Peer Support Workers to Improve Emergency Department Care Completion Among Patients Who Use Substances: A Pilot Randomized Controlled Trial

Abstract

Objectives

Assess feasibility of a randomized trial and the hypothesis that emergency department (ED) Peer support workers (Peers), who share personal lived experiences and resources, may decrease ED patients with substance use who leave before treatment completion (LBTC). Specific objectives included assessing operational feasibility, estimating recruitment rates, and seeking preliminary evidence of benefit to support the definitive trial.

Methods

Pragmatic, open-label pilot randomized trial over the initial 15 weeks of an ED Peer program. Shifts randomized 2:1 to Peer support or usual care.

We analyzed patients aged 18-75 years with the 3 most common substance use-related presenting complaints: (1) localized swelling/redness, (2) withdrawal, and (3) substance misuse/intoxication.

Feasibility outcomes were patient acceptance and recruitment rate. The primary measure was LBTC (shift level and patient level). Exploratory analyses evaluated naloxone kits and 48-hour ED revisits.

Results

There were 72 intervention shifts (391 visits) and 37 control shifts (198 visits). Patients had a median age of 44 (IQR 35-54); 71.3% were male.

Feasibility

A total of 94.9% of patients accepted Peer support. The mean recruitment rate was 30 patient visits per week.

LBTC

Mean LBTC was 17.0% (SD 18.3%) among Peer shifts, and 23.1% (SD 22.8%) among control shifts, a difference of −6.1% (95% CI −14.6%, 2.4%).

The Peer modestly impacted patient-level LBTC (OR 0.78 [95% CI 0.46, 1.32]). Among patients with substance misuse/intoxication or withdrawal, the Peer decreased odds of LBTC (21.4% vs 26.9%; OR 0.39 [95% CI 0.18, 0.84]).

Exploratory

Peer interaction increased odds of receiving a naloxone kit (OR 2.33 [95% CI 1.34, 4.04]) and decreased 48-hour ED return visits (OR 0.46 [95% CI 0.22, 0.95]).

Conclusion

An ED Peer may decrease LBTC among patients who use substances. A randomized trial is feasible.

The Bottom Line.

Peer support workers with lived experience of substance use could improve engagement among emergency department patients with substance use. This pilot randomized study comparing Peers to usual care found a trend toward decreased shift-level rates of leaving before treatment completion (LBTC) (−6.1% [95% CI −14.6%, 2.4%]) among substance use-related visits and significantly reduced LBTC among patients with misuse and withdrawal (odds ratio [OR] 0.39 [95% CI 0.18, 0.84]). Peers increased naloxone kits (OR 2.33 [95% CI 1.34, 4.04]) and decreased 48-hour revisits (OR 0.46 [95% CI 0.22, 0.95]). Findings demonstrate promise that Peers may decrease LBTC and feasibility of a definitive trial.

1. Introduction

1.1. Background

The burden of substance use in North America is disproportionately high: drug use-related deaths cause 200 deaths per day in Canada,¹ and in the United States, they account for 51.64% of the global total.² Emergency departments (EDs), often accessible 24/7, are critical locations to engage patients who may not be connected to other health or community services. People who use substances visit EDs 7 times more frequently than other patients.³ However, a large proportion of these patients leave before treatment completion (LBTC): either without being seen (LWBS; eg, after registration, prior to bedside nurse and/or physician assessment), or against medical advice (LAMA; eg, after registration and initial assessment, whether or not treatment was initiated). Alarmingly, studies have reported that 25% to 43.4% of people who use substances LBTC from hospital stays,^4,5 exceeding rates in the general ED population,⁵ and LBTC rates are rising.⁶ LBTC from both hospitalizations and ED visits has been strongly associated with overdose, ED revisits, hospital readmissions, and mortality.7, 8, 9, 10, 11, 12, 13, 14 For instance, 21% of patients who LBTC are hospitalized within 15 days, and alcohol dependence increases risk.¹⁵ ED patients presenting with drug toxicity who LBTC have a 7.1-fold increased risk of death within 12 months.¹⁶ Reasons that patients with substance use leave prematurely include undertreated withdrawal, inadequate pain management, long waits, stigma, discrimination, perceptions of neglect or judgement, prior negative experiences, uncompassionate care, patient social and other obligations, financial concerns like fear of loss of employment or income, and inflexible hospital policies (such as inability to leave, restrictive visitor, and smoking policies); negative experiences may further decrease future healthcare engagement (see Table 1).^4,17, 18, 19, 20, 21 Additionally, LBTC outcomes are influenced by multiple factors, including triage-to-bed and door-to-triage times (wait times), crowding, boarding times, volume of waiting and boarded patients, and patient acuity levels.22, 23, 24 Although underlying contributors to LBTC are multifactorial, implementing ED interventions that could improve experiences of people who use substances in EDs and thereby specifically target LBTC rates in this population has the potential to improve health outcomes of people who use substances and is an urgent priority.

Table 1.

Reasons for which patients with substance use leave the ED prior to completing care and potential role of Peers to address.

Reasons for leaving prior to completion of ED care	How ED Peers could address
Undertreated withdrawal symptoms and cravings18	Peers improve early notification of care providers to initiate acute withdrawal management and consideration of substance use dependency treatments (eg, opioid agonist therapies)
Inadequate pain management18	Peers improve early notification of care providers to initiate pain management
Stigma, discrimination, perceptions of neglect, and feeling judged by healthcare providers18,19	Peers interact directly with patients in nonstigmatizing, nondiscriminatory, and nonjudgmental manner; improve communication such that patients better understand their ED experience including expected ED course; assist with education of healthcare providers to decrease stigma and discrimination (eg, principles of trauma-informed care and nonstigmatizing language)
Long ED wait times17,19	Peers improve communication such that patients better understand their ED experience including expected ED course; improve early notification of care providers to decrease wait to receive withdrawal and pain management
Uncompassionate care17,18	Peers build rapport with patients through shared experience fostering empathetic and compassionate interactions
Prior negative experiences17	Peers improve communication and build rapport with patients through shared experience improving positive associations with their ED visit, thereby improving present and future perceptions of care
Social and other obligations21	Peers improve communication such that patients better understand their ED experience including expected ED course; liaise with multidisciplinary healthcare providers to improve connections and transitions to community support services (eg, social work, housing, financial supports)
Financial concerns like fear of loss of employment or income4	Peers improve communication such that patients better understand their ED experience including expected ED course; liaise with multidisciplinary healthcare providers to improve connections and transitions to community support services (eg, social work, housing, financial supports)
Inflexible hospital policies (such as inability to leave, restrictive visitor and smoking policies)20,21	Peers improve communication such that patients better understand their ED experience including expected ED course; communicate patient needs with healthcare providers (eg, need to step outside); accompany patients outside (eg, to smoke)

1.2. Importance

Integrating Peer support workers (Peers) with lived experience of substance use into EDs is a strategy to improve engagement of people who use substances in care and thereby decrease rates of LBTC.²⁵ Peers connect with patients through shared experience and understanding, offering validating, safe, and nonjudgmental interactions that respect patients’ autonomy.26, 27, 28 Peer roles include substance use screening, providing naloxone, harm reduction education, sitting with patients, advocating for timely symptom management (eg, pain or withdrawal), supporting the healthcare team with de-escalation, and facilitating connections to community supports.^26,29, 30, 31, 32, 33 In outpatient, community, and recovery-oriented settings, Peers have been shown to improve treatment initiation, follow-up, and retention in care34, 35, 36; reduce ongoing substance use and/or relapse35, 36, 37; decrease high-risk behaviors for HIV and hepatitis C transmission³⁶; improve relationships with treatment providers³⁵; foster stronger social supports and sense of community^35,37; reduce return to homelessness³⁶; and improve self-determination, self-efficacy, and empowerment.^36,37 Although promising, the effectiveness of Peers specifically within EDs remains unclear. For example, studies show that Peers may be associated with increased naloxone distribution,³¹ but their effect on addiction medication initiation, treatment engagement,^31,38, 39, 40 overdose prevention, ED visits, and hospitalizations^38,40,41 are variable, and studies have found no association with decreased mortality.^38,39,41,42 Few high-quality studies exist: most are underpowered, lack control groups, and focus on outcomes outside of EDs’ sphere of influence (eg, availability of follow-up).⁴³ A randomized trial is critically needed to evaluate ED Peers’ effectiveness in impacting LBTC, an important metric that directly reflects ED care and engagement, while controlling for real-world confounders, such as temporal variations in substance use, concurrent ED interventions (eg, harm reduction and addictions supports), and community services. This pilot study assesses the feasibility of a randomized trial and generates early data to address the evidence gap on ED Peer effectiveness.

1.3. Goals of This Investigation

We aimed to assess the feasibility of a randomized controlled trial evaluating effectiveness of an ED Peer program. Specific objectives were to (1) assess operational feasibility, (2) estimate recruitment rates, and (3) seek preliminary evidence of benefit to support the definitive trial. We have reported implementation and operational feasibility in greater detail in a submitted manuscript (unpublished at time of the current paper).⁴⁴ In contrast to the current study, our previous manuscript provides descriptive summaries of patient and visit characteristics among those supported by the Peer, locations and methods of patient identification and interactions, and referrals and resources provided. For the current study, we hypothesize that a randomized trial is feasible and that an ED Peer program will improve patient engagement, resulting in decreased LBTC.

2. Methods

2.1. Patient and Public Involvement

We partnered with people with lived/living experience of substance use, Peers, healthcare providers, and policy-makers in all aspects of study design and implementation. Before study rollout, we conducted focus groups with people with lived/living experience of substance use and multidisciplinary healthcare providers (physicians, nurses, pharmacists, addiction providers, security personnel, social workers) at Vancouver General Hospital. This engagement guided our Peer program and study design, including our choice of primary outcome (LBTC) and methodological approach of shift-based randomization (patient partners felt that patient-level randomization was unethical).

2.2. Study Design

This was a pragmatic, open-label pilot randomized controlled trial. Peer shifts were the unit of randomization (2:1 allocation ratio). Our study was approved by UBC Research Ethics (H20-03698). This study adhered to the CONSORT guidelines for reporting of randomized controlled trials and randomized pilot and feasibility trials.^45,46 There were no protocol changes after study initiation.

2.3. Setting

Vancouver General Hospital ED is an urban, tertiary center with over 90,000 annual visits in Vancouver, Canada. Our study period extended for the first 15 weeks of ED Peer program implementation (November 23, 2024, to March 11, 2025). We chose this study period based on available funding and our team’s availability to conduct the feasibility analyses.

2.4. Recruitment

We recruited patients from November 23, 2024, to March 11, 2025. We measured LBTC during ED visits of interest; therefore, we did not require follow-up beyond the ED visit.

2.5. Peer Identification, Hiring, and Training

Operational leadership and professional practice at our site led a standardized ED Peer identification, vetting, and hiring process, including developing a job description with required qualifications that was posted internally and externally and shared among partner networks. Required Peer qualifications included certification in a social service program; successful completion of a recognized Peer support training program; at least 2 years of recent Peer support experience or an equivalent combination of education, training, and experience; lived experience with recovery from addiction; and experience as a recipient of mental health or addiction services. The recruitment department screened applications and invited suitable candidates to interview with a panel that included a dedicated nurse supervisor, operation leader, professional practice, and a Peer representative. Interviewers used a standardized evaluation tool in accordance with union guidelines, and the recruitment panel verified references for all candidates. The Peer was hired as a paid employee. Our Peer selection and hiring compared with selection processes and requirements described in outpatient settings, including the need for lived experience of substance use, personal recovery, ability to use lived experience to support others, completion of Peer support training and certification, and training on maintaining ethical standards and boundaries.^34,35,47

The Peer’s orientation and training were multifaceted and comprehensive, including in-person orientation within the ED and specific provider groups (eg, social work), shadowing with hospital addictions teams and external partner Peer support programs, and required online training modules on Indigenous cultural safety, harm reduction services, resisting stigma on substance use, advanced overdose response, distributing naloxone to clients, and code blue response. The Peer had regular scheduled 1:1 check-in meetings with clinical nurse supervisors overseeing the program to discuss transition to the role and to provide support with self-identified learning needs.

2.6. Peer Intervention and Comparator

Intervention shifts had the Peer available to support ED patients who use substances. The Peer identified eligible candidates by electronic medical record screening using a standardized approach. Specifically, we trained the Peer to assess triage notes of all patients with a prespecified list of presenting complaints that we identified a priori as being high-yield for substance use-related concerns based on our experience. If there was an indication of substance use on the patients’ initial triage assessment, the Peer would approach them to inquire whether they would benefit from supports. The Peer also received referrals directly from providers if they noted on their assessments that patients had substance use-related issues or identified patients themselves based on signs of substance use while rounding the ED. These flexible identification methods allowed the Peer to identify and approach patients early, frequently in the waiting room and often before physician assessment. Allowing Peer flexibility to identify patients using multiple approaches has been recognized as a best practice to facilitate identification of patients who have a range of presenting complaints.²⁵

Patients provided verbal consent to receive the Peer’s services, similar to other clinical interventions. The duration and frequency of each Peer interaction was not prescribed but rather tailored to individual patients’ needs. Interactions could be brief or prolonged and could consist of a single interaction or multiple check-ins over the course of patients’ hours-long ED visits. The Peer shared personal lived experiences and provided individually tailored supports, including referrals to clinical and social services (eg, social work, addictions specialist, Indigenous patient navigator, community-based addictions resources, recovery groups, detoxification centers), resources (eg, blankets, food, naloxone kits, sterile drug use supplies), emotional support, assistance with de-escalation of agitated patients in collaboration with ED teams, and facilitated communication between patients and healthcare providers (eg, reporting withdrawal symptoms). Overall, the Peer played multiple roles within the study: patient identification, assessment, and support (as described above) as well as feedback and validation of study processes in order to refine the intervention.

Control shifts did not have the Peer available. During control shifts, patients had access to standard care: they could receive referrals and resources provided by the treating medical team, including social work referral, addiction specialist consultation, and harm reduction and other resources.

2.7. Eligibility for Analyses

We included ED patients aged 18 to 75 years with prespecified substance use-related presenting complaints in our analyses. To identify our included list of presenting complaints, we first descriptively summarized all visits supported by the Peer during our pilot study period, which allowed us to identify the 3 most common complaints made by these patients. We then focused our analysis on these 3 presenting complaints, which we identified through electronic medical records: (1) localized swelling/redness (denoting common infectious consequences of substance use), (2) withdrawal (indicating dependence), and (3) substance misuse/intoxication (denoting harmful use). These presenting complaints captured high-risk patients who, based on our experience, we presumed to be at risk for severe consequences of substance use. These patients comprised a group for whom we expected the Peer intervention would have the greatest impact. The intervention group in our analyses comprised all patients who presented with our 3 prespecified presenting complaints. To ensure comparability of the intervention and control arm, we did not include patients with other presenting complaints who were identified by provider referrals or by in-person Peer identification during intervention shifts. Eligibility criteria for our control group were identical: we included patients who presented with the same presenting complaints on control shifts (no Peer).

2.8. Data Collection

The Peer recorded medical record numbers, date, time, and details of patient interactions in a standardized log. We evaluated ED electronic medical records for all visits made during our study period. We used the randomized Peer schedule, which we created prior to hiring and which dictated our Peer work schedule, to identify intervention and control shifts by ascertaining days and times when the Peer was available in the ED. We used medical record numbers, date, and time (recorded by the Peer) to identify patients who interacted with the Peer.

2.9. Outcomes

We determined our feasibility outcomes through discussion among our study team and consultation with our partners with lived/living experience of substance use. Our feasibility outcomes were (1a) adherence to a randomized shift schedule as an indicator that the schedule was acceptable to the Peer, (1b) patient acceptance of Peer services, and (2) recruitment rate over our study period. We did not prespecify thresholds for success for our feasibility outcomes. To evaluate preliminary evidence of benefit (specific objective #3), hypothesizing that the Peer improved engagement with ED care, we assessed the percentage of visits resulting in LBTC during intervention and control shifts. We defined LBTC as a combination of patients who LWBS (eg, after registration, prior to bedside nurse and/or physician assessment) and who LAMA (eg, after registration and initial assessment, whether or not treatment was initiated). We hypothesized that the Peer could influence both outcomes because they are often able to interact with patients early in their course, such as from the waiting room and before clinician assessment. Existing literature indicates that both LWBS and LAMA are associated with poor outcomes.^13,48 Furthermore, we assessed exploratory outcomes that we hypothesized could be impacted if the Peer improved patient engagement in their care. These included dispensation of naloxone kits during patients’ ED visits and ED return visits within 48 hours. We note that not all patients supported by the Peer were eligible for a naloxone kit because eligibility for Peer support was based on use of any substance, not just opioids.

2.10. Harms

The Peer intervention posed minimal risks to participants. Patients were able to decline Peer services if not desired. We monitored for unintentional harms through regular check-ins with research team members, the Peer, and operation leadership.

2.11. Sample Size

We conducted the present analyses in a convenience sample of ED visits made during the first 15 weeks of our program. Our study period was determined by funding and resource availability for our feasibility analysis. We did not conduct a formal power calculation for the feasibility metrics.

2.12. Randomization

2.12.1. Sequence generation

We worked with human resources and ED operational leadership to develop a schedule of intervention (Peer) and control (no Peer) shifts prior to study initiation. We created 6 separate employment lines of Peer work. Each line was full-time or part-time, covering daytime (07:00-15:00), evening (15:00-23:00), and overnight (23:00-07:00) hours. We ensured that our schedule adhered to institutional employee policies in consideration of Peers personal needs. Stipulations included that Peers could not work more than 5 consecutive shifts and required at least 2 weekends and 2 statutory holidays off per 8-week period. For the full-time lines, we generated our shift sequence using block-based randomization that allocated 38 shifts across 8 weeks. Each block combined 1-5 consecutive workdays, followed by 2 days off. We then randomly assigned 2 statutory holidays into each 8-week period. The 8-week sequences were repeated. During our pilot study period, due to natural hiring lags, we had hired a single Peer who only filled the full-time daytime line (07:00-15:00).

2.12.2. Allocation and concealment

Our Peer was hired after we created the randomized schedule and received it upon hiring. Therefore, randomization was implemented at study onset. It was not pragmatic or feasible to conceal shift allocation: the Peer and healthcare providers were aware of the Peer shifts in advance to facilitate planning and referrals in support of patient care.

2.13. Blinding

Due to the nature of the intervention, it was not possible (or desirable) to blind patients or healthcare providers to whether they interacted with the Peer. Due to limited analytic resources, outcome assessors were not blinded to study group allocation.

2.14. Statistical Methods

2.14.1. Primary analysis

We used Welch’s t test to compare the difference in the average shift-level percentage of visits resulting in LBTC (primary outcome) between intervention (Peer) and control (no Peer) shifts, among visits for our prespecified presenting complaints. By comparing visits with the same presenting complaints during intervention and control shifts, we ensured that we were comparing identical patient cohorts in both groups. To mitigate selection bias, we included all visits made during Peer shifts in the analysis irrespective of whether patients interacted with the Peer, ie, an intention-to-treat approach.

2.14.2. Secondary analysis

We recognized that our primary shift-level analysis included many patients who were not targets for the intervention. This limitation was unavoidable as screening was not conducted during control shifts, precluding identification of eligible patients. While mitigating selection bias, our inclusion of nontarget patients in our primary analysis likely introduced noise and diluted the observed effect. Hence, our secondary analysis was a patient-level analysis of the primary outcome. We used a mixed-effect logistic regression to assess the probability of LBTC among patients with presenting complaints of interest who interacted with the Peer compared with those who did not during a non-Peer shift, ie, treatment effect in the treated analysis. Peer interaction logs identified exposed patients, and the control group comprised patients with the same presenting complaints during control shifts. The model adjusted for age, sex, presenting complaint, day of the week, triage acuity, and clustering by individual patients and specific shift dates.

We conducted primary and secondary analyses for all presenting complaints of interest combined and also disaggregated analyses by specific complaint. Due to the nature of our visit-based primary outcome (LBTC), which could be assessed in the electronic medical record for each ED visit in our intervention and control groups, there were no losses to follow-up during our study period.

2.14.3. Sensitivity analysis

We conducted a sensitivity analysis that separately evaluated patients who LWBS (eg, prior to clinician assessment) and LAMA (eg, after initial assessment), recognizing that these may be distinct patient outcomes. We conducted both primary and secondary analyses described above on these disaggregated outcomes.

2.14.4. Exploratory analyses

Among ED visits made during Peer (intervention) compared with no Peer (control) shifts, we conducted exploratory analyses comparing proportions of visits in which (1) a take-home naloxone kit was ordered and (2) a return visit to the ED for any reason occurred within 48 hours. For these analyses, we used a mixed-effect logistic regression with similar methodology and covariates as the treatment effect in the treated analysis described above.

3. Results

3.1. Participant Characteristics

There were 72 intervention (Peer) shifts and 37 control (no Peer) shifts (Fig 1). During intervention shifts, the Peer interacted with 335 individual patients (median age 44 [IQR 35-54], 71.3% male) over 451 visits. The 3 most common presenting complaints were localized swelling/redness, substance misuse/intoxication, and withdrawal (Table 2). All 72 intervention shifts included visits for localized swelling/redness, and 55 included visits for substance misuse/intoxication or withdrawal. Among 37 control shifts, 35 included visits for localized swelling/redness and 29 included visits for substance misuse/intoxication or withdrawal. During intervention (Peer) shifts, we included 276 visits for localized swelling/redness (n = 49 patients interacted with a Peer), and 115 visits for substance misuse/intoxication or withdrawal (n = 54 patients interacted with a Peer). During 37 control (no Peer) shifts, we included 129 visits for localized swelling/redness and 67 for substance misuse/intoxication or withdrawal. Two patients who presented during a control shift remained in the ED for over 24 hours and subsequently interacted with a Peer. In our ED, patients may occasionally have such prolonged length of stays while awaiting support services to arrange outpatient care, during intentional observation periods, or while being assessed or dispositioned by specialist and psychiatric services. We included these patients in the control group (no Peer) for the primary analysis, and in the group who interacted with a Peer in our patient-level, secondary analysis. We performed sensitivity analyses (1) excluding them from our primary and secondary analyses and (2) reclassifying these visits to the intervention (Peer) group in the primary analysis.

Figure 1.

Participant flow diagram.

Table 2.

Baseline demographic and clinical characteristics for intervention (Peer) and control (no Peer) groups.

	Overall (N = 589)	Peer shift (N = 391)		No Peer shift (N = 198)
		Local swelling/redness (n = 276)	Substance misuse or withdrawal (n = 115)	Local swelling/redness (n = 129)	Substance misuse or withdrawal (n = 69)
Age, median (IQR)	46.0 (35.0-59.0)	50.0 (39.0-64.0)	39.5 (32.0-49.8)	50.0 (35.0-64.0)	36.0 (23.0-47.0)
Sex, n (%)
Male	391 (66.4%)	194 (70.3%)	79 (68.7%)	80 (62.0%)	38 (55.1%)
Female	198 (33.6%)	82 (29.7%)	36 (31.3%)	49 (38.0%)	31 (44.9%)
Triage acuity, n (%)
1 - Resuscitation	4 (0.7%)	0 (0.0%)	2 (1.7%)	0 (0.0%)	2 (2.9%)
2 - Emergency	50 (8.5%)	6 (2.2%)	24 (20.9%)	8 (6.2%)	12 (17.4%)
3 - Urgent	381 (64.7%)	184 (66.7%)	78 (67.8%)	75 (58.1%)	44 (63.8%)
4 - Semi-Urgent	153 (26.0%)	86 (31.2%)	11 (9.6%)	45 (34.9%)	11 (15.9%)
5 - Non-urgent	1 (0.2%)	0 (0.0%)	0 (0.0%)	1 (0.8%)	0 (0.0%)
Days of the week, n (%)
Monday	93 (15.8%)	59 (21.4%)	22 (19.1%)	11 (8.5%)	1 (1.4%)
Tuesday	75 (12.7%)	55 (19.9%)	14 (12.2%)	6 (4.7%)	0 (0.0%)
Wednesday	79 (13.4%)	53 (19.2%)	20 (17.4%)	4 (3.1%)	2 (2.9%)
Thursday	76 (12.9%)	34 (12.3%)	19 (16.5%)	15 (11.6%)	8 (11.6%)
Friday	77 (13.1%)	10 (3.6%)	5 (4.3%)	44 (34.1%)	18 (26.1%)
Saturday	104 (17.7%)	23 (8.3%)	16 (13.9%)	35 (27.1%)	30 (43.5%)
Sunday	85 (14.4%)	42 (15.2%)	19 (16.5%)	14 (10.9%)	10 (14.5%)
Dispositions, n (%)
Discharged	391 (66.4%)	194 (70.3%)	68 (59.1%)	84 (65.1%)	45 (65.2%)
Left before treatment completion	110 (18.7%)	41 (14.9%)	28 (24.3%)	23 (17.8%)	18 (26.1%)
Admitted	72 (12.2%)	35 (12.7%)	16 (13.9%)	19 (14.7%)	2 (2.9%)
Transferred	11 (1.9%)	2 (0.7%)	2 (1.7%)	3 (2.3%)	4 (5.8%)
Death	1 (0.2%)	0 (0.0%)	1 (0.9%)	0 (0.0%)	0 (0.0%)
NA	4 (0.7%)	4 (1.4%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Received Peer interaction, n (%)
Yes	105 (17.8%)	49 (17.8%)	54 (47.0%)	0 (0.0%)	2 (2.9%)
No	484 (82.2%)	227 (82.2%)	61 (53.0%)	129 (100.0%)	67 (97.1%)

3.2. Feasibility Outcomes

The Peer worked the full number of scheduled hours during all shifts allocated to the intervention on the randomized schedule, demonstrating adherence and fidelity. Among all patients whom the Peer approached, 94.9% accepted their services. Over the study period, the Peer provided support during 30 visits per week, indicating an acceptable recruitment rate based on our team’s prior estimation that a Peer could be expected to support an average of 6 patients per shift and would work an average of 4.75 shifts per week during a full-time line.

3.3. LBTC Outcomes

3.3.1. Primary shift-level analysis

In our primary analyses, LBTC percentages were not statistically significantly different between Peer and control shifts. Among 72 Peer shifts, the mean LBTC percentage was 17.0% (SD 18.3%). Among 37 control (no Peer) shifts, the mean LBTC percentage was 23.1% (SD 22.8%). The difference in the mean shift-level LBTC percentage between Peer (intervention) and no Peer (control) shifts was −6.1% (95% CI −14.6%, 2.4%) (Table 3).

Table 3.

Percentages left before treatment completion (LBTC) at shift level (primary) and patient level (secondary)

Shift-level LBTC
	Total number of shifts	LBTC mean (SD)
Peer shifts	72	16.95% (18.34%)
Local swelling/redness	72	14.51% (21.9%)
Substance misuse or withdrawal	55	22.12% (33.46%)
No Peer shifts	37	23.07% (22.84%)
Local swelling/redness	35	20.54% (30.80%)
Substance misuse or withdrawal	29	32.74% (37.95%)

Patient-level LBTC
	Number of patients who LBTC	Total number of patients	LBTC percentage (95% CI)a
Peer shifts	25	105	23.81% (16.70% to 32.80%)
Local swelling/redness	13	49	26.53% (16.20% to 40.30%)
Substance misuse or withdrawal	12	56	21.43% (12.70% to 33.80%)
No Peer shifts	41	196	20.92% (15.80% to 27.10%)
Local swelling/redness	23	129	17.83% (12.20% to 25.30%)
Substance misuse or withdrawal	18	67	26.87% (17.70% to 38.50%)

^a95% CI based on Wilson’s score interval.

We disaggregated results by presenting complaints. For substance misuse/intoxication or withdrawal visits, among 55 Peer shifts, the mean LBTC percentage was 22.1% (SD 33.5%) vs 32.7% (SD 38.0%) among 29 non-Peer shifts. This corresponded to a nonsignificant difference of −10.4% (95% CI −26.8%, 6.1%). For localized swelling/redness visits, among 72 Peer shifts, the mean LBTC percentage was 14.5% (SD 21.9%) vs 20.5% (SD 30.8%) during 35 non-Peer shifts. This corresponded to a nonsignificant difference of −6.0% (95% CI −17.4%, 5.4%).

3.3.2. Secondary patient-level analysis

Among all patients who interacted with a Peer compared with those who did not, LBTC proportions were not significantly different on adjusted patient-level analyses: 23.8% (25/105) LBTC compared with 20.9% (41/196) who did not interact with a Peer (OR 0.78 [95% CI 0.46, 1.32]). Disaggregated by presenting complaint, among patients with substance misuse/intoxication or withdrawal, Peer support significantly reduced the absolute risk of LBTC by 5.5% (12/56 [21.4%] vs 18/67 [26.9%]; OR 0.39 [95% CI 0.18, 0.84]). Among patients who presented for localized swelling/redness and interacted with a Peer compared with those who did not, LBTC proportions were not significantly different: 13/49 (26.5%) LBTC, compared with 23/129 (17.8%) among those who did not interact with a Peer (OR 1.06 [95% CI 0.50, 2.25]).

3.3.3. Sensitivity analysis

3.3.3.1. LWBS vs LAMA

Our sensitivity analyses showed trends toward decrease but no significant difference in shift-level LWBS or LAMA among Peer vs control shifts. Among 72 Peer shifts, the mean LWBS percentage was 10.5% (SD 14.2%). Among 37 control (no Peer) shifts, the mean LWBS percentage was 14.0% (SD 21.8%). The difference in the mean shift-level LWBS percentage between Peer (intervention) and no Peer (control) shifts was −3.5% (95% CI −11.2%, 4.3%). The mean LAMA percentage was 6.4% (SD 10.7%) for Peer shifts and 9.2% (SD 15.1%) for control shifts. The difference in the mean shift-level LAMA percentage between Peer (intervention) and no Peer (control) shifts was −2.8% (95% CI −8.2%, 2.7%).

For substance misuse/intoxication or withdrawal visits, the difference in the mean shift-level LWBS percentage between Peer (intervention) and no Peer (control) shifts was −9.8% (95% CI −23.7%, 4.1%), and the difference in LAMA was −2.0% (95% CI −14.0%, 9.9%). For localized swelling/redness shifts, the difference in the mean shift-level LWBS percentage between Peer (intervention) and no Peer (control) shifts was −1.9% (95% CI −10.9%, 7.2%), and the difference in LAMA was −4.2% (95% CI −11.8%, 3.5%).

For the patient-level analysis, among all patients who interacted with a Peer compared with those who did not, there was a significant decrease in odds of LWBS among patients with substance misuse/intoxication or withdrawal (OR 0.33 [95% CI 0.13, 0.84]), but no difference in LWBS for the overall cohort (OR 0.57 [95% CI 0.29, 1.13]) or among patients with localized swelling/redness (OR 0.87 [95% CI 0.32, 2.35]). We found no difference in odds of LAMA among patients who interacted with a Peer compared with those who did not for the overall cohort (OR 1.12 [95% CI 0.58, 2.16]), among patients with substance misuse/intoxication or withdrawal (OR 0.68 [95% CI 0.26, 1.80]), or among patients with localized swelling/redness (OR 1.07 [95% CI 0.42, 2.74]).

3.3.4. Exploratory analysis: naloxone kit order

Among patients who interacted with a Peer during intervention shifts, 22.9% (24/105) had a naloxone kit ordered during their ED visit, compared with 13.8% (27/196) among control patients. In adjusted patient-level analysis, patients who interacted with a Peer were significantly more likely to have naloxone ordered than those who did not (OR 2.33 [95% CI 1.34, 4.04]). Results were not significantly different when patients were disaggregated by presenting complaint: substance misuse/intoxication or withdrawal (OR 1.82 [95% CI 0.73, 4.56]) or localized swelling/redness (OR 2.06 [95% CI 1.00, 4.27]).

3.3.5. Exploratory analysis: 48-hour return ED visits

Among patients who interacted with a Peer during intervention shifts, 9.5% (10/105) had a return ED visit within 48 hours, compared with 7.7% (15/196) among control patients. In adjusted patient-level analysis, patients who interacted with a Peer were significantly less likely to have a 48-hour return ED visit than those who did not for the overall cohort (OR 0.46 [95% CI 0.22, 0.95]), and among those with substance misuse/intoxication or withdrawal (OR 0.31 [95% CI 0.11, 0.90]). We found no significant difference among patients with localized swelling/redness (OR 0.78 [95% CI 0.28, 2.17]).

3.4. Harms

We did not identify unintentional harms to patients. There were no protocol deviations.

3.5. Ancillary Analyses

We conducted sensitivity analyses reclassifying the 2 visits made by patients during control (no Peer) shifts who remained in the ED for over 24 hours and interacted with a Peer during a subsequent shift, which produced similar results. First, we excluded these from the analysis and found a similar nonsignificant difference in LBTC among all visits (shift level −6.3% [95% CI −14.8%, 2.3%]; patient level OR 0.83 [95% CI 0.49, 1.40]). Among visits for substance misuse/intoxication or withdrawal we found a similar nonsignificant difference at a shift level (−11.8% [95% CI −28.5%, 4.8%]) and slightly significant decrease at a patient-level (OR 0.45 [95% CI 0.21, 0.97]). Second, we reclassified these 2 visits to the intervention (Peer) group for the primary analysis and found a similar nonsignificant difference in LBTC among all visits (−6.3% [95% CI −14.8%, 2.3]) and those for substance misuse/intoxication or withdrawal (−11.8% [95% CI −28.5%, 4.8%]).

4. Limitations

Our pragmatic study is limited by real-world confounders. Our Peer program was implemented concurrently with multiple other ED substance use services and supports, challenging determination of incremental benefit attributable to the Peer intervention. We were unable to track other substance use interventions likely deployed during Peer shifts. Additionally, we recognize that many factors contribute to our outcome of interest, LBTC: for example, whether patients LWBS or LAMA depends on many factors other than the presence of a Peer, including wait times, crowding, and boarding. The presence of confounders highlights the importance of a randomized design to address our study question. To the best of our ability, shift randomization controlled for confounding from concurrent interventions and other trends and also mitigated differences between intervention and control shifts that could have contributed to our outcome of interest. Second, in this real-world intervention, the Peer identified eligible patients through a combination of methods including electronic medical record screening, making rounds of the ED, and receiving referrals, introducing potential selection bias and making identification of a comparable control group during non-Per shifts challenging. We addressed this by evaluating shift-based outcomes for all patients with prespecified presenting complaints during intervention (Peer) and control (no Peer) shifts. Including all patients with certain presenting complaints in our primary analysis mitigated selection bias (we applied an identical method to identify intervention and control groups) and volunteer bias (we included eligible patients whether or not they interacted with a Peer, eg, intention-to-treat). Our design is limited in that there was a proportion of eligible patients within the intervention group who, although eligible, did not interact with the Peer. We also recognize that our methodological approach would dilute findings of Peer effectiveness and bias the observed effect size toward the null. Conversely, although all patients who we included in our secondary patient-level analysis interacted with the Peer, we recognize that this approach introduced selection bias among those who were identified by the Peer for the intervention. Third, this pilot study included a single urban ED with strong support from operations and institutional leadership to implement our ED Peer program (eg, dedicated recruitment, training, oversight, funding). Additionally, our study was conducted in a Canadian ED where patients have access to universal, publicly funded healthcare services. Our site specificities may limit generalizability to other settings, including to EDs in other health systems such as the United States’ private-pay model. Fourth, during our pilot period, we only hired 1 Peer who filled a single full-time daytime line. Employment of a single Peer may have introduced bias in patient selection and intervention delivery, limiting reliability and generalizability of results. Because the Peer helped to validate and refine the intervention, it may have been tailored to their individual strengths and comforts, which in turn may have influenced intervention effectiveness. This limits our ability to draw conclusions on the intervention’s feasibility for larger-scale rollout, but nonetheless, our study sets the stage for future studies that evaluate a more broadly staffed Peer program. Future studies should include a larger number of Peers in the intervention to enhance generalizability. Fourth, our analysis was not powered for our outcomes of interest. Rather, we conducted the present analyses in a convenience sample of ED visits made during our 15-week study period, which was determined by funding and resource availability for our analysis. This limitation in sample size should be addressed in a future definitive trial. Finally, due to resource constraints, we were unable to blind outcome assessors, which could have introduced bias. We will maintain blinding of outcome assessors in a future definitive trial.

To our knowledge, our study is the first to examine effectiveness of an ED Peer program at decreasing LBTC and significantly contributes to our understanding of how Peers may improve health outcomes of patients who use substances. LBTC is a critically important outcome: it is measurable within the ED visit (therefore, impervious to losses to follow-up), may be directly influenced by Peers within the ED visit, and has established clinical importance. Existing literature has strongly linked LBTC to adverse health outcomes and mortality.7, 8, 9, 10, 11, 12, 13, 14 Although preliminary, our finding of a possible signal that the Peer intervention may be associated with lower LBTC among a subset of patients with substance use-related complaints is comparable to other studies evaluating interventions to decrease LBTC in related populations.51, 52, 53, 54 Our study is aligned with previous studies that show potential benefit of ED Peers in improving patient outcomes,^31,41 but is also consistent with existing evidence that shows equivocal effectiveness.38, 39, 40^,42,43 By showing a possible signal that an ED Peer intervention may be associated with decreases in LBTC among a subset of patients with substance use, our study importantly advances current knowledge by theorizing a potential mechanism for the beneficial effect of ED Peers.

5. Discussion

5.1. Interpretation

Our pilot study demonstrated the feasibility of procedures required to operationalize a randomized trial evaluating Peer effectiveness (randomized schedule implementation) and high (94.9%) acceptability of the intervention among our target population. In our primary analysis, LBTC appeared 6.1% lower during Peer compared with control shifts, but this was a nonsignificant difference. To ensure comparability between our intervention and control groups and to avoid selection bias, we included patients with the 3 most common substance use-related complaints in our primary intention-to-treat analysis, whether or not they interacted with the Peer. We expect that this approach diluted any finding of Peer effectiveness and biased the observed effect size toward the null. Our secondary patient-level analysis showed that LBTC was no different when patients interacted with the Peer among all patients with substance use-related complaints; the nonsignificant result appeared to be driven by a null effect within the subset of patients with localized swelling and redness. In contrast, our patient-level analysis among patients with substance misuse/intoxication or withdrawal indicated that interaction with a Peer was associated with a 5.5% decrease in LBTC and significantly reduced odds of LBTC. In adjusted exploratory analyses, patients who interacted with a Peer during Peer shifts had significantly higher odds of having a naloxone kit ordered and significantly lower odds of having a return ED visit within 48 hours. If a decrease in LBTC at a shift level and/or among subsets of patients with substance use-related complaints was demonstrated in a larger sample and statistical significance was shown, or if significant improvements in patient-centered health outcomes like receipt of critical medical treatments (eg, naloxone kits) or decreased ED return visits were confirmed, validation of our preliminary findings would have clinical and public health importance. Given that substance use-related ED visits are rapidly increasing⁴⁹ and unprecedented ED strain is driving concerningly high levels of LBTC,⁵⁰ an intervention that could decrease rates by 5% to 6% could result in thousands more patients who engage in care in British Columbia yearly. If, building on our preliminary results, future studies demonstrate that a relatively low-impact Peer intervention increases patient completion of ED care and consequently improves patient-oriented outcomes, ED Peers have the potential to be a feasible solution that could be implemented on a broader scale. Our findings support the need for a well-powered future trial to confirm the magnitude of benefit of ED Peers.

5.2. Clinical Implications

This study demonstrates that an ED Peer program to support patients with substance use is feasible and may improve engagement of a subset of patients who use substances in their ED care.

5.3. Research Implications

This study highlights the need for a well-powered randomized controlled trial to evaluate effectiveness of ED Peer programs at decreasing LBTC. Our results indicate that a shift-based analysis will require a greater sample size than an analysis of individual patients. Future studies should additionally evaluate important clinical outcomes that could be impacted if Peers improved patient engagement in care. These include initiation of and retention in treatments for substance dependence (eg, medication or opioid or alcohol use disorders), future overdoses, and mortality. Our study supports feasibility of a future definitive trial at sites where there is operational support for Peer programs.

6. Conclusion

An ED Peer intervention to support people who use substances is a feasible intervention. A randomized controlled trial is feasible and necessary to assess impact of an ED Peer intervention on decreasing percentage of patients who LBTC.

Author Contributions

JM conceptualized the study. JM, HW, and EYW developed the analytic plan, linked the data, and created results and analytic summaries. LG and ZB provided operational leadership with Peer hiring, onboarding, and day-to-day management. EYW, QL, and KD assisted with implementation and data collection. ID represented Peer perspectives in Peer program development and implementation. LQ provided operational support for Peer program implementation. JH assisted with data collection tools. All authors assisted with results interpretation and have reviewed and approved the final version of the manuscript for publication. The investigators controlled all aspects of data collection, analysis, and report writing.

Funding and Support

This work received funding from the British Columbia Centre for Disease Control and Vancouver Coastal Health Research Institute.

Data Sharing Statement

Study participants did not provide consent for their data to be shared publicly; hence, due to the sensitive nature of the research, supporting data are not available.

Conflict of Interest

Jessica Moe reports financial support was provided by the British Columbia Centre for Disease Control. Jessica Moe reports financial support was provided by the Vancouver Coastal Health Research Institute. Other authors have affirmed they have no conflicts of interest to declare.