Peer providers and linkage with buprenorphine care after hospitalization: A retrospective cohort study

Abstract

Background:

People with opioid use disorder (OUD) are increasingly started on buprenorphine in the hospital, yet many patients do not attend outpatient buprenorphine care after discharge. Peer providers, people in recovery themselves, are a growing part of addiction care. We examine whether patients who received a low-intensity, peer-delivered intervention during hospitalization had a greater rate of linking with outpatient buprenorphine care relative to those not seen by a peer.

Methods:

This was a retrospective cohort study of adults with OUD who were started on buprenorphine during hospitalization. The primary outcome was receipt of a buprenorphine prescription within 30 days of discharge. Secondary outcomes included attendance at a follow-up visit with a buprenorphine provider within 30 days and hospital readmission within 90 days. Modified Poisson regression analyses tested for differences in the rate ratios (RR) of each binary outcome for patients who were versus were not seen by a peer provider. Peer notes in the electronic health record were reviewed to characterize peer activities.

Results:

111 patients met the study inclusion criteria, 31.5% of whom saw a peer provider. 55.0% received a buprenorphine prescription within 30 days of hospital discharge. Patients with versus without peer provider encounters did not significantly differ in the rates of receiving a buprenorphine prescription (RR = 1.06, 95% CI: 0.74–1.51), hospital readmission (RR = 1.45, 95% CI: 0.80–2.64), or attendance at a buprenorphine follow-up visit (RR = 1.03, 95% CI: 0.68–1.57). Peers most often listened to or shared experiences with patients (68.6% of encounters) and helped facilitate medical care (60.0% of encounters).

Conclusions:

There were no differences in multiple measures of buprenorphine follow-up between patients who received this low-intensity peer intervention and those who did not. There is need to investigate what elements of peer provider programs contribute to patient outcomes and what outcomes should be assessed when evaluating peer programs.

Background

Treatments for opioid use disorder (OUD), including methadone and buprenorphine are effective,^1–3 but access is limited.⁴ At least two thirds of people with OUD in the US are untreated,⁵ and many people do not receive treatment until years after OUD begins.⁴ To enhance access to evidence-based medications for OUD (MOUD), providers are increasingly initiating buprenorphine in acute care settings, such as the emergency department or hospital.

For patients who initiate MOUD in the hospital, linkage with outpatient buprenorphine follow-up is essential for providing continuity of buprenorphine treatment,^6–9 which in turn affects the likelihood of maintaining long-term reductions in opioid use^9,10 and reducing the risk for overdose.¹¹ Estimates of the percentages of patients who continue to receive buprenorphine after initiating buprenorphine in the hospital vary widely, from 40% to over 70%.^12–16 In one randomized controlled trial, hospitalized patients with OUD were randomized to five-day buprenorphine detoxification or buprenorphine induction and referral for outpatient follow-up.⁹ Patients randomized to buprenorphine induction were more likely to engage in outpatient buprenorphine care (72.2% versus 11.9%) and had less self-reported opioid use at one, three, and six months than patients who underwent detoxification.⁹ Among those randomized to buprenorphine induction, patients who had longer hospitalizations, more trauma symptoms, or had ever been treated with buprenorphine were more likely to attend their first follow-up appointment.¹⁷ Another study found that patients who had a buprenorphine follow-up within the first day of hospital discharge were more likely to link with care than those who had an appointment two or more days from discharge.¹⁶ Linkage to outpatient follow-up can be challenging, as many patients may face barriers to accessing care, including lack of transportation and low trust in the healthcare system.^18,19 Various approaches to facilitate this linkage after discharging from the hospital have been utilized, including “bridge” clinics,¹⁸ direct referral to an outpatient provider,²⁰ and brief counseling in the emergency department.

Peer providers could play a role in supporting patients who initiate buprenorphine in the hospital to link to outpatient MOUD treatment.²¹ Peer providers, also known as peer support specialists or recovery coaches, are typically individuals who are in recovery, have limited training in healthcare, and bring expertise from lived experience to support people with substance use disorders (SUDs).²² Peer provider activities are heterogeneous and can range from care coordination and case management to informal discussions of substance use and other life stressors²³ to more structured therapy.²⁴ Numerous qualitative studies that included interviews with hospitalized patients with SUD found that patients valued interaction with a peer while in the hospital and appreciated peer assistance with outpatient care planning.^25,26

Although the peer provider role is being scaled up nationally,^27,28 there has been limited research on peer providers and buprenorphine treatment engagement. A recent retrospective cohort study found that receiving peer services was associated with greater odds of attending OUD treatment appointments relative to usual care. In that study, peers coordinated clinic visits, accompanied patients to appointments, and met with patients at least weekly, with more intensive outreach to patients who were struggling with recovery.²⁹ Similarly, a cross-sectional sub-analysis within a larger study found that receipt of a buprenorphine prescription was higher among patients who had contact with a peer provider than those who had not.²¹ No studies have focused on peer providers to support patients who initiate buprenorphine in the hospital and receive peer services during an inpatient admission.

Accordingly, we compared linkage to outpatient buprenorphine treatment among hospitalized patients initiated on buprenorphine who did versus did not receive relatively low-intensity support from a peer provider while in the hospital. We hypothesized that contact with the peer provider would be associated with higher rates of receiving a buprenorphine prescription within 30 days of hospital discharge. To better illustrate the content of the peer intervention, we also characterized and quantified the services that peers documented providing to patients during their hospitalization. Importantly, there are many possible meaningful outcomes of peer provider programs, including satisfaction with healthcare, quality of life, and indicators of substance use. Here, we have chosen to focus on linkage with buprenorphine care, as it is associated with reductions in opioid use^9,10 and is an important step in the OUD care continuum and a place where peer providers may be able to play a valuable role.

Methods

Peer provider program

A peer provider program was established at Harborview Medical Center, an urban county hospital in Seattle, Washington, in 2017. The program was funded through external grants and is part of a larger, hospital-wide initiative to increase SUD treatment and harm reduction. This initiative included the establishment and growth of an inpatient Addiction Consult Service (ACS) and expansion of outpatient buprenorphine treatment.

Referral to the peer provider program

The peer provider program was created to broadly support addiction services at the hospital and had up to three peer providers during the study period. Peers were required to have two years of recovery and attend a 40-hour state-sponsored training on peer support.³⁰ Peers were tasked with meeting with hospitalized patients, running an outpatient mutual help group for patients on MOUD, and documenting each encounter in the electronic health record (EHR). Given the large number of patients seen by the ACS and limited number of peer providers, the peer providers were generally only able to engage with a portion of the patients on service. Patients who were referred to a peer provider typically had perceived ambivalence, prolonged hospitalization, or behavioral or emotional challenges, as assessed by the physicians, nurses, and social workers on the ACS; there were no specific criteria patients needed to meet to be assigned a peer.

Peer role in the hospital

Peers met with inpatients to have open-ended conversations about substance use and recovery, without a standardized format, and to administer the Government Performance and Results Act (GPRA) survey, which was part of mandatory reporting for the grant that funded the peer program. Peers then reported any concerns back to the ACS. Once patients were discharged, peers primarily reached out to them to encourage them to attend the weekly mutual help group for patients on MOUD. The peers were not specifically told to discuss follow-up or MOUD with hospitalized patients or provide any support around MOUD adherence or appointment attendance after discharge.

Buprenorphine follow-up appointments

When ACS providers started hospitalized patients on buprenorphine, the ACS program coordinator arranged buprenorphine follow-up for the patient. If the patient lived near the hospital and did not have a primary care provider who prescribed buprenorphine, the patient would be referred to the bridge clinic, which would then arrange follow-up with a community-based opioid treatment program or primary care clinic. If the patient lived far away from the hospital, the ACS program coordinator would arrange a buprenorphine intake visit with a provider near the patient’s home. All patients were discharged with a specific follow-up appointment as well as a buprenorphine prescription that would last them until that appointment, but patients were then responsible for attending that appointment; the clinics were not specifically tasked with doing further outreach to these patients.

Participants

We performed a retrospective cohort study of patients with OUD who were started on buprenorphine during hospitalization. Eligible patients were (1) admitted between March 1, 2019 and February 29, 2020, (2) at least 18 years of age, (3) prescribed buprenorphine for OUD at the time discharge (new start only; not including patients with buprenorphine prescriptions at admission, as determined using Washington Prescription Monitoring Program [PMP] data, into which pharmacies enter all buprenorphine prescriptions at the time they are dispensed, and medical notes documenting buprenorphine as part of the preadmission care plan), and (4) were seen by the ACS during their hospitalization. Patients were excluded if they left against medical advice, eloped, or were transferred to a long-term care or correctional facility.

Demographics and clinical characteristics

Patient information was abstracted from Government Performance and Results Act (GPRA) survey data,³¹ which was collected on most patients seen by the ACS for grant reporting purposes. Measures obtained from the GPRA included demographic information, self-reported substance use during the 30 days prior to admission (substances used and the route of administration), treatment history, and history of overdose. If the GPRA was not administered to the patient or the patient declined to complete all or part of the GPRA, researchers looked in the EHR for this information. Measures obtained from the EHR for all patients included admission and discharge dates, insurance status, and the number of interactions with peer providers during and after hospitalization. These data were obtained by chart reviews and were entered into REDCap.^32,33

Outcome measures

Primary outcome

The primary outcome was a binary measure indicating receipt of a buprenorphine prescription within 30 days of hospital discharge. Receipt of a prescription was determined using data obtained from the EHR, which included PMP data. PMP data was only accessible in the EHR for one year prior to the date of data retrieval (data retrieval 7 June 2020 through 19 August 2020), and PMP data outside the EHR was not accessed. The EHR was also examined for buprenorphine prescriptions entered by outpatient providers. Unlike PMP data, the EHR would only capture prescriptions written within a single medical system. We treated lack of records for medication orders, outpatient buprenorphine visit, or hospitalization as an indication that there was no medication prescribed, no outpatient visit, or no repeat hospitalization. Patients who were discharged from the hospital more than one year prior to the date of data retrieval and who were not scheduled for a follow-up visit within the medical system were excluded from the primary analysis. We performed a sensitivity analysis in which these participants were retained in the sample and assumed to have not received a buprenorphine prescription within 30 days to examine whether this changed the conclusions.

Secondary outcomes

Secondary outcomes included binary indicators of (1) attendance at a follow-up visit with an outpatient buprenorphine provider within 30 days of discharge, (2) number of days prescribed buprenorphine within the 90 days of discharge, and (3) hospital readmission within 90 days. For appointment attendance, patients who were referred to outside providers were excluded from this analysis (25 patients, 23% of patients included in the primary analyses) as such visits would not be captured in the EHR. For readmission, only readmissions to hospitals within the same medical system were captured.

Peer provider activities

Peers documented each encounter in a one-paragraph narrative note in the EHR. There was no standard template; therefore, documentation was at the peer’s discretion. To gather details on the content of these encounters, a single member of the research team (EDD, BAC, or DS) reviewed each note, assigned one or more pre-determined categories to the activities described, and collected illustrative quotes. The categories were deductively developed, using a process informed by rapid appraisal techniques for qualitative data,³⁴ as has been previously applied to real-world analysis of health services implementation.^35,36 Three addiction medicine physicians, two peer providers, two nurses, and the OUD program coordinator met to determine the categories of peer interaction with patients, drawing on their clinical experience, the peer provider job description, and existing literature describing peer provider activities in a general medical setting.²³ The peers were then consulted about the categories and provided feedback. The four categories were: (1) witness to life: sharing life experiences about substance use or other challenges, (2) role model and recovery coach: talking about recovery or harm reduction goals and supports, (3) facilitating medical treatment: discussing inpatient care, care transitions, and health issues, and (4) facilitating other services: helping the patient connect with social services, such as housing or transportation assistance. Descriptive statistics were used to quantify the mean number of encounters per patient and number of encounters that contained each of the four activities.

Statistical analysis

Descriptive analyses were conducted among patients with and without visits from the peer provider to characterize the demographic and clinical measures of both groups. Potential differences between these groups were evaluated using Fisher’s exact tests given the small numbers of patients expected in some demographic groups. Outpatient buprenorphine prescription receipt, appointment follow-up, and readmission were compared, and rate ratios (RRs) characterizing differences in the rates of linkage and hospital readmission for patients with versus without peer provider visits were computed using modified Poisson regression (with and without controlling for baseline covariates: age, gender, race/ethnicity, history of MOUD treatment, insurance, and length of hospitalization).³⁷ Covariates were selected based on discussion amongst investigators about what confounding factors may impact access to outpatient treatment.

A power analysis was conducted using GPower software³⁸ to determine the minimum effect size required to achieve 80% power for detecting a significant difference in the rates of buprenorphine prescriptions for patients who received peer navigation compared to patients who did not. For the primary outcome of a buprenorphine prescription within 30 days of discharge, approximately 80.7% of patients who received peer navigation (versus 53.9% of patients who did not receive peer navigation; RR = 1.50) would have needed to receive a buprenorphine prescription within 30 days to have 80% power.

The University of Washington Institutional Review Board approved this study (STUDY00010156).

Results

Sample characteristics

One hundred and eleven patients were included in the sample, 35 of whom were seen by a peer provider. Most patients were age 25–44 (51.4%) or 45–64 (38.7%). 68.5% of the patients were men, and 58.6% self-identified as white. Median length of hospitalization as 10 days (IQR 4.5–21 days), and most patients were admitted to a medical service (i.e., not a surgical or psychiatric service). More than a third were experiencing homelessness, and most were on Medicaid (70.3%). Many had a history of MOUD treatment: 27.0% had previously received buprenorphine, 11.7% methadone, and 14.4% both buprenorphine and methadone (not concurrently); 45.9% had used injection drugs in the prior 30 days. No significant differences were observed (p < .05) between patients who were seen by a peer provider and those who were not for any measures examined (Table 1). Supplemental Table 1 lists categories of discharge diagnoses for patients in both groups.

Table 1.

Descriptive statistics for the full sample, including patients who saw a peer provider while in the hospital and those who did not.

	Full sample (n = 111)		Patients with peer encounter (n = 35)		Patients without peer encounter (n = 76)		p Valuea
	n	%	n	%	n	%
Age
18–24	6	5.4%	3	8.6%	3	3.9%	.62
25–44	57	51.4%	16	45.7%	41	53.9%
45–64	43	38.7%	15	42.9%	28	36.8%
65+	5	4.5%	1	2.9%	4	5.3%
Gender
Man	76	68.5%	22	62.9%	54	71.1%	.39
Woman	35	31.5%	13	37.1%	22	28.9%
Race
American Indian or Alaskan Native	6	5.4%	1	2.9%	5	6.6%	.25
Asian or Asian American	2	1.8%	0	0.0%	2	2.6%
Black or African American	13	11.7%	4	11.4%	9	11.8%
Native Hawaiian or Pacific Islander	2	1.8%	0	0.0%	2	2.6%
White	65	58.6%	20	57.1%	45	59.2%
More than one race	3	2.7%	3	8.6%	0	0.0%
Unknown race	20	18.0%	7	20.0%	13	17.1%
Ethnicity
Hispanic or Latino	8	7.2%	3	8.6%	5	6.6%	.70
Not Hispanic or Latino	91	82.0%	27	77.1%	64	84.2%
Unknown	12	10.8%	5	14.3%	7	9.2%
Housing
Shelter/street/outdoors	41	36.9%	12	34.3%	29	38.2%	.69
Institution or housed	52	46.8%	18	51.4%	34	44.7%
Unknown/missing	18	16.2%	5	14.3%	13	17.1%
Medicaid insurance	78	70.3%	28	80.0%	50	65.8%	.18
Length of hospital stay
1–5 days	37	33.3%	9	25.7%	28	36.8%	.11
6–14 days	33	29.7%	8	22.9%	25	32.9%
15–133 days	41	36.9%	18	51.4%	23	30.3%
Hospital service
Medical service	62	55.9%	22	62.9%	40	52.6%	.66
Surgical service	38	34.2%	10	28.6%	28	36.8%
Psychiatry	8	7.2%	3	8.6%	5	6.6%
Other	2	1.8%	0	0.0%	2	2.6%
Unknown	1	0.9%	0	0.0%	1	1.3%
Trauma-related admission
Yes	36	32.4%	9	25.7%	27	35.5%	.38
No	74	66.7%	26	74.3%	48	63.2%
Unknown	1	0.9%	0	0.0%	1	1.3%
Substance use in past 30daysb
Alcohol to intoxication (5+ drinks in 1 sitting)	8	10.1%	2	8.7%	6	10.7%	1.00
Cocaine/crack	15	19.2%	4	19.0%	11	19.3%	1.00
Opioids	78	83.9%	25	83.3%	53	84.1%	1.00
Methamphetamines or other amphetamines	43	55.1%	14	63.6%	29	51.8%	.49
Benzodiazepines	10	14.1%	2	9.5%	8	14.1%	.73
Injected drugs in past 30 days
Yes	51	45.9%	15	42.9%	36	47.4%	.81
No	33	29.7%	11	31.4%	22	28.9%
Unknown	27	24.3%	9	25.7%	18	23.7%
History of MOUD treatment
Buprenorphine only	30	27.0%	13	37.1%	17	22.4%	.21
Methadone only	13	11.7%	4	11.4%	9	11.8%
Buprenorphine and methadone	16	14.4%	7	20.0%	9	11.8%
Neither	29	26.1%	7	20.0%	22	28.9%
Unknown	23	20.7%	4	11.4%	19	25.0%
History of overdose
Yes	16	14.4%	5	14.3%	11	14.5%	.46
No	26	23.4%	5	14.3%	21	27.6%
Unknown	69	62.2%	25	71.4%	44	57.9%

^ap values reflect difference between linked and non-linked patients and were estimated using Fischer’s exact tests for categorical variables.

^bData are from self-report, and levels of missing data varied by substance. Number and percentage of missing are as follows: alcohol to intoxication 32 (28.8%), cocaine/crack 33 (29.7%), opioids 18 (16.2%), methamphetamines or other amphetamines 33 (29.7%), benzodiazepines 40 (36.0%). Missing data were omitted from the calculation of the percentage and the bivariate analysis.

Peer provider activities

As shown in Table 2, among the patients who had at least one documented encounter with a peer provider, 31.4% had two or more contacts (range 1–4). In the encounter notes, peer providers often documented more than one category of activity (mean: 1.8, SD: 1.0), most frequently witness to life (68.6% of encounters) or facilitating medical treatment (60.0% of encounters). Peers documented discussing MOUD with 57.1% of patients. Quotes illustrating the categories of activities, drawn from peer notes, are shown in Table 2.

Table 2.

Number and content of peer provider encounters with patients: results of review of peer notes in the electronic health record.

	n	%
Peer provider encounters
Number of encounters for those who linked with a peer provider
One	24	68.6%
More than one	11	31.4%
Peer provider activities
Types of activities
Witness to life	24	68.6%
Role model/recovery coach	10	28.6%
Facilitate medical treatment	21	60.0%
Facilitate other services	7	20.0%
Number of activities received per person, Mean (SD)	1.8 (1.0)
Discussed buprenorphine, methadone, or naltrexone (number of patients)	20	57.1%
Discussed follow-up appointments (number of patients)	7	20.0%
Category of peer provider activity	Definition	Representative quotes from the medical record
Witness to life	Listening to the patient describe their substance use history and other challenges; peers sharing details of their own substance use history	“Initially, he did not show any interest in engaging with this writer until she self-disclosed as being a person with a history of opioid use disorder. The [patient] told this writer, ‘I’m a little tired, but let’s do this.’” “The [patient] spoke about his feelings in regards to not wanting to live any longer due to his 24-year old daughter recently committing suicide.” “When she tried to get sober before, she did not know what to do with her time, and this writer told her some of my story, and offered to help her through detoxing from the lifestyle and building a new life. Patient seemed more upbeat.”
Role model and recovery coach	Helping the patient identify recovery goals and develop recovery supports	“This writer spoke with the [patient] about the importance of counseling ...using a portion of her recovery journey in efforts to educate, inspire and encourage.” “Patient was very interested in suboxone ... this writer gave patient a flyer and a card with the info about our MARS [Medication Assisted Recovery Support] ... Group, and encouraged patient to use the support.”
Facilitating medical treatment	Discussing OUD, mental health care, primary care, harm reduction services, and the transition from inpatient to outpatient care	“Patient was ... confused about his discharge [medications] ... we took it to our team for follow-up.” “Patient was also asked how she was feeling on the micro dose [low-dose induction for buprenorphine], and she said that she feels like her dose needs to be raised.”
Facilitating other services	Supporting the patient to access benefits, housing, transportation, insurance, and other social services; discussing independent living skills, such as budgeting, diet, and exercise	“Patient was asking about getting placed into a [skilled nursing facility], and this writer did speak with patient social worker who said that they are working in finding patient a place that will accept her.”
		“This writer spoke with patient and patient had real concerns about her discharge plan. Social worker explained the plan as being [skilled nursing facility] or shelter. This writer encouraged patient to take the [skilled nursing facility] option.” “This writer called the SW [social worker] and asked that they give a referral to the HARPS program for housing.”

Buprenorphine linkage

As displayed in Table 3, 55.0% of patients received a buprenorphine prescription within 30 days of discharge, and they received it a mean of 8.3 days (SD: 6.3) from hospital discharge. There was no statistically significant difference in the rates of receipt of a buprenorphine prescription within 30 days (primary outcome) between patients who had at least one peer encounter and those who had none (RR = 1.06, 95% CI: 0.74–1.51, p = .75). Notably, 23 patients of the 134 patients who would have been eligible for inclusion in the sample were excluded, as they did not have PMP data available because they were discharged from the hospital more than one year prior to the date of data retrieval and were not scheduled to follow-up within the medical system. Results of a sensitivity analysis in which these patients were included and assumed not to have received a buprenorphine prescription were consistent with the primary analysis.

Table 3.

Comparison of clinical outcomes for patients with and without peer provider encounters.

	Peer encounter 35 (31.35%)		No peer encounter 76 (68.47%)		Unadjusted model			Adjusted modela
Primary outcome measure	N	%	N	%	RR	95% CI	p value	RR	95% CI	p Value
Buprenorphine prescription recorded within 30 days of discharge	20	57.1%	41	53.9%	1.06	0.74, 1.51	.75	1.00	0.69, 1.44	.99
Secondary outcome measures
Readmission to same medical system within 90 days of discharge	12	34.3%	19	25.0%	1.45	0.80, 2.64	.22	1.70	0.91, 3.18	.10
Attended buprenorphine follow-up visit within 30 days of dischargeb	16	53.3%	29	51.8%	1.03	0.68, 1.57	.89	1.00	0.64, 1.56	.99
	Mean	SD	Mean	SD	Difference	95% CI	p value	Difference	95% CI	p Value
Number of days of buprenorphine prescribed within 90 days of discharge	32.7	33.5	26.9	32.7	5.82	−7.66, 19.31	.39	4.37	−9.68, 18.42	.54

^aAdjusted on age (<45 years and >45 years), gender (man and woman), race/ethnicity (white non-Hispanic, all BIPOC subgroups including Hispanic, and unknown), history of MOUD (yes, no, and unknown), insurance (Medicaid/unknown or non-Medicaid), and length of hospitalization (1 to 5 days, 6 to 14 days, and 15 to 133 days).

^bSample based on those who were originally scheduled for a follow-up appointment at a clinic within the same healthcare system as the hospital admission

There were similarly no differences in the number of days of buprenorphine filled within 90 days (difference = 5.82 days, 95% CI: −7.66 to 19.31, p =.39), rates of hospital readmissions within 90 days (RR = 1.45, 95% CI: 0.80–2.64, p = .22), or rates of attendance at buprenorphine follow-up visit within 30 days among those scheduled for follow-up within the hospital system (RR = 1.03, 95% CI: 0.68–1.57, p = .89; see unadjusted models in Table 3). Patients who attended a follow-up visit within the hospital system within 30 days of discharge did so a mean of 8.4 days (SD: 6.5) from hospital discharge. When the analyses were adjusted for age, gender, race/ethnicity, history of MOUD treatment, insurance, and length of hospitalization, there remained no differences between groups in receipt of a buprenorphine prescription within 30 days, attendance at follow-up, number of days of buprenorphine filled, or hospital readmissions (see adjusted models in Table 3). Additionally, no patients were initiated on buprenorphine in this hospital more than once during the study period, so was no need to account for repeat patients in the analyses.

Conclusions

This retrospective, observational study of a relatively low-intensity peer provider program did not demonstrate a difference in the rates of receiving a buprenorphine prescription within 30 days of discharge for hospitalized patients who did versus did not have at least one encounter with a peer provider. To characterize the activities of peer providers, we analyzed the notes that peer providers wrote in the EHR, which could inform refinement of peer interventions by facilitating testing of which components contribute to outcomes.

These results differ from the two prior studies on peers and linkage to buprenorphine care, both of which found that contact with a peer was associated with greater odds of buprenorphine linkage.^21,29 Both of these studies, however, involved a multi-session peer intervention in the outpatient setting. The more rigorous of the two prior studies, which used a retrospective cohort design, had a much more intensive peer intervention than in our study. In that case, peers saw patients at least weekly and accompanied them to medical appointments. In our study, peers saw most patients only once. It is possible that there could be a dose-response for peer provider visits, with more visits leading to a greater change in outcomes of interest. There are several other possible explanations for why the patients who saw a peer provider in this study were not more likely to link with outpatient buprenorphine care. First, the ACS referred patients to a peer provider whom they believed needed additional support. As such, patients who received peer provider services were potentially at higher risk for non-engagement after discharge. Although no differences in demographic characteristics or substance use history were found between the groups, there may have been unmeasured differences. Second, the relatively high rate of follow-up for buprenorphine may make it more difficult to identify a difference associated with a peer intervention.

Third, peer provider interventions are heterogeneous and thus can have varying results.²⁷ Peer interventions take place in many different settings, from the community to a hospital. Peers can take on many different activities, including care navigation, unstructured conversation, or more structured therapy.³⁹ Strategies that have been shown to be effective with other types of providers could potentially inform what interventions are implemented by peers. For instance, a recent randomized controlled trial found that intensive case management, primarily around social services and medical follow-up, for patients with SUD being discharged from the hospital reduced readmission and increased outpatient treatment engagement.⁴⁰ While social workers led the intensive case management in that study, peers could likely conduct some similar activities. Additionally, peer support could affect a variety of meaningful, patient-centered outcomes. Qualitative and mixed methods studies that included patients who worked with a peer highlight a number of outcomes that are important to patients, including experience of or satisfaction with care,^25,26 trust in medical providers or the healthcare system,^26,41,42 mental health symptoms, and harm reduction.²³ Finally, peer providers can target a range of patients, from those most ready to change to those least engaged with care.^22,27 Peer providers are a promising option for helping patients connect with outpatient care, but the peer intervention examined here needs additional refinement.

Our examination of the activities the peers reported providing is a first step to understanding the content of inhospital peer contacts. Peers were often engaging with patients in multiple domains, most frequently sharing stories (“witness to life”) and helping facilitate medical care. To our knowledge, this is the first study to quantify peer activities, although they have previously been explored qualitatively.²³ Given the small sample, deductively defined categories, single coder, and lack of structured peer note template, we did not analyze whether specific peer activities were associated with receipt of a buprenorphine prescription. However, future studies should build on this to better characterize peer interventions and test whether specific activities are associated with different outcomes.

Limitations

This study has several limitations. First, this is a non-randomized, retrospective cohort study and was not designed to demonstrate causality. Given the paucity of literature on peer provider programs, our findings nevertheless contribute to understanding what outcomes peer provider programs may affect, which can inform future studies. Second, this study was based at a single, academic medical center that had a relatively well-developed and grant-funded addiction medicine program, which may limit the generalizability of our results to other sites. Third, some patients were excluded from the primary analysis because neither PMP nor follow-up visit data were available, but a sensitivity analysis in which these patients were assumed not to have linked yielded the same results as the initial analysis. Fourth, readmission results likely underestimate the number of patients who were readmitted since they included only patients who were readmitted to a hospital within the system. There are other hospitals in the urban center where this study took place, and patients from the large rural catchment area may initially have been admitted to an urban hospital for a complex medical need then returned to their regions of origin. Fifth, this study had a relatively small sample size so results should be interpreted with caution. However, the effect size for the primary outcome (RR = 1.06) does not suggest a clinically significant benefit.

Finally, this study took place prior to all local changes due to the COVID-19 pandemic, raising questions about whether results are generalizable to a post-pandemic setting. Qualitative and conceptual work on peer providers has suggested that peers may be particularly important during the pandemic, as peers can help vulnerable patients, who may be disproportionately marginalized by the social distancing and economic changes of COVID-19, engage with care.^43,44 In our hospital specifically, peers did only phone visits beginning in March 2020 (after the end of this study), but resumed bedside visits in early 2021, once vaccines became available. Accordingly, the content and structure of the peer program changed only briefly due to COVID-19, suggesting that our results are generalizable to the current time. Future studies should examine peer interventions tailored to a post-pandemic world—for instance, a more intensive peer intervention that blends in-person and telehealth visits and targets patients who have limited access to social or health services due to COVID.

The findings have implications for practitioners, researchers, and policymakers. Healthcare providers working with peers should acknowledge that the peer role is heterogeneous and it remains unclear which outcomes peers affect, the intensity of peer intervention needed for effect, and which patient population they are best positioned to help. Policymakers have been funding peer programs nationwide in absence of strong data for their efficacy.^21,27,45 In order to make the most of this investment in OUD treatment, there should be investment in research on peer programs alongside their implementation. Finally, there is great need for investigation into the outcomes that peer providers may affect, core components of peer interventions, and mechanisms through which peers may improve clinical outcomes. Such research should include the perspectives of people with lived experience of opioid use disorder, including both peers and their patients.