Sage Journals: Discover world-class research

Abstract

Background:

Although medication for opioid use disorder (MOUD) is the most effective form of treatment for reducing opioid misuse and related fatalities, programs continue to struggle to reach and engage those with relatively severe disorders and most in need of treatment.

Objectives:

This study assessed whether providing a rapid medical appointment conducted via telehealth could improve MOUD linkage versus the standard model of an in-person referral.

Methods and Design:

We recruited participants from a syringe service prevention program (SSP) and the surrounding communities, randomly assigning them to have an initial medical examination via telemedicine (n = 136) or an in-person (n = 135) referral. We administered structured interviews and collected urine samples for drug testing at baseline and 90 days. The primary outcome was treatment linkage within 14 days of study enrollment, measured by attendance at their first in-person appointment. Secondary outcomes included MOUD engagement (34 days), 3-month retention, and self-reported and urine-detected non-MOUD opioid use at 3-month follow-up.

Findings:

Contrary to expectations, telemedicine participants had lower odds of a successful linkage (aOR=0.42, CI=0.36, 0.49) than control condition participants. A significant interaction between the treatment condition and the timing of the initial treatment appointment indicated that having a treatment appointment scheduled within 48 hours of the medical evaluation had a more pronounced and positive impact than having a telehealth evaluation (aOR=2.9, CI=2.04, 4.12). There were no other significant differences for any secondary outcomes.

Conclusions:

Telemedicine participants were less successfully linked to MOUD than standard referrals, likely owing to experiencing longer delays between their medical evaluation and first in-person treatment appointment compared with those having an in-person medical evaluation. More effective alternatives for coordinating MOUD and SSP services include co-locating MOUD within an SSP or full delivery of MOUD via telemedicine at an SSP, both of which would further obviate delays in treatment initiation.

Trial Registration:

ClinicalTrials.gov (NCT04575324)

Keywords

telemedicine medications for opioid use disorder MOUD clinical trials opioid use disorder treatment linkage

Introduction

The most recent national estimates from the National Survey on Drug Use and Health indicate that in 2023 opioid use disorder (OUD) affected over 5.7 million or 2.0% of U.S. residents ages 12 or older.¹ Until 2022-2023, when the decades-long trend of increasing opioid-related overdose fatalities (OROF) finally abated and reversed, the US had experienced a year-over-year rise in OROF.² Even with the recent 2024 downturn in OROF, the number of fatalities attributable to opioid-related overdoses in the U. S remains substantial.³

Although medications for treating opioid use disorder (MOUD, eg, methadone, buprenorphine, and naltrexone) have proven effective, access to and utilization of MOUD treatment remains limited, with less than 13% of people with an OUD reporting treatment enrollment.⁴ To further reduce the number of OROFs, expanding access to and acceptance of MOUD remains crucial.⁵ A significant barrier faced by MOUD treatment programs – and by substance use treatment programs generally - is engaging those who deeply mistrust formal healthcare.^4,6

One possible way to improve treatment acceptance and initiation is to implement rapid introduction of treatment, as is done in emergency department (ED) settings. An advantage of the ED setting for MOUD treatment initiation is that the required initial medical evaluation is conducted shortly following admission, which is then followed closely by MOUD treatment induction.^7,8 Research has shown ED-based initiation of and linkages to MOUD treatment have successfully engaged otherwise hard-to-reach individuals with an OUD, but typically only after they have experienced a recent overdose.^8-10 A rapid medical evaluation conducted at places where opioid users are already present and receiving other services, such as syringe service programs (SSP), could provide an opportunity similar to that afforded in an ED and overcome the treatment hesitancy and institutional distrust that affects MOUD initiation in non-ED settings. Telemedicine has emerged as a promising tool to improve access to MOUD, especially in response to the COVID epidemic — but its real-world impact, particularly in SSP-based referral models, remains under-researched.^11-13

SSPs provide low-barrier services to reduce drug use risks without requiring participation in a formal substance misuse treatment or healthcare program, creating a potentially more trusted space for people who use opioids (and other drugs) where they feel greater levels of comfort and trust compared to traditional healthcare settings.^14-17 This comfort has been attributed to SSPs’ harm reduction approach, which emphasizes low-barrier services to reduce risks associated with drug use, while offering (but not requiring) connection to formal healthcare.¹⁴ Prior studies have demonstrated that SSP services can improve outcomes related to risky drug injection practices and can enhance benefits when SSP services are combined with MOUD and other treatment modalities.^18-21 However, while SSP clients often express treatment interest,^22-24 the actual treatment initiation ranges from 5% to 70%.^23,25-33

A recent scoping review of studies of SSPs as locations for providing linkages to substance use disorder (SUD) treatment generally and MOUD treatment specifically found that SSP clients could be successfully linked to SUD treatment. Beyond linkages, the review also concluded that SSPs can successfully provide co-located buprenorphine treatment.³⁴ Moreover, other studies have found that SSPs are a good location for reaching persons with relatively more severe SUDs, including OUD, and lower prior treatment participation than other groups of drug users, including injection drug users.²¹ Identifying the best approaches for linking SSP clients to MOUD has potentially far-reaching impacts, considering more than 90 countries representing an estimated 88% of the global population of people who inject drugs have SSPs.^35,36

To test these propositions, we conducted a randomized controlled trial (RCT) designated STAMINA (Syringe Service Telemedicine Access for Medication-assisted Intervention through Navigation) that was designed to offer telemedicine-based medical evaluations for persons receiving services at an SSP. Telemedicine-based MOUD models were implemented in many countries during the COVID-19 pandemic to accommodate social distancing.^37,38 US telemedicine expansion was made possible by the temporary rollback of long-standing policies preventing telemedicine-based buprenorphine prescriptions.³⁹ Since this rollback, observational and pilot studies have demonstrated high patient satisfaction, increased access, comparable retention, and cost reductions.^37,38,40

STAMINA expanded on prior studies that also sought to evaluate MOUD treatment access and uptake by initiating care at an SSP. We implemented an RCT whereby persons could select their preferred MOUD medication in contrast to prior studies, which constrained MOUD treatment to either buprenorphine or methadone, but not both. We enrolled persons with an OUD receiving supportive services at a local SSP or through community outreach and randomly assigned them to receive a medical evaluation immediately via telemedicine or a delayed in-person medical evaluation at the treatment clinic. The primary goal of STAMINA was to test whether having a medical evaluation delivered rapidly via telemedicine in an SSP office would increase MOUD treatment linkage owing to the immediacy of the telemedicine appointment compared with an in-person medical appointment, typically scheduled several days after referral. A secondary goal was to assess whether having a telemedicine examination would also increase MOUD engagement and retention subsequent to treatment initiation, as well as reduce non-prescribed opioid use.

Methods

Chestnut Health System’s Institutional Review Board (IRB) approved all procedures, and a reliance agreement was used to establish a single IRB process with the University of Illinois Chicago.

Setting and Participants

We recruited 274 participants from 2 Chicago SSP sites and their surrounding communities, with one SSP located in a predominantly African American community on Chicago’s west side, and the other in a predominantly Latino community on the northwest side. Table 1 provides the study’s eligibility criteria. SSP clients were recruited to the study through program staff, self-referral after seeing recruitment materials posted in the SSP offices, or by outreach workers who recruited additional participants from the surrounding neighborhoods. Our recruitment target of 274 was determined by an a priori power analysis as adequate for detecting an odds ratio of at least 2.0 for a dichotomous predictor (the intervention effect) in a binary logistic regression with 80% power at a p < .05 significance level.⁴¹ However, 3 participants began the study before the block randomization by preferred MOUD was implemented (see procedures below). As this variable was included in all regression models, we could not retain these non-randomized participants in the analyses, reducing our final analytic N to 271.

Table 1.

Study Inclusion and Exclusion Criteria.

Inclusion	Exclusion
18+ years of age	Experiencing severe opioid withdrawal symptoms indicating need for immediate referral to medical care^a
Communicated verbally in English	Already receiving medication for opioid use disorder treatment.
Met DSM-5 clinical criteria for a past year opioid use disorder of any severity level ⁴⁵	Planned to move outside of Cook County, IL within the next 6 months.
Expressed interest in receiving any medication for opioid use disorder treatment	Required to serve a prison sentence within 6 months of enrollment
Resides in Cook County, IL	Demonstrated inadequate ability to provide informed consent as indicated by inability to answer 3 clarifying questions after receiving an explanation of the study.

Assessed using the Clinical Opiate Withdrawal Scale.

Interventions

STAMINA comprised the following components: (a) a brief interview to assess MOUD interest; (b) a same-day secure, video-based telemedicine appointment with a provider who can prescribe or refer to all 3 MOUD types; (c) one-time rideshare-based transportation assistance to either pick up a prescription for buprenorphine or naltrexone induction or to receive a first methadone dose; (d) instructions for how to locate the pharmacy on-site at the FQHC and how to complete induction steps. An FQHC partner provided MOUD treatment assessment and referral for this trial. The collaborating FQHC had an on-site pharmacy that dispensed buprenorphine and naltrexone and offered daily on-site methadone dosing.

Participants assigned to the usual referral arm who were interested in buprenorphine treatment were given a standard FQHC referral for an in-person intake appointment. Patients interested in methadone treatment were referred to a provider who was accessible to them. Staff attempted to schedule all referral appointments within 2 business days. Usual referral arm participants were given 2 bus passes to assist with transportation to their first appointment.

Measures

Primary Outcome

Our primary outcome was a binary (yes/no) indicator of a successful in-person treatment linkage, defined following national care standards as attending one treatment appointment within 14 days of enrollment and verified by the electronic health record.⁴² We selected ‘in-person’ linkage over ‘medication-based’ (ie, receipt of an initial medication dose) as the primary outcome because the collaborating FQHC’s telemedicine policies mandated a post-induction in-person appointment for continuing on buprenorphine. This ensured our linkage definition reflected the need for all participants to complete a first in-person treatment appointment following their medical evaluation.

Secondary Outcomes

We assessed 4 binary (yes/no) secondary outcomes: (1) MOUD treatment engagement, defined as attending a minimum of 2 OUD-related treatment appointments within 34 days of enrollment;⁴² (2) MOUD treatment retention, defined as not exceeding 14 consecutive days without medication during the 90-day observation window; (3) any self-reported past 30-day non-MOUD opioid use assessed at follow-up using questions derived from the National Survey on Drug Use and Health Questionnaire;⁴³ and (4) any use of non-MOUD opioids such as fentanyl, detected by urinalysis at follow-up. We collected treatment data from FQHC’s medical records and the city’s largest methadone provider. We collected self-reported treatment linkage and engagement data to understand whether participants engaged in treatment with other providers. The Cook County Medical Examiner provided mortality data.

Covariates

We evaluated 15 covariates for inclusion in the logistic regression models. Demographic measures captured in interviews included: age in years (continuous), gender (male/female), race (African American/other), ethnicity (Latino/other), employment status (employed/unemployed), insurance status (insured/uninsured), and housing status (stable/unstable). We identified whether the participant was an SSP client (yes/no) by asking if they had ever used services at either site.

Additional covariates included self-reported number of lifetime overdoses, recruitment location (SSP vs community), and severe psychological distress (yes/no) as assessed at baseline using the Kessler Psychological Distress Scale (K6) and the recommended threshold of scores greater than 12 indicating severe distress.⁴⁴ Opioid use severity was assessed as a continuous measure yielding scores from 0/lowest to 11/highest using a DSM-5 OUD screening instrument.⁴⁵ Withdrawal severity was assessed as a continuous measure from 0–48 using the COWS.^45,46 Variables related to study involvement included: enrollment arm (telehealth/in-person medical exam); whether the participant had an in-person appointment scheduled within 48 hours of study enrollment (yes/no); block randomization (methadone only/open to any MOUD). We used enrollment location (west or northwest Chicago community) as a clustering variable in the multivariable models.

Procedures

Study recruitment ran from August 24th, 2020, through June 30th, 2022. Trial participants were identified in 2 ways: (1) onsite at the SSP program and (2) street outreach activities within the geographic service areas of the SSP sites. Research staff conducted an eligibility screener and walked participants through a treatment decision guide explaining the 3 available MOUDs. We determined arm allocation using randomization with stratification by site and whether the participant indicated a strong preference for methadone treatment. This was because the FQHC referred patients to a different methadone provider per patient preference if it was geographically easier to access for daily dosing. We determined allocation in advance of enrollment by randomizing pre-established study identification numbers sequentially assigned to participants as they were enrolled. The arm assignment was revealed to the on-site research assistant by phone after the data were collected. This was conducted as an open trial, as we made no efforts to conceal the assignment after baseline data collection. Figure 1 provides an overview of the main study steps from enrollment through analysis showing the study N at each step and conforms to CONSORT reporting guidelines for randomized trials.

Figure 1.

STAMINA trial CONSORT diagram.

Research assistants collected baseline data using a structured interview questionnaire and administered a point-of-care drug screening test. All participants were asked to complete these procedures again 3 months after enrollment. We were able to locate and reinterview 217 (80.0%) participants at follow-up, with 191 interviews conducted in person and 26 via telephone for those who could not return to the SSP. Participants received a $50 incentive for completing baseline data collection and $35 for the shorter follow-up. With participant consent, administrative data reflecting 6 months of post-enrollment treatment were extracted from medical records. The county medical examiner provided mortality data using a deterministic linkage algorithm based on patient name, demographics, and social security number. Research staff recorded all interviews, urinalysis results, medical records data, and data from the coroner’s office electronically using the UIC installation of the Research Electronic Data Capture System (REDCap), a secure web-based HIPAA-compliant application for capturing health-related data.⁴⁷

These procedures include protocol modifications to address barriers stemming from the COVID-19 pandemic, which shifted some study aspects toward a more pragmatic approach. Due to considerable reductions in client volume experienced by our SSP and others nationally,⁴⁸ we decided not to recruit at a third planned site with the lowest volume of referrals. Instead, we increased the number of recruitment days at the 2 higher volume locations. We also added outreach-based recruitment to address the early impact of low enrollments on our ability to meet the planned sample size and increased baseline interview incentives from $25 to $50 to further improve the recruitment rate. Initially, there was a planned 6-month interview, which we eliminated to accommodate time and space considerations related to social distancing.

Our FQHC partner was also scheduled initially to implement methadone dosing at multiple other locations near the recruitment sites, which was delayed considerably due to the pandemic’s impact on the federal approval process. As sending participants to inaccessible treatment locations was not ethical, we referred them to alternative methadone providers with whom we did not always have an established relationship to obtain treatment records. This resulted in more missing data related to treatment linkage than we anticipated, for which we adopted a conservative approach, assuming no linkage where data were missing, as described below.

Analyses

Our reporting of study findings follows CONSORT guidelines.⁴⁹ We conducted an intention-to-treat analysis with all statistical modeling done using Stata v18.5.⁵⁰ We used generalized estimating equation (GEE) binary logistic regression for all analyses with SSP interview site as the clustering variable. For each outcome, the GEE models provide population-averaged fixed-effect estimates for included covariates.⁵¹ We estimated all models assuming an exchangeable correlation structure, robust standard errors, and Stata’s ‘nmp’ method call to correct for the small number of clusters.^52,53

Primary Outcome

We included 2 covariates in the initial analysis: treatment condition and MOUD randomization block to test the main study hypothesis. Of the 271 participants, treatment linkage was definitively determined for 209 (77.1%). For the remaining 62 (22.9%) participants, there was no administrative record of the participant showing for an in-person treatment appointment within 14 days of the medical examination. This was either because the participant never showed up for their scheduled in-person treatment appointment, or an appointment was made with a provider outside the study network whose treatment records were not accessible to the study. We conservatively assumed no treatment linkage occurred for those missing any record of an initial in-person treatment appointment and assigned a value of zero (no appointment). However, we conducted a sensitivity analysis to test this assumption using complete case analysis as described below.

Following estimation of the regression model that tested our main hypotheses, we conducted further binary logistic regressions that retained treatment condition and randomization block as model predictors but added covariates. Although randomization was stratified by site and MOUD preference, additional covariates were included in secondary and sensitivity analyses to explore their contribution to outcomes and improve model precision. We conducted preliminary bivariate analyses to screen additional covariates for inclusion in a multivariable GEE logistic model, selecting covariates significant at p <= .20 or lower for each outcome’s bivariate model. This approach, while sequential, did not involve automatic stepwise variable selection procedures. We then used the reduced covariate set in a preliminary multivariable model, followed by a final GEE logistic model that excluded non-significant covariates in the preliminary multivariable model. Preliminary analyses showed a strong association between scheduling an in-person appointment within 48 hours and treatment linkage, as well as with arm assignment. This supported the inclusion of an interaction term between these variables in the multivariable model.

Secondary Outcomes

We applied the same modeling strategy used for the primary outcome, including covariate selection procedures and model structure, to each of the secondary outcomes, using the appropriate dependent variable in each case. For each secondary outcome, we ran a model with predictors for treatment assignment and MOUD preference randomization block. We then ran a series of models to assess the effects of additional covariates, with the final model for each secondary outcome including treatment assignment, MOUD block, and any covariates significant in preliminary bivariable regression models. To facilitate cross-model comparison, we conducted parallel analyses, including all covariates that were significant in at least one model, and present these in a supplemental table that shows the same set of covariates for each model, regardless of significance in any given model. The results of these parallel analyses are provided in Supplemental Table 1. For treatment engagement and retention, we made the same assumption as for treatment linkage (ie, that the participant was not engaged or retained in the absence of an indication they had ever begun in-person treatment). We also tested this assumption by conducting a complete case analysis for these 2 outcomes as described in the next section.

Sensitivity Analysis

Although only 3 covariates had a small amount of missing data (insurance status [N = 10, 3.7%], opioid use severity [N = 3, 1.1%], and withdrawal severity [N = 5, 1.8%]), our primary outcome, treatment linkage, was missing/indeterminate for 22.8% (N = 62) of participants. To assess the effect of including or excluding participants with an indeterminate linkage in the primary outcomes regression models, we conducted a sensitivity analysis by running 2 models for this outcome. For the primary outcome (treatment linkage), the main analysis conservatively assumed that missing data indicated no linkage (coded as 0). A sensitivity analysis excluded these cases from the model. As the 2 models converged substantively regarding the effects of treatment condition and MOUD preference on the estimated probability of being successfully linked to treatment, we emphasize the results for the main model, where an indeterminate linkage status was assumed to mean no linkage.

We assessed whether the data met criteria for the missing completely at random (MCAR) assumption using Little’s chi-square test, which returned a non-significant result, suggesting that the assumption held. Separate bivariate tests of missingness by covariate revealed only isolated, marginal associations, which we judged insufficient to warrant multiple imputation. Because no strong predictors of missingness emerged and auxiliary variables were weak or unavailable, we opted for a complete case analysis under the MCAR assumption. We acknowledge this analytic choice as a potential limitation and addressed it further in the Discussion section.

Results

Descriptive Statistics

Sociodemographics and baseline opioid use disorder characteristics. Table 2 provides descriptive statistics for sociodemographics and opioid use disorder characteristics by treatment condition. There were no statistically significant differences by treatment condition for any baseline measures, supporting successful allocation balance. Most participants were recruited via outreach (67.2%) versus through being an existing SSP client (32.8%). A majority of participants in both treatment conditions (54.2% overall) expressed a preference for methadone treatment. Most participants were male (78.2%), African American (61.6%), non-Latino (79.0%), unemployed (74.2%), had health insurance (86.2%), and were stably housed (55.0%). The average age at recruitment was 48.2 years (SD = 10.4). Just over one-third of participants (33.9%) evidenced symptoms consistent with severe psychological distress and, per self-report, had experienced an average of over 4 lifetime opioid-related overdoses (m = 4.4, SD = 10.1). On average, participants evidenced only mild withdrawal symptoms (m = 4.5, SD = 4.5) but met criteria for a severe OUD (m = 10.4, SD = 1.1).

Table 2.

Participant Baseline Demographics and Characteristics by Study Arm.

	In-person	Telehealth	Total	Sig
	(N = 135)	(N = 136)	(N = 271)
Recruitment location
Community outreach	90 (66.7%)	92 (67.6%)	182 (67.2%)	0.864
SSP client	45 (33.3%)	44 (32.4%)	89 (32.8%)
MOUD preference
Any medication^a	60 (44.4%)	64 (47.1%)	124 (45.8%)	0.666
Methadone^b	75 (55.6%)	72 (52.9%)	147 (54.2%)
Gender at birth
Male	107 (79.3%)	105 (77.2%)	212 (78.2%)	0.682
Female	28 (20.7%)	31 (22.8%)	59 (21.8%)
Race
Black/African American	83 (61.5%)	84 (61.8%)	167 (61.6%)	0.962
Other	52 (38.5%)	52 (38.2%)	104 (38.4%)
Ethnicity
Non-Hispanic/Latino	105 (77.8%)	109 (80.1%)	214 (79.0%)	0.632
Hispanic/Latino	30 (22.2%)	27 (19.9%)	57 (21.0%)
Age (Mean [SD])	48.81 [10.55]	47.53 [10.15]	48.17 [10.35]	0.308
Employment status
Unemployed	96 (71.1%)	105 (77.2%)	201 (74.2%)	0.252
Other	39 (28.9%)	31 (22.8%)	70 (25.8%)
Insurance Status
No insurance	18 (13.8%)	18 (13.7%)	36 (13.8%)	0.980
Insurance	112 (86.2%)	113 (86.3%)	225 (86.2%)
Homelessness status
Not homeless	75 (55.6%)	74 (54.4%)	149 (55.0%)	0.850
Homeless	60 (44.4%)	62 (45.6%)	122 (45.0%)
Severe Psychological Distress
Absent	85 (63.0%)	94 (69.1%)	179 (66.1%)	0.285
Present	50 (37.0%)	42 (30.9%)	92 (33.9%)
Lifetime Opioid Overdoses (Mean [SD])	3.83 [9.24]	4.90 [10.93]	4.37 [10.12]	0.383
Clinical Opioid Withdrawal Scale Score (Mean [SD])	4.00 [3.74]	5.00 [5.16]	4.50 [4.53]	0.070
Opioid Use Severity (Mean [SD])	10.41 [1.05]	10.47 [1.20]	10.44 [1.12]	0.660

All figures shown are based on information collected at the baseline interview and, unless otherwise indicated, reflect N’s and percents. Significance levels are based on chi-square tests for categorical variables and on t-tests for interval level measures.

Prior to randomization, the participant indicated their willingness to discuss all medication options with a provider.

Participant indicated prior to randomization they had a strong preference for methadone treatment.

Appointment scheduling, primary, and secondary outcomes

As shown in Table 3, 2 factors differed significantly by treatment condition; participants in the telemedicine condition (27.2%) were less likely to have an in-person treatment appointment scheduled within 48 hours of enrollment compared with controls (54.8%), ( $χ^{2}$ ₍₁₎ = 21.3, P < .001) Additionally, fewer participants in the telemedicine condition (17.6%) completed their initial in-person treatment appointment within 14 days compared with control condition participants (34.1%). Telehealth participants also had more indeterminate linkages (31.6% vs 14.1%) ( $χ^{2}$ ₍₂₎ = 16.2, P < .001). Otherwise, the 2 treatment groups did not differ in the proportions engaging or retained in treatment, or in the proportion detected as using non-MOUD opioids and self-reported opioid use at follow-up.

Table 3.

Appointment Scheduling and Primary and Secondary Outcomes by Study Arm.

	In-person	Telehealth	Total	Sig
	(N = 135)	(N = 136)	(N = 271)
First in-person treatment appointment within 48 hours
Longer than 48 hours	61 (45.2%)	99 (72.8%)	160 (59.0%)	<0.001
Within 48 hours	74 (54.8%)	37 (27.2%)	111 (41.0%)
Linkage
Not linked	70 (51.8%)	69 (50.8%)	139 (51.3%)	<0.001
Linked	46 (34.1%)	24 (17.6%)	70 (25.8%)
Indeterminate^a	19 (14.1%)	43 (31.6%)	62 (22.8%)
Engagement
Not engaged	106 (78.5%)	117 (86.0%)	223 (82.3%)	0.105
Engaged	29 (21.5%)	19 (14.0%)	48 (17.7%)
Retention
Not retained	121 (89.6%)	124 (91.2%)	245 (90.4%)	0.666
Retained	14 (10.4%)	12 (8.8%)	26 (9.6%)
Urinalysis-detected opioid use^b
No non-moud opioids detected	20 (26.3%)	21 (28.8%)	41 (27.5%)	0.738
Any non-moud opioids detected	56 (73.7%)	52 (71.2%)	108 (72.5%)
Self-reported opioid use^c
No self-report opioid use	15 (13.9%)	17 (16.0%)	32 (15.0%)	0.659
Self-report opioid use	93 (86.1%)	89 (84.0%)	182 (85.0%)

All data shown are N’s and percents and are based on data collected from administrative records extraction at follow-up with 2 exceptions: self-reported opioid use, which was obtained during the follow-up interview and urinalysis-detected opioid use, which is based on urine samples also collected and tested at follow-up.

An indeterminate linkage was due to there being no indication of a scheduled initial in-person appointment in the administrative records of the collaborating MOUD provider.

122 participants had missing data on urinalysis results (n = 149).

57 participants had missing data on self-reported opioid use (n = 214).

Multivariable GEE Models

Sensitivity Analysis

The sensitivity model, which treated missing/indeterminate data for the primary outcome (treatment linkage) as missing, yielded similar patterns but differences in the magnitude of effects from the main analysis, which assumed missing data indicated no linkage. We conducted parallel sensitivity analyses for the engagement and retention outcomes as well, which also relied on EHR-based confirmation. These sensitivity models produced results that were directionally consistent with the main models but showed differences in estimated odds ratios. We note these differences below, where relevant, and have provided 2 additional supplemental tables: Supplemental Tables 2 and 3 show the sensitivity results using the alternate missing data assumption (complete case analysis) and correspond to the main models shown in Tables 4 and 5.

Table 4.

GEE Binary Logistic Regression Outcomes – Restricted Covariate Set.

	Linkage		Engagement		Retention		Urinalysis-Detected Opioid Use^a		Self-Reported Opioid Use^b
	Odds Ratio[95% CI]	Sig	Odds Ratio[95% CI]	Sig	Odds Ratio[95% CI]	Sig	Odds Ratio[95% CI]	Sig	Odds Ratio[95% CI]	Sig
Treatment condition
Telehealth	0.42	***	0.60	NS	0.86	NS	0.88	NS	0.84	NS
	[0.36, 0.49]		[0.31, 1.16]		[0.72, 1.03]		[0.47, 1.66]		[0.46, 1.56]
MOUD preference
Methadone	1.45	**	1.50	***	3.95	***	1.69	**	0.96	NS
	[1.12, 1.87]		[1.24, 1.81]		[2.00, 7.81]		[1.19, 2.39]		[0.67, 1.36]
Intercept	0.43	**	0.22	***	0.05	***	2.07	NS	6.36	**
	[0.23, 0.82]		[0.12, 0.39]		[0.04, 0.05]		[0.66, 6.52]		[2.06, 19.59]
N	271		271		271		149		214

Parameter estimates were obtained using generalized estimating equations for bivariate logistic regression with study site as the clustering variable. The correlational structure was assumed to be independent with variance estimation using robust standard errors and the ‘nmb’ correction for small cluster size. Treatment linkage was defined as attending one treatment appointment within 14 days of enrollment and verified by the electronic health record. The linkage, engagement, and retention models assume that participants with indeterminate chart data (ie, missing linkage, engagement and retention statuses) did not attend treatment. Sensitivity analyses treating these cases as missing are described in the text with corresponding results provided as Supplemental Table 2.

There were 122 cases with missing data for urinalysis-detected opioid use owing to no follow-up interview conducted, urinalysis collected, or the interview was conducted on the telephone.

There were 57 cases with missing data for self-reported opioid use owing to no follow-up interview conducted.

NS = non-significant.

***

P < .001, ** P < .01.

Table 5.

GEE Binary Logistic Regression Outcomes – Extended Covariate Set.

	Linkage		Engagement		Retention		Urinalysis-Detected Opioid Use^a		Self-Reported Opioid Use^b
	Odds Ratio[95% CI]	Sig	Odds Ratio[95% CI]	Sig	Odds Ratio[95% CI]	Sig	Odds Ratio[95% CI]	Sig	Odds Ratio[95% CI]	Sig
Treatment condition
Telehealth	0.37	***	0.64	NS	0.81	NS	0.84	NS	0.88	NS
	[0.30, 0.45]		[0.27,1.54]		[0.35, 1.86]		[0.44, 1.61]		[0.49, 1.57]
MOUD preference
Methadone	1.25	**	1.28	NS	3.73	NS	1.97	***	1.01	NS
	[1.06, 1.47]		[0.93, 1.77]		[0.83, 16.70]		[1.38, 2.81]		[0.64, 1.60]
Recruitment location
SSP client	2.06	**	2.19	**	—		1.70	***	1.98	*
	[1.19, 3.55]		[1.29, 3.73]				[1.35, 2.14]		[1.11, 3.50]
Gender at birth
Female	1.52	***	0.89	***	2.10	***	—		0.70	***
	[1.27, 1.82]		[0.86, 0.92]		[1.87, 2.36]				[0.65, 0.75]
Ethnicity
Hispanic/Latino	—		0.55	***	1.33	***	4.18	***	—
			[0.43, 0.69]		[1.15, 1.54]		[1.94, 9.00]
Age	1.04	**	1.04	***	1.09	***	1.04	***	—
	[1.01, 1.06]		[1.04, 1.05]		[1.07, 1.11]		[1.04, 1.05]
Employment status
Other	—		—		—		0.47	**	—
							[0.29, 0.77]
Health insurance status
Insured	1.71	***	—		—		—		—
	[1.27, 2.30]
Severe psychological distress
Present	—		—		—		0.68	**	—
							[0.51, 0.90]
Lifetime opioid overdoses	—		—		—		0.97	***	0.99	**
							[0.96, 0.97]		[0.98, 0.99]
Clinical opioid withdrawal scale score	—		—		—		0.91	***	—
							[0.89, 0.94]
First treatment appointment within 48 hours
Within 48 hours	2.32	NS	2.77	***	2.38	***	0.43	***	—
	[0.82, 6.57]		[1.77, 4.31]		[2.09, 2.72]		[0.33, 0.57]
First treatment appointment within 48 hours # Telehealth	2.20	**	1.57		1.85	NS	—		—
	[1.23, 3.93]		[0.67, 3.68]		[0.67, 5.10]
Intercept	.02	*0	0.01	***	0.00	***	0.63	NS	5.89	*
	[0.00, 0.50]		[0.01, 0.02]		[0.00, 0.00]		[0.32, 1.22]		[1.52, 22.87]
N	260		270		270		148		214

Parameter estimates were obtained using generalized estimating equations for bivariate logistic regression with study site as the clustering variable. The correlational structure was assumed to be independent with variance estimation using robust standard errors and the ‘nmb’ correction for small cluster size. Cells with a dash (—) indicate the corresponding variable was not significant in a preliminary multivariable model and was dropped from the final model represented in this table. Treatment condition and randomization block were retained in all models regardless of significance. The linkage, engagement, and retention models assume that participants with indeterminate chart data (ie, missing linkage, engagement and retention statuses) did not attend treatment. Sensitivity analyses treating these cases as missing are described in the text with corresponding results provided as Supplemental Table 3.

There were 122 cases with missing data for urinalysis-detected opioid use owing to no follow-up interview conducted, urinalysis collected, or the interview was conducted on the telephone.

There were 57 cases with missing data for self-reported opioid use owing to no follow-up interview conducted.

NS = non-significant.

***

P < .001, ** P < .01, * P < .05.

Two of the other secondary outcomes collected at follow-up, urinalysis-detected opioid use (N = 122, 45.0%) and self-reported opioid use (N = 57, 21.0%), also had large amounts of missing data. For self-reported opioid use, this was owing exclusively to the lack of a follow-up interview. For urinalysis-detected opioid use, data were missing due to a combination of no follow-up interview, refusal to provide a urine sample at follow-up, or the follow-up interview was conducted by telephone.

Primary and Secondary Outcomes, Restricted Covariate Set

Results for the GEE regressions with trial arm and randomization block as the sole covariates are shown in Table 4. For the primary outcome, telemedicine participants had significantly lower odds of treatment linkage (aOR = 0.42, CI = 0.36, 0.49) than participants in the control condition. Conversely, participants expressing a preference for methadone at enrollment had a higher odds (aOR = 1.45, CI = 1.12, 1.87) of a successful treatment linkage than those expressing no preference. Re-running the model for treatment linkage using complete case analysis (Supplemental Table 2) and excluding participants with no available treatment records did not change the substantive findings and, in fact, magnified the odds that methadone preference was associated with increased odds of treatment linkage (aOR = 3.82, CI = 1.79, 8.16). At the same time, persons in the telehealth condition continued to have lower odds of linking to treatment (aOR = 0.54, CI = 0.33, 0.91).

With one exception, the results for medication preference were consistent across secondary treatment outcomes; participants who preferred methadone had higher odds of both treatment engagement (aOR = 1.50, CI = 1.24, 1.81) and retention (aOR = 3.95, CI = 2.00, 7.81). When we reran the models for these 2 outcomes, and did not assume missing data necessarily meant a missing appointment, the results were similar to those of the sensitivity analyses of treatment linkage (Supplemental Table 2). The effects of medication preference were significant and in the same direction as for the models inferring no treatment, but magnified for both engagement (aOR = 2.43, CI = 1.87, 3.16) and retention (aOR = 3.95, CI = 2.00, 7.81). Participants who expressed a preference for methadone also had higher odds of continued opioid use at the 3-month follow-up as verified by urinalysis (aOR = 6.27, CI = 3.35, 11.76) but not self-report. Treatment condition was non-significant across all secondary outcomes in the models where we inferred that no chart information equated to no MOUD treatment received.

The sensitivity model differed from the main model only with respect to the treatment condition predictor and retention as the outcome. In the analysis using inferred non-retention, which assumed missing data indicated no retention, treatment condition was non-significant (aAOR = 0.86, CI = 0.72, 1.03). Allowing the cases with missing chart information to be missing yielded a significant and positive effect for participants in the telehealth condition relative to those in the in-person medical visit condition (aAOR = 1.22, CI = 1.11, 1.32).

Primary and Secondary Outcomes, Extended Covariate Set

Adjusting for additional covariates yielded differences (Table 5) from the regression models limited to treatment condition and randomization block as the sole covariates. Expressing a preference for methadone was significant for the primary and all secondary outcomes in the more restricted models. However, while it remained statistically significant for the primary outcome, treatment linkage (aOR = 1.45, CI = 1.12, 1.87), it remained significant for only a single secondary outcome, urinalysis-detected opioid use (aOR = 1.97, CI = 1.38, 2.81). Participants expressing a preference for methadone at baseline were more likely to be successfully linked to treatment, but also more likely to be using non-MOUD opioids at the 3-month follow-up. The positive effects of preferring methadone on being linked to treatment did not persist over time through engagement and retention.

Scheduling a first in-person treatment appointment within 48 hours of referral was a strong positive predictor across all outcomes, from linkage to retention. Participants with such an appointment had over 3 times the odds of being linked to care compared to those without one (aOR = 3.09, 95% CI: 1.49–6.38, p = .002), were more likely to be engaged (aOR = 2.77, 95% CI: 1.77–4.31), retained in treatment (aOR = 2.40, 95% CI: 2.07–2.80), and less likely to have a urine test positive for non-MOUD opioids (aOR = 0.43, 95% CI: 0.33–0.57).

This scheduling variable also interacted with treatment condition to predict linkage (aOR = 2.20, 95% CI: 1.23–3.93). As plotted in Figure 2, the interaction effect was strongest for participants in the telemedicine condition, though early appointments facilitated linkage in both groups.

Figure 2.

Estimated Marginal Mean Probabilities for Treatment Linkage by Appointment Scheduling and Treatment Condition.

Participants who were already clients of the SSP similarly had higher odds of being linked (aOR = 2.06, 95% CI: 1.19–3.55) and engaged in treatment (aOR = 2.19, 95% CI: 1.29–3.73), though not retained. SSP clients also had higher odds of urine-confirmed opioid use (aOR = 1.70, 95% CI: 1.35–2.14) and self-reported opioid use at follow-up (aOR = 1.98, 95% CI: 1.11–3.50).”

Among most of the other covariates assessed, effects varied inconsistently across outcomes. An exception worth noting is that older age was consistently associated with higher odds of treatment linkage (aOR = 1.04, CI = 1.01, 1.06), engagement (aOR = 1.04, CI = 1.04, 1.05), and retention (aOR = 1.09, CI = 1.07, 1.10). Older participants also had higher odds of non-MOUD opioid use detected through urinalysis (aOR = 1.04, CI = 1.04, 1.05).

Discussion

Contrary to our main hypothesis, telemedicine was less effective than a standard in-person referral at linking persons to MOUD treatment. Several reasons could have contributed to this result. One factor was the FQHC’s policy to delay the first in-person treatment appointment for those beginning on buprenorphine until post-induction, contrasting with control participants whose appointment was scheduled within 2 business days. Given that the FQHC stopped dispensing methadone after 2 pm, telehealth methadone participants who completed the urine screen and structured interview questionnaire in the afternoon were unable to receive their first dose of methadone until the following day, preventing the rideshare option for transportation assistance, which was only available same-day, and which also possibly contributed to a higher rate of missed appointments. Unfamiliarity with the FQHC compared to more established methadone providers may have also played a role. Several participants indicated previous experience with established methadone clinics, to which research medical staff were more likely to refer control participants.

While methadone treatment was associated with higher odds of verified linkage, it is important to emphasize that some participants referred to methadone were sent to external programs outside our data-sharing agreements. For these individuals, as noted, we could not verify treatment attendance, and they were conservatively coded as non-linked in the primary analysis. However, this subset represented a minority of the methadone-assigned participants. The majority were still referred within the FQHC system, and we were able to obtain treatment records for those individuals. Additionally, methadone is often administered in tightly structured programs with high contact and frequent dosing, which may have improved the likelihood of verified linkage for those treated in-network.

Additional results support the notion that delayed in-person appointments, particularly for buprenorphine patients, negatively impacted outcomes, as 48-hour in-person appointment scheduling had a more significant impact on linkage success than arm assignment. This finding aligns with previous studies indicating shorter appointment wait times improve linkag.^54,55 The significant interaction term in our model indicates that telemedicine participants benefited more from 48-hour scheduling, given that this led to a higher odds of treatment linkage within 14 days for this group. Moreover, the positive effect of 48-hour scheduling seems to have translated into better treatment linkage, engagement, and retention outcomes. Despite a lower linkage rate, the telemedicine arm had higher engagement at 34 days compared with controls (57% vs 44%), suggesting either more positive post-linkage treatment experiences or that the telemedicine participants who did make it to their first in-person appointment could have been more highly motivated by virtue of the fact they had to wait longer for the appointment but still managed to persevere.

These results underscore the importance of rapid treatment scheduling in improving initiation, engagement, and retention. Ironically, our intent to achieve rapid treatment initiation using telehealth resulted in the opposite occurring for participants assigned to this condition, owing to unforeseen logistical hurdles and provider treatment initiation policies. These results, however, do support the original hypothesis that treatment initiation rates can be improved by speeding up the enrollment process in this hard-to-reach and engage population. In our opinion, it is possible that if the logistic and administrative hurdles participants in the telehealth condition encountered disproportionately in STAMINA could be overcome, a rapid telehealth medical appointment followed by an equally rapid induction or medication disbursement appointment would have resulted in improved initiation rates. While the significance of treatment effects diminished in some models after adjusting for additional covariates, the consistency in the direction and magnitude of the coefficients across models suggests the underlying associations remained stable, even if statistical power was reduced due to the modest sample size.

The study findings also support our notion that initiating treatment in an SSP increases the chances of successful MOUD treatment linkage and engagement. This was demonstrated by the higher odds of these 2 outcomes for SSP clients compared with persons recruited from the surrounding communities who did not have an existing relationship with the SSP staff or location. Given that participants recruited via outreach were not as familiar with the SSP sites and services provided, this could have been a barrier to overcoming treatment hesitancy and institutional distrust, thus reducing the impact of telemedicine treatment linkage, engagement, and retention. As noted, we intended to recruit all participants from the SSP program, but had to adjust our recruitment strategy owing to the effects of the COVID pandemic on SSP operations and client attendance. However, these results warrant additional study of SSPs as viable settings for MOUD treatment recruitment and initiation.

Prior studies have reported negative buprenorphine experiences, such as an unpleasant taste and precipitated withdrawal symptoms.^56,57 Home induction protocols, like the one in this study, primarily designed for heroin use, may heighten the risk of precipitated withdrawal for individuals with greater opioid tolerance due to fentanyl use.⁵⁸ The prior non-randomized studies of telemedicine buprenorphine linkage in SSP settings have focused on buprenorphine and included components likely to translate into better treatment outcomes, such as not requiring an in-person follow-up for continued prescribing and patient navigation.^54,55 This is compatible with US emergency rules that allowed pandemic telemedicine buprenorphine prescribing. On April 4, 2024, these same prescribing flexibility levels were permanently extended for opioid treatment programs.⁵⁹ As such, additional policies to support navigation and support services might have mitigated barriers encountered after the initial STAMINA telemedicine engagement.⁶⁰ Based on other prior research, these services can likely be successfully delivered virtually and require no or minimal in-person appointments.^54,55

Also aligning with previous research,⁶¹ our findings suggest participants who preferred methadone and were assigned to a methadone treatment program had a higher odds of MOUD treatment linkage. This is noteworthy considering the many patient complaints about restrictive methadone treatment policies represented in prior literature.^62,63 As such, this positive outcome may be partially related to emergency pandemic regulations that made methadone more flexible for some patients under certain conditions.⁶⁴ In contrast, the FQHC’s complex buprenorphine induction follow-up requirements might have confused treatment expectations and course.

Limitations

This study’s pragmatic approach limited our control over intervention delivery and the treatment protocols of our FQHC partner, potentially affecting internal validity. In the broader context, the study was conducted during the COVID-19 pandemic, which disrupted service delivery, altered participant behavior, and contributed to challenges in treatment access and data collection. In the narrower context, one policy at the FQHC site delayed first in-person appointments for telemedicine-initiated buprenorphine patients until after medication induction, whereas in-person control participants were typically scheduled within 2 business days. While this reflects real-world implementation barriers at the time, it may have introduced structural differences in treatment access and limited generalizability to settings with more streamlined intake procedures.

As a counterpoint, operating under real-world linkage and treatment conditions likely improved generalizability. Other site-specific factors, such as on-site methadone treatment, may not be replicable elsewhere. Additionally, providing participants with a choice among all 3 MOUD options introduced heterogeneity that complicated analyses but better reflected shared decision-making models emerging in SSP-based telemedicine programs.

Although prescriptions were written for buprenorphine in the telemedicine arm before the required in-person intake, full dispensation data were unavailable, preventing deeper analysis of medication initiation. We were also unable to systematically capture prescription duration or refill timing due to EHR limitations, which restricted our ability to assess how initial prescribing practices may have influenced linkage or continuation.

While telemedicine programs now increasingly support fully virtual treatment initiation, this was not the case during the STAMINA trial period. At the time, in-person attendance remained the standard at both study sites. As such, our primary outcome—attendance at an in-person appointment—was a valid and meaningful measure of treatment linkage within the care structures that existed. Moreover, although engagement and retention outcomes could theoretically include both in-person and telehealth visits, the EHRs we reviewed did not differentiate between visit types/locations. Based on communication with site partners, these secondary and tertiary outcomes are almost certainly reflected in in-person treatment and the state of telehealth implementation during the study period, which had not yet fully extended to post-initiation care. However, as fully virtual models of care continue to evolve, outcomes such as continuous buprenorphine access and medication coverage may become more relevant indicators of treatment success.

We noted above the data limitations concerning medical records abstraction in assessing treatment attendance from linkage through retention. Because of delays in the FQHC obtaining licenses to dispense methadone at multiple locations, we had to refer some study participants to treatment programs with which we did not have participant consent to access their medical records. We opted for a conservative intent-to-treat strategy and assumed all such patients for whom we did not have a treatment attendance record did not attend any appointments. Finally, we also noted the large amount of missing data for the urinalysis-detected and self-reported opioid use at follow-up. Consequently, the results for these outcomes are less compelling than for the primary and other secondary outcomes assessed and should be viewed as suggestive only.

Conclusion

The STAMINA study is, to our knowledge, the most rigorous investigation of a telemedicine-based MOUD linkage intervention conducted to date. Our findings suggest that initial telemedicine linkage alone may be insufficient to address access barriers for individuals with high-severity OUD. Many participants had recent histories of overdose, criminal justice involvement, or unstable housing. For individuals with high severity OUD, institutional mistrust and logistical instability may complicate treatment initiation, and a virtual-only referral model may be insufficient. Additional strategies—such as warm handoffs, peer support, or flexible scheduling—may be needed to optimize the effectiveness of telemedicine in this population.

Distinct medication- and provider-based differences in MOUD initiation also need to be accounted for when designing telemedicine interventions to ensure their benefits are realized equitably. Our results support the importance of rapid treatment connection, suggesting that, ceteris paribus, the speed of the referral process is a critical determinant of success. Exploring models in which MOUD treatment is initiated directly at SSP sites, either through in-person services or telehealth, may offer additional promise for improving linkage, engagement, and retention.

Supplemental Material

sj-docx-1-sat-10.1177_29768357251372336 – Supplemental material for Syringe Service Program-Based Telemedicine Linkage to Opioid Use Disorder Treatment: Results From a Pragmatic Randomized Trial of the STAMINA Intervention

Supplemental material, sj-docx-1-sat-10.1177_29768357251372336 for Syringe Service Program-Based Telemedicine Linkage to Opioid Use Disorder Treatment: Results From a Pragmatic Randomized Trial of the STAMINA Intervention by Dana Franceschini, James A. Swartz, Dennis P. Watson, Mary Ellen Mackesy-Amiti, Lisa Taylor, Peipei Zhao, Sarah Messmer, Antonio D. Jimenez and Nicole Gastala in Substance Use: Research and Treatment

Supplemental Material

sj-docx-2-sat-10.1177_29768357251372336 – Supplemental material for Syringe Service Program-Based Telemedicine Linkage to Opioid Use Disorder Treatment: Results From a Pragmatic Randomized Trial of the STAMINA Intervention

Supplemental material, sj-docx-2-sat-10.1177_29768357251372336 for Syringe Service Program-Based Telemedicine Linkage to Opioid Use Disorder Treatment: Results From a Pragmatic Randomized Trial of the STAMINA Intervention by Dana Franceschini, James A. Swartz, Dennis P. Watson, Mary Ellen Mackesy-Amiti, Lisa Taylor, Peipei Zhao, Sarah Messmer, Antonio D. Jimenez and Nicole Gastala in Substance Use: Research and Treatment

Footnotes

Acknowledgements

Kim Erwin and Adrian Politzer assisted with the design of study protocols and participant-facing materials. The Cook County Medical Examiner’s Office provided mortality data for this study. Mona Stivers provided detailed pre-submission editing services.

ORCID iDs

Dana Franceschini

James A. Swartz

Dennis P. Watson

Ethical Considerations

All procedures involving human subjects were approved by the Institutional Review Board of Chestnut Health Systems (#1138-0420) on April 14, 2020, and a reliance agreement was used to establish a single IRB process with the University of Illinois Chicago.

Consent to Participate

Written informed consent was obtained from all participants prior to trial enrollment.

Author Contributions

Dana Franceschini: Methodology, Formal Analysis, Investigation, Writing - Original Draft; James A. Swartz: Conceptualization, Methodology, Formal Analysis, Writing – Original Draft - Review & Editing, Funding Acquisition; Dennis P. Watson: Conceptualization, Methodology, Formal Analysis, Investigation, Writing - Original Draft, Funding Acquisition; Mary Ellen Mackesy-Amiti: Methodology, Formal Analysis, Writing - Review & Editing; Lisa Taylor: Data Curation, Project Administration; Peipei Zhao: Formal Analysis; Sarah Messmer: Writing - Review & Editing; Antonio D. Jimenez: Writing - Review & Editing; Nicole Gastala: Conceptualization, Data Curation, Writing - Review & Editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was funded by Arnold Ventures (grant number not applicable). Opinions are those of the authors and do not necessarily reflect those of the funder.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

Due to the sensitivity of information related to substance use and treatment, the dataset is available only upon request and with proper data use agreements in place. All materials developed for the study are available upon request.

Supplemental Material

Supplemental material for this article is available online.

References

Substance Abuse and Mental Health Services Administration [SAMHSA]. Key substance use and mental health indicators in the United States: Results from the 2023 National Survey on Drug Use and Health (HHS Publication No. PEP24-07-021, NSDUH Series H-59). 2024. Accessed June 7, 2025. https://www.samhsa.gov/data/report/2023-nsduh-annual-national-report

Garnett

Miniño

AM.

Drug overdose deaths in the United States, 2003–2023. NCHS Data Brief. 2024;(522).

Centers for Disease Control and Prevention. Provisional drug overdose death counts. Accessed June 7, 2025. https://www.cdc.gov/nchs/nvss/vsrr/drug-overdose-data.htm

Krawczyk

Negron

Nieto

Agus

Fingerhood

MI.

Overcoming medication stigma in peer recovery: A new paradigm. Subst Abuse. 2018;39(4):404-409.

Renner

Crawford

MB.

Medication for the treatment of opioid use disorder. In: Kelly

Wakeman

, eds. Treating Opioid Addiction. Springer International Publishing; 2019:89-107.

Treloar

Rance

Yates

Mao

Trust and people who inject drugs: The perspectives of clients and staff of Needle Syringe Programs. Int J Drug Policy. 2016;27:138-145.

Reuter

Smith

McKinnon

Varley

Jouriles

Seaberg

Successful medication for opioid use disorder (MOUD) program at a community hospital emergency department. Acad Emerg Med. 2020;27(11):1187-1190.

Swartz

Franceschini

Marino

Call

Rosenberger

Whitehouse

Barriers and facilitators to prescribing buprenorphine for treating opioid use disorder among emergency department and other practice setting physicians. AIMS Public Health. 2025;12(1):56-76.

D’Onofrio

O’Connor

Pantalon

, et al Emergency department–initiated buprenorphine/naloxone treatment for opioid dependence: A randomized clinical trial. JAMA. 2015;313(16):1636.

10.

Scott

Grella

Nicholson

Dennis

ML.

Opioid recovery initiation: Pilot test of a peer outreach and modified Recovery Management Checkup intervention for out-of-treatment opioid users. J Subst Abuse Treat. 2018;86:30-35.

11.

Nordeck

Buresh

Krawczyk

Fingerhood

Agus

Adapting a low-threshold buprenorphine program for vulnerable populations during the COVID-19 pandemic. J Addict Med. 2021;15(5):364-369.

12.

Jones

Shoff

Blanco

Losby

Ling

Compton

WM.

Association of receipt of opioid use disorder-related telehealth services and medications for opioid use disorder with fatal drug overdoses among Medicare beneficiaries before and during the COVID-19 pandemic. JAMA Psychiatry. 2023;80(5):508-514.

13.

Frost

Zhang

Kim

Lin

LA.

Use of and retention on video, telephone, and in-person buprenorphine treatment for opioid use disorder during the COVID-19 pandemic. JAMA Netw Open. 2022;5(10):e2236298.

14.

Ashford

Curtis

Brown

AM.

Peer-delivered harm reduction and recovery support services: Initial evaluation from a hybrid recovery community drop-in center and syringe exchange program. Harm Reduct J. 2018;15(1):52.

15.

Allen

Grieb

O’Rourke

, et al Understanding the public health consequences of suspending a rural syringe services program: A qualitative study of the experiences of people who inject drugs. Harm Reduct J. 2019;16(1):33.

16.

Treloar

Newland

Maher

A qualitative study trialling the acceptability of new hepatitis C prevention messages for people who inject drugs: Symbiotic messages, pleasure and conditional interpretations. Harm Reduct J. 2015;12(1):5.

17.

Macneil

Pauly

Needle exchange as a safe haven in an unsafe world: Needle exchange as a safe haven. Drug Alcohol Rev. 2011;30(1):26-32.

18.

Platt

Minozzi

Reed

, et al Needle and syringe programmes and opioid substitution therapy for preventing HCV transmission among people who inject drugs: Findings from a Cochrane Review and meta-analysis. Addiction. 2018;113(3):545-563.

19.

Behrends

Gibson

DR.

Decreased odds of injection risk behavior associated with direct versus indirect use of syringe exchange: Evidence from two California cities. Subst Use Misuse. 2017;52(9):1151-1159.

20.

Fernandes

Cary

Duarte

, et al Effectiveness of needle and syringe programmes in people who inject drugs: An overview of systematic reviews. BMC Public Health. 2017;17(1):309.

21.

Kidorf

King

Pierce

Kolodner

Brooner

RK.

Benefits of concurrent syringe exchange and substance abuse treatment participation. J Subst Abuse Treat. 2011;40(3):265-271.

22.

Kidorf

King

VL.

Expanding the public health benefits of syringe exchange programs. Can J Psychiatry. 2008;53(8):487-495.

23.

Heimer

Can syringe exchange serve as a conduit to substance abuse treatment?

J Subst Abuse Treat. 1998;15(3):183-191.

24.

Fox

Sohler

Frost

Lopez

Cunningham

CO.

Development and evaluation of a community-based buprenorphine treatment intervention. Harm Reduct J. 2017;14(1):23.

25.

Hagan

McGough

Thiede

Hopkins

Duchin

Alexander

ER.

Reduced injection frequency and increased entry and retention in drug treatment associated with needle-exchange participation in Seattle drug injectors. J Subst Abuse Treat. 2000;19(3):247-252.

26.

Bråbäck

Nilsson

Isendahl

Troberg

Brådvik

Håkansson

Malmö Treatment Referral and Intervention Study (MATRIS)—effective referral from syringe exchange to treatment for heroin dependence: A pilot randomized controlled trial. Addiction. 2016;111(5):866-873.

27.

Henderson

Vlahov

Celentano

Strathdee

SA.

Readiness for cessation of drug use among recent attenders and nonattenders of a needle exchange program. J Acquir Immune Defic Syndr. 2003;32(2):229-237.

28.

Kidorf

King

Neufeld

Peirce

Kolodner

Brooner

RK.

Improving substance abuse treatment enrollment in community syringe exchangers. Addiction. 2009;104(5):786-795.

29.

Kidorf

Challenges in motivating treatment enrollment in community syringe exchange participants. J Urban Health. 2005;82(3):456-467.

30.

Kuo

Brady

Butler

, et al Feasibility of referring drug users from a needle exchange program into an addiction treatment program: Experience with a mobile treatment van and LAAM maintenance. J Subst Abuse Treat. 2003;24(1):67-74.

31.

Lloyd

Ricketts

Strathdee

, et al Social contextual factors associated with entry into opiate agonist treatment among injection drug users. Am J Drug Alcohol Abuse. 2005;31(4):555-570.

32.

Riley

Safaeian

Strathdee

Brooner

Beilenson

Vlahov

Drug user treatment referrals and entry among participants of a needle exchange program. Subst Use Misuse. 2002;37(14):1869-1886.

33.

Strathdee

Ricketts

Huettner

, et al Facilitating entry into drug treatment among injection drug users referred from a needle exchange program: results from a community-based behavioral intervention trial. Drug Alcohol Depend. 2006;83(3):225-232.

34.

Jakubowski

Fowler

Fox

AD.

Three decades of research in substance use disorder treatment for syringe services program participants: a scoping review of the literature. Addict Sci Clin Pract. 2023;18(1):40.

35.

Colledge-Frisby

Ottaviano

Webb

, et al Global coverage of interventions to prevent and manage drug-related harms among people who inject drugs: a systematic review. Lancet Glob Health. 2023;11(5):e673-e683.

36.

Harm Reduction International. The global state of harm reduction 2022. Published 2022. Accessed November 2022. https://hri.global/wp-content/uploads/2022/11/HRI_GSHR-2022_Full-Report_Final-1.pdf

37.

Chan

Bougatsos

Priest

McCarty

Grusing

Chou

Opioid treatment programs, telemedicine and COVID-19: a scoping review. Subst Abus. 2022;43(1):539-546.

38.

Guillen

Reddy

Saadat

Chakravarthy

Utilization of telehealth solutions for patients with opioid use disorder using buprenorphine: a scoping review. Telemed J E Health. 2022;28(6):761-767.

39.

Lambdin

Bluthenthal

Tookes

, et al Buprenorphine implementation at syringe service programs following waiver of the Ryan Haight Act in the United States. Drug Alcohol Depend. 2022;237:109504.

40.

McKellar

Des Marais

Chen

, et al Providing medication for opioid use disorder and HIV pre-exposure prophylaxis at syringe services programs via telemedicine: a pilot study. Harm Reduct J. 2024;21(1):69.

41.

Watson

Swartz

Robison-Taylor

, et al Syringe service program-based telemedicine linkage to opioid use disorder treatment: protocol for the STAMINA randomized control trial. BMC Public Health. 2021;21(1):630.

42.

National Committee on Quality Assurance (NCQA). HEDIS Measure: Initiation and engagement of alcohol and other drug dependence treatment (IET). Accessed September 5, 2025. https://www.caresource.com/documents/ga-mcd-iet-hedis-measure/

43.

Substance Abuse and Mental Health Services Administration (SAMHSA). 2018 National Survey on Drug Use and Health: Methodological summary and definitions. Published 2019. Accessed September 5, 2025. https://www.samhsa.gov/data/

44.

Kessler

Green

Gruber

, et al Screening for serious mental illness in the general population with the K6 screening scale: results from the WHO World Mental Health (WMH) survey initiative. Int J Methods Psychiatr Res. 2010;19(suppl 1):4-22.

45.

American Psychiatric Association, ed. Diagnostic and Statistical Manual of Mental Disorders. 5th ed. American Psychiatric Association; 2013.

46.

Wesson

Ling

The Clinical Opiate Withdrawal Scale (COWS). J Psychoactive Drugs. 2003;35(2):253-259.

47.

Harris

Taylor

Thielke

Payne

Gonzalez

Conde

JG.

Research electronic data capture (REDCap)—a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform. 2009;42(2):377-381.

48.

Glick

Prohaska

LaKosky

Juarez

Corcorran

Des Jarlais

DC.

The impact of COVID-19 on syringe services programs in the United States. AIDS Behav. 2020;24(9):2466-2468.

49.

Moher

Hopewell

Schulz

, et al CONSORT 2010 explanation and elaboration: updated guidelines for reporting parallel group randomised trials. J Clin Epidemiol. 2010;63(8):e1-e37.

50.

Stata 19.5 for Mac. StataCorp; 2025.

51.

Zorn

CJW

. Generalized estimating equation models for correlated data: a review with applications. Am J Polit Sci. 2001;45(2):470-490.

52.

Zorn

Comparing GEE and robust standard errors for conditionally dependent data. Polit Res Q. 2006;59(3):329-341.

53.

NIH Pragmatic Trials Collaboratory. Small-sample robust variance correction for generalized estimating equations for use in cluster randomized clinical trials. Published 2015. Accessed September 5, 2025. https://dcricollab.dcri.duke.edu/sites/NIHKR/KR/Variance-correction-for-GEE_V1.R0.pdf

54.

Jakubowski

Norton

Hayes

, et al Low-threshold buprenorphine treatment in a syringe services program: program description and outcomes. J Addict Med. 2022;16(4):447-453.

55.

Suarez

Bartholomew

Plesons

, et al Adaptation of the tele-harm reduction intervention to promote initiation and retention in buprenorphine treatment among people who inject drugs: a retrospective cohort study. Ann Med. 2023;55(1):733-743.

56.

Muthulingam

Bia

Madden

Farnum

Barry

Altice

FL.

Using nominal group technique to identify barriers, facilitators, and preferences among patients seeking treatment for opioid use disorder: a needs assessment for decision making support. J Subst Abuse Treat. 2019;100:18-28.

57.

Sud

Salamanca-Buentello

Buchman

Sabioni

Majid

Beyond harm-producing versus harm-reducing: a qualitative meta-synthesis of people who use drugs’ perspectives of and experiences with the extramedical use and diversion of buprenorphine. J Subst Abuse Treat. 2022;135:108651.

58.

Srivastava

Kahan

Leece

McAndrew

Buprenorphine unobserved “home” induction: a survey of Ontario’s addiction physicians. Addict Sci Clin Pract. 2019;14(1):18.

59.

Department of Health and Human Services (DHHS). Medications for the treatment of opioid use disorder. 42 CFR 8. DHHS; 2024:7528-7563.

60.

Simpson

Hordes

Blum

Rinehart

Al-Tayyib

Barriers to engagement in opioid use disorder treatment after buprenorphine induction. J Addict Med. 2022;16(4):479-482.

61.

Degenhardt

Clark

Macpherson

, et al Buprenorphine versus methadone for the treatment of opioid dependence: a systematic review and meta-analysis of randomised and observational studies. Lancet Psychiatry. 2023;10(6):386-402.

62.

Pasman

Kollin

Broman

, et al Cumulative barriers to retention in methadone treatment among adults from rural and small urban communities. Addict Sci Clin Pract. 2022;17(1):35.

63.

Frank

Mateu-Gelabert

Perlman

Walters

Curran

Guarino

“It’s like ‘liquid handcuffs’”: the effects of take-home dosing policies on methadone maintenance treatment (MMT) patients’ lives. Harm Reduct J. 2021;18(1):88.

64.

Gressick

Fiorillo

Richardson

, et al Telemedicine to improve access to medications for opioid use disorder in Illinois, 2022-2024. Int J Drug Policy. 2025;137:104729.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.03 MB