequences were positive, including a statistically significant decrease in the likelihood of reporting an opioid overdose within 30 days. The odds ratio for reporting an overdose within 30 days of the baseline visit, compared to the follow-up visit, was 4.03 (95%CI: 1.27–12.75). Participants reported a mean of 2.9 (95%CI: 1.95–3.89) fewer days of non-medical opioid use per week (p<0.001) at the follow-up visit. Approximately half of the urine toxicology tests at follow-up (N=35, missing were 3 remote assessments, 2 lost to follow-up) were negative for illicit opioids, and 80% confirmed the presence of BUP.

Participants perceived ED-initiation of BUP favorably; 67.5% were satisfied or very satisfied with how BUP works. Participants mean rating, on a scale of 1-10, of likely are they would be to recommend BUP to others was 7.9 (SD 2.57). There was a marginally significant improvement in health-related quality of life, as measured by the EQ-5D-3L.¹⁷

Across all sites, N=80 and N=83 ED staff, including physicians, APPs, nurses, social workers, and administrators, completed pre- and post-study surveys, respectively. Readiness to provide ED initiated BUP increased from medians of 1.92/10 to 6.95/10 for a median difference of 5.02 (95%CI: 3.56–6.47) with statistically significant improvements seen at each site.

There were no study or treatment-related serious adverse events or deaths. Suicide risk assessments were triggered for 12 participants based on their responses to the PHQ-9 questionnaire (11 at baseline, 5 at follow-up).¹⁸

Limitations

This multicenter implementation feasibility study was designed to balance scientific rigor with the urgent public health need for rapid dissemination of evidence and strategies to effectively implement BUP in EDs that are representative of where the majority of people receive care. While the distinct characteristics of the three Northeastern sites and specific challenges of implementing de novo programs at each represents a sizable proportion of ED settings, further study of other ED contexts is warranted to broaden generalizability. Several limitations reflect our decision to prioritize the primary evaluation of implementation under real-world conditions ahead of our secondary patient-level outcomes. Specifically, our focus on minimizing the influence of research on clinical behaviors during patient encounters is reflected in the decision to delay the approach of research coordinators for recruitment activities until after ED visit completion. This resulted in missed enrollment opportunities and diminished ability to resolve ambiguous or incomplete EMR documentation. A consequence of sites’ changing EMR systems and experiencing clerical staff losses was that patients’ contact information were rarely updated, which often precluded offering enrollment to potential participants who presented outside of recruiting shifts. The risk of selection bias is mitigated by this occurring at a systems-level (i.e., whether or not patients actually had contact information) and by recruitment shifts spanning wide-ranges of times and days of the week. Together, these factors contributed to the small sample of enrolled participants necessary to evaluate effectiveness. As such, patient-level outcomes, including those reaching statistical significance, should be interpreted as preliminary.

There were no missing data necessary for the primary implementation endpoint. Further, the capture of BUP administrations and prescriptions should be reliable and complete given electronic ordering. The risk of abstraction error is mitigated further by the substantial agreement between independent blinded reviewers. However, defining and operationalizing the denominator (i.e., patients for whom ED-initiated BUP is clinically appropriate) remains an ongoing struggle for researchers and policy-makers seeking to specify meaningful quality measures that can be feasibly ascertained using EMR and administrative data for large-scale clinical application. As we learned in this study, clinical documentation often lacks sufficient detail to draw conclusions with certainty. The large number of patients whose reason for not being BUP candidates was not clearly documented likely reflects some missed opportunities and may positively skew the primary implementation outcome, as patients with insufficient clinical documentation were classified as non-candidates. Importantly, however, as none of the sites initiated BUP prior to the study, the raw number of initiations remains an unbiased indicator of program implementation and reach. Further, the change in rate of ED-initiated BUP is the primary outcome for at least two large implementation-effectiveness trials.^14,15 Our study, thus, contributes to developing quality measures for ED-based treatment initiation for OUD that are feasible, meaningful, and attributable to the ED.

Discussion

This study demonstrated that a period of implementation support of ED-initiated BUP was successful in rapidly and effectively implementing de novo, ED-based programs for the initiation of BUP and referral in rural and urban settings with high-need, limited resources, and varied staffing structures. Importantly, it linked implementation to clinical practice change, implementation process measures, and patient-level outcomes. The implementation of ED-initiated BUP required strategies to address considerable barriers, including broad initial reluctance to change practice, absence of existing referral networks, an existing policy preventing the use of BUP, and an overall absence of experience with BUP. Each site successfully developed and implemented a clinical protocol for BUP initiation and integrated it into routine ED-based clinical care. Within 6 months of clinical protocol introduction, 52 (41.6%) unique clinicians initiated BUP and referral for 112 patients – reaching an average of 4.4 patients per week. Given the absence of BUP use in each of these EDs prior to study implementation, and low initial readiness to implement ED-initialed BUP (1.92/10), it is highly unlikely that these patients would have received BUP through any of these EDs in absence of external IF support. Further, approximately half of the enrolled patient-participants were confirmed to be engaged in formal addiction treatment on Day-30, and there were statistically significant decreases in non-medical opioid use and overdose events. These outcomes are particularly remarkable considering that all patient-participants were enrolled after the clinically-driven ED care was completed, in contrast to patient-level intervention studies.

As expected, given the considerable methodological and contextual differences between the Yale efficacy trial and our work-more consistent with real-word implementation, our 30-day treatment engagement rates lagged their 78% engagement.³ Retrospective and observational reports of ED-initiated BUP have been reported, but in the absence of a prospective implementation framework the ability to describe prospective implementation strategies and evaluate secondary outcomes is limited.^20–25 Our study’s rates of linkage to and 30-day engagement in OUD treatment are similar to those found in a recent government-funded demonstration project in which 77% of patients receiving ED-initiated BUP in conjunction with peer-recovery support in five South Carolina EDs were linked to treatment, and 43% remained engaged at Day-30.²⁰ Further, these rates are similar, and often superior, to those of completed referrals from the ED for chronic diseases, like asthma and cardiovascular disease, particularly amongst studies involving populations, like ours, having well-established barriers to aftercare access because of underinsurance, lack of existing care, rural geography, and social determinants.^26–29 The decrease in past 30-day self-report of opioid overdose rates amongst patients receiving ED-initiated BUP and other improvements in patient-reported outcomes are consistent with BUP’s well-established safety profile, effects on reducing opioid-related morbidity and mortality, and wide-ranging psychosocial and economic benefits.^2,30–32

The practice of ED-initiated BUP has increased substantially since the 2015 publication of the Yale trial and endorsements of several professional societies; however, diffusion remains limited, particularly among rural and community EDs.^33–35 Delays in the translation of knowledge to routine clinical practice are widely recognized, with the gap between knowledge and clinical practice averaging 17 years.^36,37 In fact, a 2021 national capabilities assessment conducted of 300 primarily small community and rural EDs participating in the ACEP E-QUAL Opioid Network initiative found that only 6.3% of EDs reported having a protocol for ED-initiated BUP.⁷ Changes in clinician knowledge, attitudes, and behavior are thought to be key factors in the implementation of practice change.³⁸ Our experience through this study reinforces our belief that external facilitation is of critical importance, at least in these early stages of implementation and particularly in the absence of a highly engaged local champion. The application of IF strategies– external facilitation, stakeholder engagement, formative evaluation, learning collaboratives, academic detailing and performance monitoring and feedback– over less than a one-year period translated into practice change and improved patient-level outcomes. This study, and its clinical protocols, educational resources and training materials, helps provide a roadmap for integrating an evidence-based practice, ED-initiated BUP, into various of high-need, low-resource settings. We have made these and other resources freely available: [Appendix-Resources].

By demonstrating the feasibility and acceptability of program development and implementation across these diverse settings and generating preliminary evidence of effectiveness, this study represents an important step towards widespread implementation and dissemination of ED-based treatment initiation for patients with OUD. Future research should optimize the application of IF at scale, particularly within the context of continually increasing opioid-associated mortality. Longitudinal evaluation of downstream impacts of treatment initiation in the ED is warranted also; however, outcomes would be less attributable to the ED, reflecting influences beyond the scope of this study. Additionally, the high prevalence of thoughts of self-harm/suicide emphasizes existing concerns about the precarious mental well-being of this population and should generate hypotheses about how to support its needs. Also, our observation that white patients were 22% more likely to receive BUP than non-white patients is consistent with prior literature on racial disparities and calls for proactive approaches and further study with regards to disparities in access to treatment.^25,39–41Although we did not provide a racial equity intervention, IF methods are highly-suited for this purpose. Our confidence in this assertion is boosted further by our positive experience in using IF to address stigma, which was recognized to be apparent and prevalent.¹⁹

Conclusion

Building upon a randomized trial of initiating BUP in a resource-rich, urban academic ED, our multicenter implementation study provides a critical next step for prospective implementation and evaluation of programs across a variety of clinical settings, including community, rural critical access, and public safety-net EDs. Our IF approach enabled us to learn how to effectively implement a known efficacious treatment across heterogeneous ED settings rapidly, which was associated with promising patient-level outcomes.

Table 5:

Summary of Secondary Patient-Participant Outcomes (N=40)

	Baseline N (%) Mean (SD)	Day 30 N (%) Mean (SD)	Change1 Mean (SD); Mean (95% CI); Odds Ratio (95% CI)
N	40	371
Days of Opioid Use, self-reported for past 7 days
Mean (SD); Mean (95% CI)	4.0 (2.82)	1.2 (2.13)	−2.9 (−3.89 - −1.95)
UDS results:
Opioids (Opiates 2000ng, oxycodone, fentanyl)	—	17 (48.6)
Buprenorphine	—	28 (80.0)
Overdose Events in past 30 days
0	29 (72.5)	35 (92.1)	[ref]
1+	11 (27.5)	3 (7.9)	Odds Ratio (95% CI) 4.03 (1.27, 12.75)
Time to 1st formal addiction treatment
Missing	—	2 (5.0)
Never	—	12 (30.0)
Within 24 hours	—	11 (27.5)
Within 3 days	—	5 (12.5)
Within 8 days	—	6 (15.0)
Longer than 8 days	—	4 (10.0)
Quality of Life: EQ-5D-3L Health State Index Score2
Mean (SD), Mean (95% CI)	0.68 (0.28)	0.76 (0.22)	0.08 (<0.01- 0.16)
Acceptability
On a scale of 1-10, how likely are you to recommend to others that they consider this treatment? Mean (SD)	—	7.9 (2.57)
If you received buprenorphine, how satisfied or dissatisfied are you with how buprenorphine worked or is working?
Very dissatisfied	—	0 (0)
Dissatisfied	—	4 (10.8)
Somewhat satisfied	—	8 (21.6)
Satisfied	—	9 (24.3)
Very satisfied	—	16 (43.2)
Safety
Related Serious Adverse Events or Deaths	—	0 (0)
Suicide Assessment Plan Activated: Endorsed any frequency of having thoughts of being better off dead or of hurting themselves in some way over the last 2 weeks (PHQ-9 Assessment)	11 (27.5)	5 (13.2)

¹Follow-up N=38 with the exception of urine toxicology (N=35) and EQ-5D-3L (N=37). Two participants were lost to follow up (1 incarcerated); 3 participants completed follow up visits remotely without collection of biological specimens; 1 declined to complete EQ-5D-3L.

²The Eq-5D-3L index score is a measure of health state with a score of 0 being equivalent to death, 1 being perfect health, and negative scores being a health state worse than death. It has been validated for use in OUD; clinically meaningful changes are disease specific and not well established for OUD.