Association of Primary Care Engagement with Initiation and Continuation of Medication Treatment for Opioid Use Disorder Among Persons with a History of Injection Drug Use

Abstract

Background:

For patients with opioid use disorder (OUD), primary care can serve as a pathway to medication for OUD (MOUD). No community-based studies have examined whether people with OUD engaged in primary care are more likely to a) initiate or b) continue MOUD.

Methods:

Data were collected 2014–2020 from two subsamples of the AIDS Linked to the Intravenous Experience (ALIVE) cohort, a community-recruited cohort of people from Baltimore who have injected drugs: 1) people who reported past-six-month illicit opioid use and no MOUD (360 participants, 789 study visits), and 2) people who reported MOUD and no illicit opioid use in the past six months (561 participants, 2,027 visits). Logistic regression was used to estimate associations of past six-month self-reported primary care engagement, respectively, with a) initiating MOUD, b) continuing MOUD, and c) cessation from illicit opioid use without initiating MOUD.

Results:

Among 360 persons not on MOUD treatment (28% female, 26% under 50, 59% actively injecting drugs), primary care engagement was not associated with either cessation from illicit opioid use or initiating MOUD. Similarly, among persons on MOUD (40% female, 22% under 50, 6% actively injecting drugs) primary care engagement was not associated with continued treatment.

Conclusions:

Our findings implicate missed opportunities to initiate and maintain buprenorphine treatment in primary care settings.

1. Introduction

The most effective medications for opioid use disorder (MOUD) incorporate prescribing of methadone and buprenorphine. As compared to both no treatment and to other treatment modalities, patients receiving MOUD are less likely to overdose, less likely to use illicit opioid drugs, more likely to remain in engaged in care, less likely to contract HIV infection, more likely to adhere to HIV medications, and more likely to experience other positive social and functional outcomes.^1,2 Because of their effectiveness, expanding the use of these medications has been identified as a key public health strategy for reducing the United States’ national overdose rate, which reached record levels of more than 100,000 per year in 2021.³

However, despite the effectiveness of medications for treating opioid use disorder, only a small fraction of adults with an opioid use disorder in the United States receive medication treatment.^4,5 This low utilization in the United States occurs, at least in part, because substance use treatment has historically been segregated in its own system, with patients required to seek care in specialty settings; these specialty settings may be difficult to access or may impose additional conditions for patients to receive treatment such as mandates that patients attend counseling or complete frequent urine drug screens.⁶

Consequently, some experts have identified better integration of substance use treatment with primary care as a key strategy for addressing the treatment gap for opioid use disorder.⁷ Patients in substance use treatment who are more regularly engaged in general medical care are more likely to achieve sustained cessation of substance use and reduced substance use severity.^8,9 These benefits of general medical care engagement for reducing substance use are particularly great for patients with substance use associated physical health conditions.¹⁰ However, in the United States, integration of treatment with medications for opioid use disorder in primary care settings has been impeded by regulatory requirements governing methadone and, to a lesser extent, buprenorphine. In particular, methadone may not be prescribed by clinicians; it may only be dispensed from specially licensed Opioid Treatment Programs and, prior to 2020, could usually only be dispensed in daily doses that patients take while under observation by a clinician. Unlike methadone, buprenorphine may be prescribed by a clinician. However, until the end of 2022, only clinicians with a special license known as the X-waiver¹¹ could prescribe buprenorphine, and additional training was required for waivered prescribers to prescribe to more than 30 patients.¹¹

Despite these historic restrictions, buprenorphine remains the foundation of office-based treatment of opioid use disorder in the United States, and buprenorphine prescribing presents a significant opportunity to integrate treatment for opioid use disorder in primary care. Indeed, there is evidence this promise is increasingly being realized. From 2009 to 2018, buprenorphine prescriptions per capita in the United States more than doubled,¹² and providers with a waiver permitting them to prescribe buprenorphine quadrupled.¹³ These prescribing and access increases occurred even though buprenorphine has historically been more restricted than other Schedule II drugs. In December of 2022, Congress eliminated these additional requirements so that any physician with a controlled substance license may now prescribe buprenorphine. This potentially further expands the opportunity to integrate substance use treatment into a broader set of primary care practices. Further, in 2024, the Substance Abuse and Mental Health Services Administration (SAMHSA) revised federal regulations to permit greater dispensing of methadone, including immediately permitting limited “take home” methadone dosing, expanding eligibility for treatment, eliminating counseling requirements, and allowing for screening for methadone eligibility via telehealth. While this policy liberalization still does not permit prescribing of methadone by primary care physicians, it may facilitate new opportunities for closer collaboration between primary care providers and OTPs, for example by capitalizing on telehealth provisions.

Most studies on treating substance use disorder in primary care focus on quality of care for populations already enrolled in specialty substance use treatment. However, despite the potential benefits of integrating substance use treatment with primary care, a decade of expansion in buprenorphine prescribing, and the promise of liberalized buprenorphine prescribing and methadone dispensing regulations, to our knowledge, there are no studies in community samples (where patients may move in and out of treatment) on whether patients who are engaged in primary care are actually more likely to utilize MOUD. Addressing this gap in knowledge is important, both to understand whether primary care providers who treat patients using illicit opioids are effective at helping their patients initiate and remain on buprenorphine treatment, and also to establish a baseline for future evaluation of the effects of the United States’ new buprenorphine policy.

This study uses data from an active, community-recruited cohort study of adults who have injected drugs to examine the role of primary care providers in connecting persons with opioid use disorder to medication treatment. Specifically, we evaluated whether engagement in primary care was associated with: a) initiation of MOUD following a period when participants were not receiving any MOUD and b) maintenance on MOUD among participants already on MOUD. We also examined, as a secondary analysis, whether primary care engagement was associated with cessation from illicit opioid use without initiation of MOUD. Based on the literature summarized above and the context of U.S. regulatory policy, we hypothesized that patients engaged in primary care would be modestly more likely to initiate and remain on buprenorphine treatment than patients not engaged in primary care, but that primary care engagement would not be associated with methadone treatment.

2. Methods

2.1. Study Sample

Study participants were from the AIDS Linked to the Intravenous Experience (ALIVE) Study. Complete details of the ALIVE Study, which began recruitment in 1988, are described elsewhere.¹⁴ Briefly, the ALIVE Study is an ongoing, community-based, prospective cohort study of people who have a history of injection drug use, living in and around Baltimore, Maryland. Approximately 2,600 adults were originally recruited in 1988, and subsequent waves of recruitment occurred in 1994–1995, 2005–2008, and 2015–2018. Participants attend twice-annual study visits where they complete a physical exam, standardized interviewer-administered and audio-computer assisted surveys about drug use, related health-risk behaviors, and physical and mental health outcomes, and provide a blood sample for HIV, Hepatitis C (HCV) antibody testing, and other biomarker measurements. The Johns Hopkins University Institutional Review Board has continuously approved the ALIVE study. All study procedures were conducted in accordance with Johns Hopkins University Institutional Review Board’s standards and the Declaration of Helsinki. All participants provided written informed consent.

Our analyses in this study included ALIVE participants who attended at least three study visits between 2014 to March 2020, the period during which the main study variables were measured consistently. Separate analytic samples were used for each study outcome (see details below). For each analysis, participants’ status on the primary exposure of interest – primary care engagement – and other analytic covariates were drawn from responses at the visits meeting the analysis-specific inclusion criteria described below (hereafter “index visit”). Outcome measures – initiation or retention in treatment, respectively (as defined below) – were drawn from participants’ next study visit (hereafter “follow-up visit”), to maintain the temporal relationship necessary to establish initiation (transitioning from no medication at one visit to medication at the next) and discontinuation (transitioning from medication at one visit to no medication at the next).

2.1.1. Initiation analytic sample

To study initiation of MOUD, we included as index visits all visits at which a participant reported: a) any past-six-month illicit opioid use (including speedball, injecting or ingesting painkillers not prescribed by a doctor, and smoking, snorting, or injecting heroin) and b) no use of MOUD in the past six months (see “Measures” for definition). This resulted in 491 participants contributing 1,321 visits. We then limited the analytic sample to those who had complete data for all covariates used in the analysis. Our final sample for the initiation analysis comprised 360 participants contributing 789 visits.

2.1.2. Discontinuation analytic sample

To study discontinuation of MOUD, we included as index visits all study visits at which a participant reported a) past-six-month use of MOUD and b) no illicit opioid use in the past six months. This resulted in 631 participants contributing 2,695 visits. After limiting to those participants who had complete data on all covariates, our final analytic sample for the discontinuation analysis comprised 561 participants contributing 2,027 visits. We also conducted two sub-analyses: a) including only persons reporting past-six-month methadone treatment (429 participants; 1,608 visits), and b) including only persons who reported past-six-month buprenorphine treatment (155 participants; 419 visits).

2.2. Measures

2.2.1. Exposure of interest

The exposure of interest was primary care engagement. During each study visit, participants responded to three separate questions: (1) “Do you have one doctor’s office or clinic location where you go for most of your medical care?” (2) “When you go there, do you usually (more than 90% of the time) see the same doctor, nurse, or physician’s assistant?” and (3) “Other than emergency rooms, have you been seen or received care at a doctor’s office or clinic during the last six months?” Participants who responded “yes” to all three questions were considered “engaged” in primary care during the six months preceding their index visit, whereas all other response patterns were coded as “not engaged” in primary care.

2.2.2. Outcomes

During each study visit, participants were asked, using separate questions, whether they received methadone or buprenorphine treatment (including prescriptions) in the past six months.

2.2.2.1. Initiation of medication for opioid use disorder

Among participants who reported illicit opioid use and no use of MOUD at their index visit (see “Initiation Analytic Sample” above), we created a two-category outcome variable based on participants’ self-reported behavior at their follow-up visit: 1) reported no MOUD and continued illicit opioid use (“continued use without medication”) or 2) reported medication treatment at the follow-up visit (“initiated treatment”). For a secondary analysis focused on initiation by medication type, we separated “initiated treatment” into a) “initiated methadone treatment” and b) “initiated buprenorphine treatment.”

2.2.2.2. Discontinuation of medication for opioid use disorder

Among participants who reported no illicit opioid use and use of MOUD at their index visit (see “Continuation Analytic Sample” above), we created a two-category outcome variable based on participants’ self-reported behavior at their follow-up visit: 1) the participant reported MOUD (“continued medication treatment”), or 2) the participant reported no MOUD, or the participant died within six months of the index visit (“stopped medication treatment, or death”). For sub-analyses focused on specific medication types, the outcome was the same, but restricted to the relevant medication type.

2.2.2.3. Cessation from illicit opioid use without MOUD

Among participants who reported illicit opioid use and no use of MOUD at their index visit, we also created a two-category outcome variable based on participants’ self-reported illicit opioid use at their follow-up visit: 1) reported illicit opioid use and no MOUD (“continued use without medication”) or 2) reported cessation from illicit opioid use and no MOUD (“cessation without MOUD”).

2.2.3. Confounders

We included various sociodemographic and clinical characteristics that may confound the associations tested for each study goal. Sociodemographic characteristics included self-reported sex, race (i.e., Black versus other), age (split into tertiles), cohort enrollment period (1988–1989, 1994–2000, 2005, 2018), five dummy-coded variables for calendar years 2014–2019, and homelessness in the past six months. Clinical and behavioral characteristics included HCV status (from antibody test at index visit), as well as self-reported injection drug use, cigarette use, alcohol use, other non-opioid drug use (i.e., cocaine, sedatives, tranquilizers, stimulants, and hallucinogens), emergency room visit, non-fatal overdose, and HIV status. Finally, to adjust participants receipt of MOUD, we adjusted for whether participants reported any methadone or buprenorphine treatment at their most recent study visit immediately preceding their index visit.

2.3. Statistical Analysis

Prior to analyses, we described the visit-level prevalence of participant characteristics (see Tables 1 and 3). We also calculated the prevalence of primary care engagement by outcome status for all models (see Tables 2 and 4).

Table 1.

Visit-level prevalence of characteristics among 360 ALIVE participants reporting illicit opioid drug use and no medication treatment for opioid use disorder

	Full Sample (789 visits)	Primary Care Engagement (490 visits)	No Primary Care (299 visits)
Sex
Male	72%	71%	72%
Female	28%	29%	28%
Race
Black	86%	89%	80%
Other	14%	11%	20%
Age
Lower Tertile (Under 50)	26%	19%	38%
Middle Tertile (50–59)	49%	53%	43%
Upper Tertile (60–74)	24%	28%	19%
Cohort
1988–1989	21%	24%	16%
1994–1995	7%	6%	8%
1998–2000	4%	4%	5%
2005	42%	43%	41%
2018	26%	24%	30%
Calendar Year
2014	8%	9%	8%
2015	17%	18%	17%
2016	19%	18%	20%
2017	21%	19%	23%
2018	23%	22%	23%
2019	12%	14%	9%
Prior MOUD
Yes	21%	21%	21%
No	79%	79%	79%
Cigarette use
Yes	86%	83%	93%
No	14%	17%	7%
Alcohol use
Yes	75%	77%	72%
No	25%	23%	28%
Other illicit drug use
Yes	64%	61%	67%
No	36%	39%	33%
Injection drug use
Yes	59%	56%	63%
No	41%	44%	37%
ER Visit
Yes	27%	29%	23%
No	73%	71%	77%
Non-fatal overdose
Yes	8%	10%	6%
No	92%	90%	94%
Homelessness
Yes	12%	9%	18%
No	88%	91%	82%
HCV antibody
Yes	71%	72%	69%
No	29%	28%	31%
HIV
Negative	77%	69%	89%
Positive, RNA level < 400 copies/mL	15%	22%	4%
Positive, RNA level ≥ 400 copies/mL	8%	9%	8%

Note. Age tertiles are approximations after rounding. Behavioral characteristics (e.g., ER visit, injection and other drug use, overdose, homelessness, HIV status) reflect self-report data in the past six-months. Values may not add to 100 due to rounding. MOUD = medication treatment for opioid use disorder.

Table 3.

Visit-level prevalence of characteristics among 561 ALIVE participants reporting medication treatment for opioid use disorder and no illicit opioid use

	Full Sample (2,027 visits)	Primary Care Engagement (1,593 visits)	No Primary Care (434 visits)
Sex
Male	60%	59%	66%
Female	40%	41%	34%
Race
Black	87%	90%	79%
Other	13%	10%	21%
Age
Lower Tertile (Under 50)	22%	18%	35%
Middle Tertile (50–59)	47%	49%	42%
Upper Tertile (60–75)	31%	33%	23%
Cohort
1988–1989	27%	28%	22%
1994–1995	12%	12%	12%
1998–2000	8%	7%	10%
2005	43%	43%	40%
2018	11%	10%	16%
Calendar Year
2014	8%	8%	9%
2015	20%	20%	21%
2016	19%	18%	20%
2017	20%	20%	18%
2018	21%	22%	20%
2019	12%	12%	12%
Prior MOUD
Yes	87%	87%	85%
No	13%	13%	15%
Cigarette use
Yes	80%	79%	82%
No	20%	21%	18%
Alcohol use
Yes	36%	35%	38%
No	64%	65%	62%
Other illicit drug use
Yes	23%	21%	30%
No	77%	79%	70%
Injection drug use
Yes	6%	5%	8%
No	94%	95%	92%
ER Visit
Yes	26%	26%	22%
No	74%	74%	78%
Non-fatal overdose
Yes	1%	1%	1%
No	99%	99%	99%
Homelessness
Yes	7%	6%	12%
No	93%	94%	88%
HCV antibody
Yes	87%	89%	82%
No	13%	11%	18%
HIV
Negative	67%	63%	81%
Positive, RNA level < 400 copies/mL	24%	27%	11%
Positive, RNA level ≥ 400 copies/mL	9%	9%	9%

Note. Age tertiles are approximations after rounding. Behavioral characteristics (e.g., ER visit, injection and other drug use, overdose, homelessness, HIV status) reflect self-report data in the past six-months. Values may not add to 100 due to rounding. MOUD = medication treatment for opioid use disorder.

Table 2.

Results from logistic regression models testing the association between primary care engagement and probability of initiating medication treatment for opioid use disorder, and cessation from illicit opioid use

	Primary Care	No Primary Care	Crude OR [95% CI]	Adjusted OR [95% CI]
Model Set 1
Stopped illicit opioid use without medication	6%	6%	1.04 [0.56, 1.93]	1.19 [0.60, 2.36]
Model Set 2
Initiated medication treatment	25%	30%	0.79 [0.57, 1.09]	0.81 [0.56, 1.16]
Model Set 3
Initiated Buprenorphine	10%	9%	1.17 [0.71, 1.92]	1.16 [0.67, 2.01]
Model Set 4
Initiated Methadone	15%	21%	0.67 [0.46, 0.97]	0.69 [0.46, 1.05]

Note. Analyses consist of 360 participants across 789 visits. Bold indicates statistical significance (p < 0.05). OR = Odds Ratio. Values in primary care columns reflect prevalence of each outcome within exposure groups across all visits. 95% Confidence Intervals are calculated based on robust standard errors. Adjusted models include the following covariates: sex, race, age, cohort, calendar year, HCV antibody status, and past six-month medication treatment for opioid use, homelessness, injection drug use, HIV status, ER visit, non-fatal overdose, and other drug use.

Table 4.

Results from logistic regression models testing the association between primary care engagement and probability of continuing medication treatment for opioid use disorder

	Primary Care	No Primary Care	Crude OR [95% CI]	Adjusted OR [95% CI]
Model Set 4 (n = 561; 2,027 visits)
Continued medication treatment	88%	89%	0.98 [0.71, 1.37]	0.88 [0.62, 1.26]
Model Set 5 (n = 429; 1,608 visits)
Continued Methadone treatment	91%	90%	1.04 [0.71, 1.54]	0.89 [0.59, 1.36]
Model Set 6 (n = 155; 419 visits)
Continued Buprenorphine treatment	77%	70%	1.43 [0.77, 2.65]	1.33 [0.69, 2.60]

Note. Bold indicates statistical significance (p < 0.05). RRR = Relative Risk Ratio. Values in primary care columns reflect prevalence of each outcome within exposure groups across all visits. 95% Confidence Intervals are calculated based on robust standard errors. Adjusted models include the following covariates: sex, race, age, cohort, calendar year, HCV antibody status, and past six-month medication treatment for opioid use disorder, homelessness, non-opioid injection drug use, HIV status, ER visit, and other drug use. Non-fatal overdose was not adjusted for in these models due to limited variability in responses.

For subsequent inferential analyses, we estimated the association of primary care engagement with each outcome using logistic regression. The resulting (exponentiated) regression coefficients we present are interpreted as odds ratios (ORs). Our analytic approach was similar for analyses of discontinuation. In each instance, we tested a model without adjusting for potential confounders, followed by an adjusted model with all confounders included. The method of generalized estimating equation (GEE) with robust standard errors and an assumption of independent correlation structures was used to account for the fact that participants contributed repeated study visits. For each measure of association, we also report a 95% Wald confidence interval.^15,16 All analyses were conducted using R statistical software, using the packages and geepack.¹⁶

3. Results

3.1. Initiation of Medication for Opioid Use Disorder

Among persons not on MOUD, median age at participants’ index visit was 52.3 years (interquartile range [IQR] = 11); 72% were male and 84% self-identified as Black. Participants reported primary care engagement at approximately 62% of index visits. At follow-up visits, 17% initiated methadone treatment, 10% initiated buprenorphine treatment, and 6% stopped using opioid drugs without medication treatment. Table 1 presents the visit-level prevalence of study characteristics among these participants.

There was no evidence to suggest that participants who reported primary care engagement were significantly more likely to report cessation from illicit opioid use without initiating MOUD compared to participants without primary care engagement (Adjusted OR = 1.19, 95% CI = 0.60–2.36) (Table 2). Similarly, there was no evidence that primary care engagement was significantly associated with initiation of any MOUD (Adjusted OR = 0.81, 95% CI = 0.56–1.16) or buprenorphine initiation (Adjusted OR = 0.69, 95% CI = 0.46–1.05) (Table 2). There was a crude association between primary care engagement and lower odds of methadone initiation (OR = 0.67, 95% CI = 0.46–0.97); however, this association was null after adjustment for confounders (Adjusted OR = 0.69, 95% CI = 0.46–1.05) (Table 2).

3.2. Discontinuation of Medication for Opioid Use Disorder

Among persons already on MOUD, median age at participants’ index visit was 53.3 years (IQR = 11); 63% of the sample was male and 84% self-identified as Black. Participants reported primary care engagement at approximately 79% of study visits. At follow-up visits, 88% remained on medication treatment. Table 3 presents the visit-level prevalence of study characteristics among these participants; Tables S1–S2 present visit-level prevalence of study characteristics among the subsamples already in buprenorphine and methadone treatment, respectively.

Primary care engagement was not associated with discontinuation of MOUD (Adjusted OR = 0.88, 95% CI = 0.62–1.26] (Table 4). Results were similar when splitting the sample by medication treatment type – there was no evidence that primary care engagement was significantly associated with a discontinuing buprenorphine (OR = 1.33, 95% CI = 0.69–2.60) or methadone use (Adjusted ORR = 0.89, 95% CI = 0.59–2.60) (Table 4).

4. Discussion

In our first prospective analysis of study visits by adults who likely needed treatment for opioid use disorder but who were not engaged in medication treatment, respondents reported primary care engagement at close to two thirds of visits, a level of engagement we find encouraging for a population often perceived as being “disconnected” from essential services. Just over one quarter of these visits at which participants needed treatment were followed by participants starting either methadone or buprenorphine treatment for opioid use disorder, and another five percent by stopping illicit opioid use without medication. However, after adjusting for potential confounders, having a primary care provider and a recent outpatient visit was not associated with initiation of buprenorphine or methadone, nor was it associated with stopping illicit opioid use without treatment. Further, in our second analysis, focused on patients who were already engaged in methadone or buprenorphine treatment, primary care engagement was reported at three-quarters of visits. However, more than one-in-ten participants were no longer engaged in treatment at their next six-month study visit, and primary care was also not associated with retention in either form of treatment.

These findings run counter to our a priori hypothesis that primary care engagement would improve buprenorphine engagement. We believe the contrast seen in this study – relatively high levels of primary care engagement but low levels of MOUD uptake and no association between care-engagement and uptake – represents a missed opportunity. If people who use opioids are choosing to receive outpatient medical care and are engaged with a primary care provider – and our findings suggest that more than three in five may meet these criteria – this should be a pathway to MOUD. Our findings suggest that, from 2014–2020, primary care engagement was not such a pathway. Instead, for the participants in this study, participation in evidence-based substance use disorder treatment appears to have remained largely independent of participation in general medical care.

This shortfall is likely at least in part attributable to the low percentage of primary care providers who had waivers to prescribe buprenorphine prior to recent changes in federal law. Unfortunately, we cannot know if participants’ primary care physicians were waivered providers (see “Limitations” below). However, in 2012 (shortly before the start of data collection for this study) fewer than 1 in 30 primary care providers in the United States had a waiver to prescribe buprenorphine.¹⁷ The fact that waivers are no longer required to prescribe buprenorphine should make it easier for primary care providers to engage patients in buprenorphine treatment in particular. However, that so few providers were waivered before likely reflects, at least in part, other barriers to treating patients with opioid use disorder. Qualitative studies have found many primary care providers are reluctant to prescribe buprenorphine because of concerns about attracting patients with substance use disorder into their practice, concerns patients will give or sell their buprenorphine to others, and concerns about their own efficacy to prescribe and manage buprenorphine treatment correctly. Providers might also perceive addiction treatment to be more appropriately the purview of specialists. Ending the waiver requirement is a step in the right direction.¹⁸ Past studies suggesting these concerns are prevalent among primary care providers suggests liberalized buprenorphine policies on their own may not be a panacea for addressing the missed treatment opportunities identified in this study. While recent federal changes also should also result in expanded access to, and stability of, methadone treatment through OTPs, federal law still does not permit prescribing of buprenorphine, so these changes have limited implication for integration of MOUD into primary care.

Although there are challenges in engaging primary care providers to prescribe buprenorphine or link patients to methadone, it is notable that in the decade leading up to 2022 (when waivers were eliminated) the number of physicians with waivers to prescribe buprenorphine grew dramatically, and most of this growth was among primary care practitioners and non-psychiatrist specialty practitioners.¹⁹ Research also shows that, when physicians actually begin prescribing buprenorphine to patients with opioid use disorder, many of the aforementioned negative attitudes about patients with substance use disorder and substance use treatment become more positive.¹⁸ However, at minimum, our finding that participants engaged with primary care were no more likely to initiate or remain on medication treatment that participants not engaged in primary care implies that, as U.S. health care systems and providers seek to capitalize on new flexibility to prescribe buprenorphine for treatment of opioid use disorder and expanded access to methadone through OTPs, we are starting from a baseline with substantial room for growth and improvement.

4.1. Limitations

First, in this study, both primary care engagement and substance use treatment were measured using self-report which may suffer from recall bias. Our measure of primary care engagement suffers from a second additional source of measurement error, because participants were not actually asked if they saw their primary care provider in the past six months, only whether they had an outpatient visit. These measurement shortfalls may erode our ability to detect an association between primary care engagement and MOUD. Although a strength of this study is the fact that we are able to observe the behavior of people who use opioids who are not recruited from, and not necessarily engaged with, any healthcare services, future research should seek to replicate these findings using more objective measures of care engagement, such as insurance billing records or electronic health records. Second, this study is missing important information that could help explain why primary care providers were unable to connect their patients to MOUD, including whether participants saw providers waivered to prescribe buprenorphine, and what attitudes providers held about prescribing. Third, participants in this study – who were recruited from a city in the Northeast United States because of a history of injecting drugs specifically – many not be representative of the broader population of adults who use opioid drugs illicitly. In particular, nearly all ALIVE participants are insured, with such a high prevalence that we cannot study the role of insurance coverage in influencing access to care; insurance coverage of similar populations in different states may be different if those states have less expansive Medicaid eligibility. Fourth, as noted in Methods, our definition of “initiating” MOUD is self-reported past six-month use of medications following an at least six-month period in which no MOUD was used. Participants may have used MOUD at some point in the past: thus, our definition of “initiation” encompasses both first-time users and persons restarting treatment after a gap of at least six months, although we did adjust for participants’ recent history of medication treatment in our analysis. Fifth, our study did not include naltrexone as an outcome. This was because the prevalence of naltrexone use in this cohort was exceedingly low and could not be analyzed. Sixth, while our study design used a longitudinal design to ensure that the assessment of primary care engagement preceded the initiation of MOUD, some MOUD providers may require patients to have a primary care visit before initiating treatment. If participants received primary care only for the purpose of planned future MOUD initiation, this could bias our estimates due to reverse-causation. Seventh, in our analysis of treatment retention, we could only examine whether participants reported any MOUD at consecutive visits: we could not establish the precise duration of treatment, nor could we determine if participants experienced gaps in their treatment throughout either period. Finally, though we adjusted for numerous potential confounding variables, there may be unmeasured participant attributes associated with both primary care and substance use engagement: for example, self-efficacy, disease burden, or socioeconomic factors not captured by the variables in this analysis.

5. Conclusion

Evidence supports the effectiveness of integrating MOUD into primary care practice for improving both substance use and general health outcomes. Liberalized federal regulations governing buprenorphine and methadone present a profound opportunity to expand access to care to the great majority of adults with opioid use disorder who are currently not receiving evidence-based treatment and advance a key public health strategy for addressing the overdose crisis. However, our research finds that, in the years leading up to this new policy, patients using opioids illicitly who were recently engaged in primary care were no more likely to initiate either buprenorphine or methadone treatment than patients not engaged with primary care. This is further evidence that the United States is starting from a weak baseline in terms of integrating MOUD into primary care. Educational support, financial incentives, and proactive support from health systems may be needed to help primary care providers capitalize on their new prescribing freedoms and provide evidence-based care to their patients. Policy makers should also consider whether methadone liberalization goes far enough to facilitate access outside OTPs, since methadone is the MOUD with the longest history of use, is associated with superior treatment retention compared to buprenorphine, and outside the United States may be prescribed similar to other opioid medications.²⁰ In addition, future research should track primary care engagement and the link between primary care and buprenorphine initiation and maintenance as key measures for evaluating the effectiveness of newly liberalized federal buprenorphine in ameliorating the U.S. overdose crisis.

Figure 1.

Prevalence of Medication Treatment Use Over Time Among Full Study Sample

Highlights.

• Examined primary care, MOUD uptake among people who have injected drugs

• Visit-level prevalence of primary care engagement was 62%

• Primary care engagement not associated with MOUD initiation or continuation

Appendix Table 1. Visit-level prevalence of characteristics among 155 ALIVE participants reporting buprenorphine use and no illicit opioid use

	Full Sample (419 visits)	Primary Care Engagement (363 visits)	No Primary Care (56 visits)
Sex
Male	63%	62%	70%
Female	37%	38%	30%
Race
Black	93%	93%	96%
Other	7%	7%	4%
Age
Lower Tertile (Under 50)	12%	10%	20%
Middle Tertile (50–59)	56%	58%	48%
Upper Tertile (60–79)	32%	32%	32%
Cohort
1988–1989	25%	26%	18%
1994–1995	10%	10%	9%
1998–2000	7%	6%	14%
2005	48%	49%	43%
2018	10%	9%	16%
Calendar Year
2014	8%	9%	9%
2015	20%	20%	21%
2016	15%	14%	25%
2017	19%	19%	16%
2018	22%	23%	16%
2019	16%	16%	13%
Prior MOUD
Yes	76%	77%	70%
No	24%	23%	30%
Cigarette use
Yes	78%	77%	79%
No	22%	23%	21%
Alcohol use
Yes	38%	37%	45%
No	62%	63%	55%
Other illicit drug use
Yes	13%	12%	18%
No	87%	88%	82%
Injection drug use
Yes	3%	2%	5%
No	93%	98%	95%
ER Visit
Yes	21%	21%	25%
No	79%	79%	75%
Non-fatal overdose
Yes	1%	1%	0%
No	99%	99%	100%
Homelessness
Yes	5%	4%	9%
No	95%	96%	91%
HCV antibody
Yes	87%	88%	79%
No	13%	12%	21%
HIV
Negative	51%	49%	66%
Positive, RNA level < 400 copies/mL	38%	41%	20%
Positive, RNA ≥ 400 copies/mL	11%	10%	14%

Note. Age tertiles are approximations after rounding. Behavioral characteristics (e.g., ER visit, injection and other drug use, overdose, homelessness, HIV status) reflect self-report data in the past six-months. Values may not add to 100 due to rounding. MOUD = medication treatment for opioid use disorder.

Appendix Table 2. Visit-level prevalence of characteristics among 429 ALIVE participants reporting methadone use and no illicit opioid use

	Full Sample (1,608 visits)	Primary Care Engagement (1,230 visits)	No Primary Care (378 visits)
Sex
Male	59%	57%	66%
Female	41%	43%	34%
Race
Black	86%	89%	76%
Other	14%	11%	24%
Age
Lower Tertile (Under 50)	25%	21%	38%
Middle Tertile (50–59)	45%	46%	41%
Upper Tertile (60–75)	31%	33%	21%
Cohort
1988–1989	27%	29%	22%
1994–1995	12%	12%	12%
1998–2000	8%	7%	10%
2005	41%	42%	40%
2018	11%	10%	16%
Calendar Year
2014	9%	7%	10%
2015	20%	20%	21%
2016	19%	20%	19%
2017	20%	21%	18%
2018	21%	21%	20%
2019	11%	11%	12%
Prior MOUD
Yes	90%	90%	88%
No	10%	10%	12%
Cigarette use
Yes	81%	80%	82%
No	19%	20%	18%
Alcohol use
Yes	35%	35%	37%
No	65%	65%	63%
Other illicit drug use
Yes	25%	23%	31%
No	75%	77%	69%
Injection drug use
Yes	7%	6%	8%
No	93%	94%	92%
ER Visit
Yes	27%	28%	22%
No	73%	72%	78%
Non-fatal overdose
Yes	1%	1%	1%
No	99%	99%	99%
Homelessness
Yes	8%	6%	12%
No	92%	94%	88%
HCV antibody
Yes	87%	89%	82%
No	13%	11%	18%
HIV
Negative	71%	67%	83%
Positive, RNA level < 400 copies/mL	20%	23%	9%
Positive, RNA level ≥ 400 copies/mL	9%	9%	8%

Note. Age tertiles are approximations after rounding. Behavioral characteristics (e.g., ER visit, injection and other drug use, overdose, homelessness, HIV status) reflect self-report data in the past six-months. Values may not add to 100 due to rounding. MOUD = medication treatment for opioid use disorder.