A scalable, automated warm handoff from the emergency department to community sites offering continued medication for opioid use disorder: Lessons learned from the EMBED trial stakeholders

Abstract

Background

In order to streamline the emergency department (ED) referral process in a multi-network automated opioid treatment referral program, we performed a needs assessment of community providers for Medication for Opioid Use Disorder (MOUD) in the EMergency department-initiated BuprenorphinE for opioid use Disorder (EMBED) trial network.

Methods

A needs assessment was conducted in two phases: (1) key stakeholder meetings and (2) a survey of community sites offering MOUD. Stakeholder meetings were conducted with five key stakeholder groups: 1) ED clinicians and staff, 2) community sites offering MOUD, 3) the investigative team, 4) health system IT staff, and 5) medical ethics experts. Meetings continued until each stakeholder group stated that their priorities and needs were understood. Major categories of needs were extracted pragmatically based on recurrence across stakeholder groups. Informed by needs expressed by IT and MOUD site stakeholders, nineteen MOUD sites were surveyed to better characterize information needs of community sites offering MOUD when receiving an ED referral.

Results

Three major categories of needs for referral system were identified: 1) The system to be automated, flexible and allow multiple channels of referral, 2) Referral metrics are retrievable in a HIPAA compliant manner, 3) Patients are scheduled into community sites offering MOUD as urgently as possible. Of the MOUD sites surveyed, 68.4% (13/19) responded. Based on the responses, specific patient identifiers were required for most MOUD site referrals, and encrypted emails and EHR were the preferred methods of communication for the handoff. 53.8% (7/13) of the sites were able to accept patients within 3 days with only 1 site requiring more than 7 days.

Conclusion

These findings can inform IT solutions to address the discordant priorities of the ED (rapid and flexible referral process) and the community sites offering (referrals minimize variability and overbooking). To prevent drop-out in the referral cascade, our findings emphasize the need for increased availability and accessibility to MOUD on demand and protected communication channels between EDs and community providers of MOUD.

BACKGROUND

Medication for opioid use disorder (MOUD) is a safe and effective method for treating opioid addiction. (Mattick, Breen, Kimber, & Davoli, 2014) Buprenorphine (BUP), a partial opioid agonist often combined with the antagonist naloxone, and methadone, a full opioid agonist, are MOUDs that are highly effective in decreasing the rates of non-fatal and fatal overdose, withdrawal symptoms, craving, opioid use, risks of infectious-disease transmission, and criminal activity engagement as well as increasing addiction treatment retention. (Kakko, Svanborg, Kreek, & Heilig, 2003; Larochelle et al., 2018; Mattick et al., 2014; Sullivan & Fiellin, 2008; Volkow, Frieden, Hyde, & Cha, 2014) Currently, only 2 in 5 people with OUD are being treated with MOUD. (Volkow et al., 2014) Meanwhile, EDs in the U.S. experienced a 30% increase in patients presenting with opioid overdose from 2016–2017. (Vivolo-Kantor et al., 2018) To address the opioid crisis in the ED, we demonstrated in 2015 that initiation of BUP in the ED followed by referral for ongoing MOUD at an office-based treatment program was nearly twice as effective as referral alone in retaining patients with opioid use disorder (OUD) in formal addiction treatment at 30 days (78% with BUP and referral vs. 37% with referral alone and 45% with referral and brief intervention, p < 0.001). (D’Onofrio et al., 2015)

Increasing need and demand for MOUD with a relatively unchanging supply of community sites offering MOUD treatment have created a bottleneck effect in the MOUD referral process. (Roman, Abraham, & Knudsen, 2011) Between 2003 and 2012 the rate of OUD increased from 634 to 831 individuals per 100,000; however, the number of opioid treatment programs only increased from 1067 to 1167 across the nation (Jones, Campopiano, Baldwin, & McCance-Katz, 2015). In 2015, 38 states reported that 75% of their opioid treatment programs were operating at 80% or higher capacity. (Jones et al., 2015) Moreover, only 44–75.5% of DATA-waived physicians actually prescribe BUP due to barriers such as lack of institutional support, reimbursement, limited access to addiction specialists and low provider confidence in addressing addiction;, and those that do prescribe BUP prescribe well below their patient limit. (Hutchinson et al. 2014; Cunningham et al. 2006; Kissin et al. 2006; Walley et al. 2008; Jones and McCance-Katz 2019) These issues complicate the ED referral process and necessitate appropriate coordination of referral to community sites offering continued MOUD that will have the capacity to treat them.

Given the current landscape of OUD treatment options and to help coordinate ED referral to community sites offering continued MOUD, we aim to improve and streamline the referral process as a part of our upcoming pragmatic trial of user-centered clinical decision support to implement EMergency department-initiated BuprenorphinE for opioid use Disorder (EMBED). This parallel cluster randomized trial of a user-centered health IT intervention aims to increase adoption of ED-initiated BUP and referral for ongoing MOUD into routine emergency care in 20 EDs across five healthcare systems. Since all data for the trial will be collected pragmatically from each site’s EHR without research staff in the ED, the referral component of the intervention must be well-integrated into the emergency clinician’s EHR workflow. The Agency for Healthcare Research and Quality recommends that as a part of the success in the patient referral process, providers should establish referral agreements that delineate mutual expectations and responsibilities (including what information is necessary for and after a referral) and help patients through the referral process. (“Make Referrals Easy: Tool #21 | Agency for Healthcare Research & Quality,” 2015) This Brief Article reports a pragmatic needs assessment of multiple, key stakeholders in order to develop a scalable, automated warm handoff in the referral process from the ED to community sites offering MOUD.

METHODS

A needs assessment was conducted in two phases to create a scalable, automated solution for warm handoffs from trial site EDs to surrounding community sites offering MOUD in preparation for the upcoming EMBED trial: (1) key stakeholder meetings and (2) a survey of community sites offering MOUD. Stakeholders were recruited from the EDs in the EMBED trial network and the lead site’s local providers for continued MOUD.

Stakeholder meetings were conducted with five key stakeholder groups to identify their needs and preferences in the ED to MOUD site referral process: (1) ED clinicians and staff (attending and resident physicians and ED addiction counselors at the Yale New Haven Hospital ED, n = 26), (2) clinicians and staff at community sites offering MOUD (attending physicians and front desk and scheduling staff of two community MOUD sites one in the Yale-New Haven System and the other outside of the system, n = 7), (3) the EMBED investigative team (n = 8), (4) Yale- New Haven Health system IT staff (n = 4), and (5) medical ethics experts (Yale’s IRB, NIH Health Care Systems Research Collaboratory Ethics Core). Stakeholder meeting participants included both staff involved with EMBED sites as well as unaffiliated staff at Yale-New Haven Health and the Yale School of Medicine. Meetings were approximately one hour in duration, and conducted by OMA (medical student / student researcher with business training) and/or ERM (emergency physician / clinical informatics researcher / EMBED PI). Extensive, handwritten notes of issues raised during the meetings were taken by OMA in real time. Due to time constraints regarding trial initiation, a formal qualitative analysis was not feasible. Instead, meetings were conducted with each stakeholder group (multiple times if necessary) until the stakeholders expressed that their priorities and needs were adequately understood. Whenever possible, stakeholder meetings were conducted as part of that group’s regularly scheduled staff meetings. Following each meeting, although a formal qualitative analysis was not performed, the investigative team did have formal debriefing sessions to discuss major themes and issues raised during the meeting. During the debriefing sessions, the team synthesized the top needs that were expressed via an informal consensus process. OMA and ERM met weekly to analyze and further synthesize these needs based on their recurrence across all stakeholder groups.

The investigators identified solutions to each primary need expressed with input from the EMBED IT and design teams at biweekly meetings.

Following the stakeholder meetings, based on needs and priorities expressed by the clinicians and staff at community sites offering MOUD and the Yale-New Haven Health System’s IT leadership (lead EMBED system), it became apparent there were several potential channels of communication between the ED and the MOUD providers. Given this fact and a goal of a scalable solution in the EMBED trial network and beyond, a survey was designed to better characterize and quantify information needs of community sites offering MOUD when receiving an ED referral for ongoing MOUD. The survey (Appendix 1) was designed by OMA, JAM, and ERM based on input from the IT and MOUD stakeholders and consists of 7 questions regarding communication method preferences, scheduling preferences, patient-specific information needs, and types of treatment offered. It was created and distributed to 19 MOUD sites in the EMBED network with Qualtrics Survey Software (Qualtrics International Inc., Provo, UT). Follow-up phone call reminders were conducted to optimize survey completion. Concurrently, we also identified which data elements were necessary for a successful referral process and to monitor handoff quality while remaining in compliance with regulatory statutes 42 CFR part 2 and MOUD site leaders needs. Data were analyzed using descriptive statistics.

RESULTS

Results from the stakeholder meetings are summarized in Table 1. ED physicians and residents were clear on their needs for an automated referral system that minimizes disruption to the ED workflow. The community providers for MOUD also prioritized an efficient process with minimal workflow disruption. The EMBED investigative team wanted a referral system that is scalable and able to collect data on the effectiveness of the referral process. IT staff expressed a need for specific list of information MOUD sites required for a referral. Biostatisticians on the trial’s data coordinating team stated a requirement for a method of communication to collect data points. The medical ethics experts required that patient privacy and protect patient consent.

Table 1:

Needs expressed by various stakeholders in the EMBED project regarding a referral system to MOUD treatment

Stakeholder	Role	Needs	How they were expressed	Solutions
Emergency department clinicians and staff	Attending physicians	Automate referral	“I do not have time to call sites for external referral in a busy ED”	Referral automated and implemented into EMBED CDS
	Resident physicians	Minimize disruption to workflow	“Using the EMBED CDS is already a new workflow for us, so the referral process needs to be integrated and quick”
	ED addiction counselors	Best match of MOUD site to patient’s needs	“Some patients will take one look at [the local OTP] waiting room and never come back”	Include more than one option for referral site selection
Clinicians and staff at community sites offering MOUD	Attending physician	Minimize disruption to workflow	“We can slightly overbook urgent referrals, but patient scheduling should not grossly over-utilize clinic capacity”	Set a limit on how many patients can be overbooked per week by our referral system to ensure a balance between quick referral and manageable workload
	Front desk staff	Efficiency	“Some referrals can be tough to coordinate as patients bounce between outside scheduling services and our office”	Create a standardized flow of how patients will be referred and booked for every case in order to minimize variability in the process
	Scheduling staff	Minimize disruption to workflow	“The easiest way to refer patients to our site is to send them via the same system that we currently use for all referrals”	Work with IT staff to ensure the system has multiple modes of sending out referrals and tailor each MOUD site to its specific preference
Investigative team	Principal Investigator	Scalability	“System should be broadly applicable to any ED and MOUD site”	1. Referral can be sent out via multiple channels, e.g., e-mail, EHR message, or fax. 2. Collect survey data from MOUD sites to determine their preferences.
		Quality Assurance	“We need to have a way to measure the effectiveness of the referral process at each step (i.e. referral, scheduling, appointment visit, medication started)”	Build referral network with the capability to collect aggregate data on % of referrals who were scheduled at MOUD sites, those who attended and those who were started on medication
	Biostatistician	Collect information on referral effectiveness	“We must have a two-way communication system with scheduling service s to collect data on appointments made/attended”	For MOUD sites with EHR linkage, we created an automated data pull that can extract referral usage metrics. For non-EHR linked sites, agree with administrative staff to send us usage data
Health system IT staff	Local EHR programmer s	Specificity of the automation process	“We need a very specific set of commands to automate the referral”	Acquire an exact list of patient information that MOUD sites need in order to make a very specific request to IT to generate an automated referral message
Medical ethics experts	Our institution’s IRB	Ensuring patient privacy	“Any communication with external sites needs to be performed in a HIPAA compliant manner”	1. Worked with IT to encrypt automated email referrals sent to MOUD sites 2. Fax is considered HIPAA compliant
	NIH Ethics Core	Patient consent and collecting data	“In collecting patient data, you must comply with 42 CFR part 2 and have proper consent processes if you collect patient data for research purposes”	Since collecting patient consent for measuring referral efficiency would be too cumbersome, we collected data from MOUD sites as aggregate, deidentified data for QA/QI purposes, which does not require consent.

Results from the MOUD community site survey are summarized in Table 2. Thirteen of 19 (68.4%) surveyed MOUD sites responded. Of the thirteen site that responded, 4/13 (30.77%) were academically affiliated and 9/13 (69.23%) were private MOUD sites compared to 1/6 (16.66%) of the non-responders was academically affiliated and 5/6 (83.33%) were private. 6/13 (46.15%) of the sites that responded were urban (population >50,000) and 7/13 (53.85%) were suburban/rural (population <50,000) compared to 5/6 (83.33%) of the non-responders were urban and 1/6 (16.66%) was suburban/rural.

Table 2:

MOUD Survey Results

Item	Response	N (% of 13 total)
Site Affiliation	Academic	4 (30.77%)
	Private	9 (69.23%)
Site Population	Urban (population >50,000)	6 (46.15%)
	Suburban or Rural (population <50,000)	7 (53.85%)
Location	North Carolina	6/13 (46.15%)
	Connecticut	3/13 (23.08%)
	New Jersey	1/13 (7.69%)
	Minnesota	1/13 (7.69%)
MOUD type	Academically affiliated	4/13 (30.77%)
	Community/Rural	9/13 (69.23%)
Referral method accepted	Fax	8/13 (61.54%)
	Encrypted e-mail	4/13 (30.77%)
	EHR message	4/13 (30.77%)
Scheduling	Can accept patient on an urgent basis (within 3 days)	7/13 (53.85%)
	Require longer than a week to see patients	1/13 (7.69%)
Days of Operation	Monday - Thursday appointments	9/13 (69.23%)
	Friday appointments	6/13 (46.15%)
	Saturday/Sunday appointments	1/13 (7.69%)
Intake hours	Accepts intake before 8:00 am	5/13 (38.46%)
	Accepts intake after 8:00 am	6/13 (46.15%)
	Variable	2/13 (15.38%)
Medication Provided at MOUD	BUP	12/13 (92.31%)
	Methadone	5/13 (38.46%)
Patient information needed for referral	Patient name	13/13 (100%)
	Patient date of birth	12/13 (92.31%)
	Patient contact (phone)	13/13 (100%)
	Patient home address	8/13 (61.54%)
	Patient sex	5/13 (38.46%)
	OUD medication/dosage/prescription given in ED	9/13 (69.23%)
	List of other medications that patient takes	7/13 (53.85%)
	ED provider name	6/13 (46.15%)
	Patient insurance status (medicare/medicaid/private/none)	10/13 (83.33%)
	Opioid use history (date of first use/type of opioid/pattern of usage)	10/13 (83.33%)
	Other substance abuse issues	8/13 (61.54%)
	Medical/psychiatric history	8/13 (61.54%)
	COWS score/DSM Mini-SCID score/Urine Drug Screen/Liver Function Test	6/13 (46.15%)

Respondents preferred fax transmission of referral 8/13 (61.54%). 53.8% (7/13) MOUD sites were able to accept patients on an urgent basis from the ED and see them within 3 days, 12 of 13 (92.3%) sites provided BUP, and 38.5% (5/13) of sites provided methadone. Eight of 13 (61.5%) sites requested at least the following data to be given: patient name, contact number, address, date of birth, insurance status, opioid use history, OUD medication given in the ED, and medical and psychiatric history, including other substance use issues.

DISCUSSION

Stakeholder meetings identified needs and priorities that can be characterized into three domains: (1) The need for an automated, rapid, and flexible referral system that allows referrals to be sent through multiple channels (e.g. email, EPIC, fax); (2) retrieving aggregate referral usage metrics for quality assurance in a HIPAA-compliant manner; and, (3) providing appointments for patients at MOUD sites as urgently as possible without significantly disrupting the workflow of the clinic.

Participating emergency physicians were willing to refer patients with OUD to local MOUD sites on an urgent basis. However, for this practice to be more readily adopted, they require a referral system that is rapid, automated, and flexible. On the other hand, MOUD sites were interested in facilitating urgent ED referrals but had numerous logistical and feasibility issues in adopting the practice. Among the 13 responding MOUD sites, 92% were willing to accept patients in a one-week timeframe via numerous referral methods. Yet most sites required numerous patient-level data elements in order to initiate a referral - something that would be burdensome for ED providers to provide manually. Furthermore, MOUD sites most commonly preferred fax as a means of referral (61.54% of sites). This method is time-consuming, inefficient, and requires multiple steps to compile and transmit the data, which is at odds with the need for efficiency and automation expressed by ED providers. The complexity of this process demonstrated a need for an automated referral system. Our needs assessment interviews and site visits helped us realize the need to use established, confidential, and appropriate referral processes for referrals. Although overbooking MOUD sites is an option, alternative solutions include: (1) bridge clinics for urgent MOUD appointments until care can be established with a long-term provider of MOUD (Martin, Mitchell, Wakeman, White, & Raja, 2018) and (2) streamlined ED referral systems to primary care providers who would be willing to provide MOUD services. (Wakeman & Barnett, 2018)

These findings can inform IT solutions to address the discordant priorities of the ED (rapid and flexible referral process) and the MOUD sites (referrals minimize variability and overbooking). There are technical limitations to a one-size-fits-all solution for all stakeholders. So, for the EMBED pilot study, we are exploring automated EHR messaging from the ED to the local community MOUD sites, automated notification of addiction counselors and follow-up nurses to help facilitate the referral process during and after the ED visit, and automating printing of high quality MOUD site information in ED patient discharge instructions.

In our previous 2015 study, we tracked how many individuals continued to follow up with outpatient MOUD treatment after being started on BUP at the ED. In this trial, follow up was arranged with a primary care office and specific times were arranged within 72 hours of the index ED visit. In routine emergency care practices, resources available to each ED and community providers and programs offering MOUD site to ensure follow-up varies, and this difference has yet to be analyzed. Additionally, there are multiple points in the referral process in which contact between patients and the healthcare system may be derailed. This includes patients either not receiving a referral, obtaining a referral without an actual scheduled appointment, or patients not attending their appointment. The referral process for OUD patients has not been well documented in the literature and adoptions has lagged behind compared to that of other chronic diseases. The “cascade of care” framework, initially introduced in the treatment of HIV/AIDS, can be exemplified as a tool to identify major gaps in OUD treatment.(Kerkerian et al., 2018) Under this framework, standardized metrics could be used to track the OUD treatment progress over time, such as when patients drop out of treatment as well as days between referral and scheduled appointment. Identification of shortfalls along the cascade of care can help identify future areas of improvement in the MOUD referral process. (Eugenia Socías, Volkow, & Wood, 2016; “To Battle the Opioid Overdose Epidemic, Deploy the ‘Cascade of Care’ Model,” 2017) Streamlining the referral process and collecting performance metrics for quality improvement could increase the likelihood of successful warm handoffs from the ED to community MOUD sites in real-world settings.

We believe that a scalable, automated warm handoff from the ED to community sites offering continued MOUD can only be successful if the solution adequately meets the priorities and needs of all stakeholders in the process. Our study’s strengths derive from the engagement of an array of stakeholders involved throughout the process of the ED visit to the community MOUD site referral and learning how to align and meet their needs and priorities. These needs were further characterized quantitatively for the EMBED trial. The generalizability of the quantitative results may be limited due to this selection bias. This may overestimate the availability and willingness to receive ED referrals of MOUD sites in other communities. Regardless, our findings help to understand the requirements for a rapid, automated, and flexible electronic referral in areas where MOUD community sites are available. Furthermore, our initial IT solutions will utilize a diverse, multimodal approach using existing IT infrastructure that can be adopted in most EDs.

We plan on using these results to inform the referral process in the EMBED trial which will streamline ED identification, treatment, and referral for OUD patients. The needs assessment guide the development of a multi-modal ED referral process through direct messaging in the EHR to in-network sites and external messaging via encrypted email to out-of-network sites that use a different EHR. For in-network sites, we plan to collect aggregate metrics to determine the cascade of care effect and see where patients are lost to referral. Ultimately, this assist in achieving regional partnerships between EDs and community sites offering MOUD to assure the availability of continued MOUD, and help similar efforts be rapidly scalable through mirroring our methods and experience.

Highlights.

• ED referral for ongoing MOUD needs an automated, rapid, flexible referral system.

• MOUD sites are interested and willing to accommodate ED referrals urgently.

• Most MOUD sites require patient identifiers for referral.

• Encrypted emails and EHR communication are the preferred methods of referral.

Appendix 1: MOUD Site Survey

1. How would you be willing to accept OUD referrals from the Emergency Department? (Select all that apply)

   • Encrypted E-mail
   • Fax
   • Direct EHR Message (e.g. referral within same EHR as the Emergency Department; please elaborate below)
   • Other (please explain)

2. What software/platform do you use for scheduling? (e.g. EPIC, Microsoft Outlook)

3. What patient information do you require in your referral process? (Select all that apply)

   • Patient Name
   • Patient Date of Birth
   • Patient Contact (Phone)
   • Patient Home Address
   • Patient Sex
   • OUD Medication/Dosage/Prescription given in ED
   • List of other medications that patient takes
   • Emergency department provider name
   • Patient insurance status (Medicare/Medicaid/Private/None)
   • Opioid use history (Date of first use/Type of opioid/pattern of usage)
   • Other substance abuse issues
   • Medical/Psychiatric history
   • COWS score/DSM Mini-SCID score/Urine Drug Screen/Liver Function Test
   • Other (Please specify below)
4. How early are you able to schedule your patients following a referral?

   • Within 1 Day
   • Within 2 Days
   • Within 3 Days
   • Within 1 Week
   • Longer than 1 Week
5. What days are you open to accept a referral?

   • Monday
   • Tuesday
   • Wednesday
   • Thursday
   • Friday
   • Saturday
   • Sunday

6. What times are you open to accept referrals? (e.g. 9AM - 6PM)

7. Medication provided at center (select all that apply)

   • Buprenorphine formulations (e.g. Suboxone)
   • Methadone
   • Naltrexone
   • No Medication/Abstinence