Use of telehealth for opioid use disorder treatment in safety net primary care settings: A mixed-methods study

Abstract

Introduction:

The COVID-19 pandemic resulted in a marked increase in telehealth for the provision of primary care-based opioid use disorder (OUD) treatment. This mixed methods study examines characteristics associated with having the majority of OUD-related visits via telehealth versus in-person, and changes in mode of delivery (in-person, telephone, video) over time.

Methods:

Logistic regression was performed using electronic health record data from patients with ≥1 visit with an OUD diagnosis to ≥1 of the two study clinics (Rural Health Clinic; urban Federally-Qualified Health Center) and ≥1 OUD medication ordered from 3/8/2020-9/1/2021, with >50% of OUD visits via telehealth (vs. >50% in-person) as the dependent variable and patient characteristics as independent variables. Changes in visit type over time were also examined. Inductive coding was used to analyze data from interviews with clinical team members (n=10) who provide OUD care to understand decision-making around visit type.

Results:

New patients (vs. returning; OR=0.47;95%CI:0.27-0.83), those with ≥1 psychiatric diagnosis (vs. none; OR=0.49,95%CI:0.29-0.82), and rural clinic patients (vs. urban; OR=0.05; 95%CI:0.03-0.08) had lower odds of having the majority of visits via telehealth than in-person. Patterns of visit type varied over time by clinic, with the majority of telehealth visits delivered via telephone. Team members described flexibility for patients as a key telehealth benefit, but described in-person visits as more conducive to building rapport with new patients and those with increased psychological burden.

Conclusion:

Understanding how and why telehealth is used for OUD treatment is critical for ensuring access to care and informing OUD-related policy decisions.

INTRODUCTION

Opioid use disorder (OUD) treatment has historically been delivered in-person. COVID-19 restrictions on in-person visits exposed the limitations of relying on one delivery mode for OUD care. To ensure treatment access during the pandemic, insurers, including many state Medicaid programs, reimbursed for both video and telephone-based telehealth visits at the same rate as in-person visits.¹ The government also allowed for remote buprenorphine induction and prescription refills if patients could be adequately evaluated and monitored via telehealth. The American Society of Addiction Medicine (ASAM) recommended that telehealth should be considered for OUD care patients stable in their recovery if care (medical follow-up, behavioral health and case management) could be delivered at the same level as in-person visits.² These changes allowed for comprehensive OUD treatment via telehealth and the opportunity to explore treatment patterns and real-world experiences related to care delivered via telehealth.

Since the COVID-19 pandemic began, studies have examined the impact of telehealth on retention on medications for OUD (MOUD), OUD treatment access, and factors influencing OUD treatment via telehealth.^3-8 While most studies identify benefits of telehealth for OUD treatment, some have highlighted disparities in use of telehealth for certain populaions.^9-11

Most prior studies were conducted early in the COVID-19 pandemic when limitations on in-person visits were in place and did not examine how the type of visit (in-person, telephone, video) for OUD care delivery has changed over time.³ More research is needed to fully understand the factors influencing OUD care via telehealth using quantitative and qualitative data.

To address this gap, we used quantitative data to examine patient characteristics associated with more telehealth versus in-person visits for OUD care and to examine the trends in OUD visit type (in-person, telephone, video) over the 18-months after initial COVID-19 restrictions. To inform these findings, we interviewed clinical team members to examine decision making related to modality of OUD care.

MATERIALS AND METHODS

Study Design

This was an iterative observational mixed methods study.

Setting

This study was conducted in a Rural Health Clinic¹² (RHC) and an urban Federally Qualified Health Center (FQHC) within one academic health system in the United States (US). The clinics’ OUD treatment program was adapted from the Family Health Center of Worcester model,¹³ and included interdisciplinary team-based care from primary care clinicians who prescribe MOUD [e.g., medical doctors (MDs), nurse practitioners (NPs), physician assistants (PAs)]; behavioral health clinicians (BHCs) (e.g., psychologists, social workers) who work with patients on improving coping skills, relapse prevention, and resilience strategies; and nurses (RNs) who care manage and engage patients in care. The program uses a tier-based approach, with decreasing visits and longer time between MOUD prescriptions refills in the highest tier for those furthest along in their recovery.¹⁴

This study was approved by the Oregon Health & Science University Institutional Review Board (IRB).

Study population

Electronic health record (EHR) data was extracted on patients who had 1) one or more visits to at least one of the two clinics between March 8, 2020 (COVID-19 state of emergency declaration in the clinics’ state) and September 1, 2021; 2) an OUD diagnosis in the problem or encounter lists during the study period identified using the International Classification of Diseases 10th Revision (ICD-10) codes F11.*); and 3) at least one MOUD order in the medication list during the study period, defined as a medication containing buprenorphine in sublingual film, tablet, or injectable suspension, or naltrexone in tablets or extended release intramuscular suspension.

To examine decision-making around visit type (i.e., in-person, telephone, video), we purposively selected a sample OUD care team members who varied by professional background (i.e., MD, NP, PA, RN, BHC) to participate in semi-structured interviews (n=10, 4 rural and 6 urban clinic team members. Everyone we contacted agreed to participate in an interview.

Variables

Patient characteristics.

Discrete data on characteristics often associated with OUD and treatment¹⁵ were extracted from the EHR (See Table 1) and included: sex, age, race/ethnicity by self-identification (using combined Uniform Data System race and ethnicity categories), health insurance coverage, smoking status as identified in a discrete data field in the vital signs or social history, and medical and psychiatric diagnoses identified via ICD-10 codes. Medical and psychiatric (including substance use disorders other than OUD) comorbidity variables were dichotomized (yes if ≥1 diagnosis; otherwise no). The location of the patient’s primary clinic (urban/rural) and whether the patient was identified as new to the clinic during the study period (yes/no) were also included as independent variables.

Table 1.

Patient characteristics and adjusted odds ratios of telehealth vs. in-person primary care visits for opioid use disorder (OUD) among patients with an OUD diagnosis and ≥1 medication prescription for OUD from 3/8/2020-08/31/2021 (N=781)

Characteristic	Total sample (N=781)	Equal in- person-to- telehealth visits (not in model) (n=40)	Majority (>50%) of PC visits for OUD In- person (n=149)	Majority (>50%) of PC visits for OUD telehealth (n=592)	Odds Ratio (95% CI) Ref (in-person) (n=741)
Sex
Female	445 (57.0)	19 (47.5)	77 (51.7)	349 (59.0)	Ref
Male	336 (43.0)	21 (52.5)	72 (48.3)	243 (41.1)	0.93 (0.60-1.46)
Age
18-29	135 (17.3)	<10 (17.5)	33 (22.2)	95 (16.1)	Ref
30-49	460 (58.9)	18 (45.0)	94 (63.1)	348 (58.8)	1.13 (0.64-1.99)
50 and older	186 (23.8)	15 (37.5)	22 (14.8)	149 (25.2)	1.81 (0.85-3.84)
Race/Ethnicity
NH White	573 (73.4)	31 (77.5)	112 (75.2)	430 (72.6)	Ref
Hispanic or not NH White		<10 (2.5)			0.84 (0.38-1.85)*
NH American Indian/Alaskan Native	28 (3.6)	0	<10 (4.0)	22 (3.7)
NH Asian or Pacific Islander	<10 (0.6)	0	<10 (1.3)	<10 (0.5)
NH Black	34 (4.4)	0	<10 (0.7)	33 (5.6)
Hispanic	23 (2.9)	0	<10 (2.0)	19 (3.2)
Unknown/Missing	118 (15.1)	<10 (20.0)	25 (16.8)	85 (14.4)	0.50 (0.27-0.93)
Clinic Location
Urban	503 (64.4)	20 (50.0)	24 (16.1)	459 (77.5)	Ref
Rural	278 (35.6)	20 (50.0)	125 (83.9)	133 (22.5)	0.05 (0.03-0.08)
Insurance
Private	166 (21.2)	10 (25.0)	26 (17.5)	130 (22.0)	Ref
Public (Medicaid, Medicare, Tricare)	592 (75.8)	28 (70.0)	121 (81.2)	443 (74.8)	0.85 (0.48-1.51)
Self-pay/Unknown	23 (2.9)	<10 (5.0)	<10 (1.3)	19 (3.2)	0.69 (0.11-4.04)
Current smoking
No	312 (40.0)	19 (47.5)	58 (38.9)	235 (39.7)	Ref
Yes	469 (60.1)	21 (52.5)	91 (61.1)	357 (60.3)	0.89 (0.57-1.40)
≥1 medical diagnosis ±
No	438 (56.1)	25 (62.5)	94 (63.1)	319 (53.9)	Ref
Yes	343 (43.9)	15 (37.5)	55 (36.9)	273 (46.1)	1.00 (0.59-1.70)
≥1 psychiatric diagnosis ‡
No	290 (37.1)	19 (47.5)	56 (37.6)	215 (36.3)	Ref
Yes	491 (62.9)	21 (52.5)	93 (62.4)	377 (63.7)	0.49 (0.29-0.82)
New patient
No	679 (86.9)	38 (95.0)	111 (74.5)	530 (89.5)	Ref
Yes	102 (13.1)	<10 (5.0)	38 (25.5)	62 (10.5)	0.47 (0.27-0.83)

Note. OUD=Opioid use disorder, NH=Non-Hispanic, CI=confidence interval; Ref=reference

^*Categories other than non-Hispanic White and Unknown/Missing were collapsed in adjusted analysis to account for the low numbers of racial/ethnic minority patients

^±Includes chronic viral hepatitis C, HIV, chronic pain, fibromyalgia, migraine, low backpain, temporomandibular joint disorder, chronic fatigue syndrome, liver disease, cardiovascular diseases, diabetes, hypertension, COVID-related diagnoses, insomnia/sleep disorders, sexually transmitted diseases, non-skin malignancy

^‡Includes substance use disorders (other than OUD), depressive disorders, anxiety disorders, bipolar disorder, post-traumatic stress disorder, schizophrenia, schizoaffective disorder, attention deficit hyperactivity disorder

OUD care visit type.

We extracted all patient visits (e.g., medical, behavioral health) with an OUD diagnosis noted in the visit encounter list and then categorized as in-person, telephone, or video, using a combination of scheduling and level of service (LOS) codes for services rendered in cases where video visits were converted to telephone visits. We could not distinguish the services rendered at each visit.

Data Collection

Quantitative data were extracted by the health system data team, who masked identifiers and transferred data via secure email to the research team.

Interviews with clinical team members followed a semi-structured guide (see Supplemental Material) and were conducted by experienced qualitative researchers. Interviewers explored clinical staff members’ attitudes, beliefs and experiences delivering OUD treatment using telehealth, as well as the factors that influence care modality decisions. After a series of open-ended questions, we asked more directed questions to examine the patient, organizational, and contextual factors that influence choice of visit type for OUD care. The interview guide was pilot tested and refined to improve question format and flow.

Interviews were conducted virtually between December 2021-April 2022 and were approximately 60 minutes long. Audio-recordings were professionally transcribed, reviewed for accuracy and de-identified.

Analysis

Descriptive statistics were calculated for the study sample (i.e., patients with an OUD diagnosis and ≥1 MOUD order during the study period), overall and by visit type. Race categories other than non-Hispanic white were collapsed in adjusted analysis to account for low numbers of racial/ethnic minority patients. To examine characteristics associated with visit type, logistic regression was performed with >50% of OUD visits via telehealth (either telephone or video) vs. >50% of OUD visits in-person as the dependent variable. Due to low numbers of video visits, telephone and video were combined into a ‘telehealth’ category within our modeling. Data from patients who had an equal number of telehealth and in-person visits are included descriptively but excluded from the adjusted analysis (n=40 patients). To determine if characteristics associated with majority telehealth vs. in-person visits would differ if majority were defined as a more pronounced percent difference in visits, we conducted the same analysis but with the cut-point of ≥60% as a sensitivity analysis. Statistical analyses were conducted in 2022 using RStudio, version 2022.02.0+443, and Stata, version 15.1. All statistical tests were two-sided and significance was defined as p<0.05.

Qualitative transcripts were entered into ATLAS.ti Version 9 (Scientific Software Development GmBH, 2021) for data management and analysis. Using an inductive approach,¹⁶ three team members (SB, TWL, SW) listened to and analyzed two interviews, tagging text with “codes” that named the emerging insights and defined the code for future use. Using this code list, two additional interviews were analyzed by TWL and SW independently, who then compared and discussed their analyses, and refined the codebook, as needed. When analyst consensus was reached, each analyst independently analyzed a portion of the remaining data. The team then shifted to conducting comparative analyses at the clinic and clinician levels, which involved examining code output related to visit and telehealth modality decision making. Preliminary findings were examined side-by-side with quantitative data showing differences in frequency of telephone vs. video visits over the study period, as well as patient characteristics associated with more telehealth vs. in-person visits. The full team discussed these data to identify clinic and clinic team member experiences and decision-making related to visit type and telehealth use.

RESULTS

EHR study sample characteristics

Of patients seen during the study period (N=26,289), 4.6% (n=1,204) had an OUD diagnosis; 5.4% in the FQHC (n=748 of 13,843 patients) and 3.7% in the RHC (n=456 of 12,446 patients). Our study sample consisted of the 781 patients with an OUD diagnosis who had ≥1 MOUD prescription (64.9% of patients with an OUD diagnosis). Of those with ≥1 MOUD order, 57.0% were female, 59.6% were 30-49 years of age, 73.1% identified as non-Hispanic white, 87.8% were publicly insured, 73.3% had ≥1 psychiatric diagnosis, 38.9% had ≥1 medical comorbidity, and over half had a status indicating current smoking (see Table 1 for study sample characteristics). Patients had a range of 1-63 visits with a median of 11 visits (interquartile range=13).

In the sensitivity analysis which defined majority of visits as ≥60%, results were similar (see Appendix Table 1). Given the need to exclude 104 people from the sensitivity analysis because they had no visit type majority (e.g., those with relatively equal visit ratios of 41:59, 45:55, etc.), we report findings from the regression performed with >50% of OUD visits via telehealth.

Delivery of OUD care by visit type (in-person, telephone, video) over time

There were 10,642 total OUD visits by patients in our sample. While the most frequent type of visit for OUD care delivery varied over time, with in-person and telehealth variation more pronounced at the rural clinic, over 75% (n=592) of patients in our study sample had >50% of all OUD visits conducted via telehealth. The majority of telehealth visits were delivered via telephone (Figure 1).

Figure 1. Monthly visits for opioid use disorder-related care at each study clinic by mode and modality.

Note: FQHC: Federally Qualified Health Center

Clinical teams described an immediate transition to telehealth at the beginning of the pandemic. Once restrictions began to be lifted, clinic team members may have had a preference for visit type, but neither clinic had a protocol for who should be seen in-person vs. via telehealth, nor did they have a way to track patient visit preferences. Medication could be prescribed via telehealth or in-person and neither clinic withheld medication due to a patient not coming in for an in-person appointment. They did report observing some loss to follow-up when visits were primarily limited to telehealth only.

We were trying to get people to reconnect because we had noticed we had had a significant number of people who were lost to follow up during 2020, when we were trying to promote only virtual visits… I think it's more of on a case-by-case basis where, if we find that virtual is not working, we'll ask somebody if they can be coming in for their visits.

[D2; Clinician]

To address this, and as in-person restrictions were lifted, clinics conducted outreach and blended in-person and telehealth visits, which parallel EHR findings (Figure 1).

Compared to in-person visits, noted benefits of telehealth included decreased disruption to patients’ daily routines and increased flexibility, as coming into the clinic on a regular basis for OUD-related care could be burdensome: “…Getting to the clinic, taking four or five buses to get to your visit, and then having to go to the pharmacy. It’s hard. It’s difficult for them. Them being able to just do it on the phone or virtually, it knocked down a lot of barriers.” [D4; Clinician]

Characteristics associated with delivery of OUD care by visit type (in-person vs telehealth)

We found associations between majority telehealth vs. in-person visits and the following characteristics: new vs. returning patients, patients with ≥1 psychiatric diagnosis vs. none, and location of patient’s clinic (see Table 1).

Newer patients were less likely to have the majority of their visits via telehealth as compared to returning patients (OR=0.47, 95%CI:0.27-0.83). Clinical staff at both study clinics stated that building rapport with new patients was especially important, and this was more difficult to do via telehealth than in-person, though not impossible.

Our people who are really unstable as part of our—especially in the early phases of our program, those are the ones we're trying to prioritize, keeping them coming in face-to-face, building those relationships, and staying engaged with care.

[D8; BHC]

As patients became more established, telehealth modalities were used more. As one team member explained, “The better I know them, the better the video visit works. The more that I have human connections established, the more we can use humor, prior context, check-in… “ [D7; MD]

Compared to patients with no psychiatric diagnosis, patients with ≥1 psychiatric diagnosis had lower odds (OR=0.49, 95%CI:0.29-0.82) of having the majority of their visits via telehealth than in-person. Clinical interviewees did not explicitly address this finding, but noted that among those who need more support, there is a need for in-person visits that foster whole-person treatment.

We found that patients in the rural clinic (vs. urban FQHC; OR=0.05, 95%CI:0.03-0.08) had lower odds of having the majority of their visits via telehealth vs in-person compared to patients in the urban clinic. Figure 1 shows this difference in visit type patterns at the clinic level over time by total number of visits each month. Within the urban clinic, telehealth was the primary visit mode over most of the study period, with few in-person visits. In the RHC, both in-person and telehealth (primarily telephone) visits were common, with telehealth used more often in the first year of the study period and more in-person visits in the latter part of the study period. Rural clinic team members reported barriers to telehealth (e.g., inconsistent access to internet and/or phone service, and lack of technological literacy).

We've gone through periods of fluctuation along with COVID as restrictions have gone up and down. Whenever restrictions have dropped, we try to get as many face-to-face as possible […] We don't have a good telecommunications infrastructure out here to do virtual visits. Even with that, sometimes telephone visits aren't feasible for my patient population.

[D8, BHC]

And, they noted the ease of in-person visits (“…for a lot of my patients, it’s actually easier just to come to clinic…” [D10; Clinician]) especially for those that live close to the clinic.

Delivery of OUD care by telehealth visit type (telephone vs video)

Between the options of telephone or video, telephone visits were the primary telehealth delivery mode (see Figure 1). While clinic team members preferred video visits because they are a richer, more personal modality, telephone visits were easier for patients and seemed as effective. Several barriers to video visits contributed to the significantly higher rate of telephone visits.

Video's better than telephone. I think I have been able to successfully have five video appointments [laughs] since COVID started…because of barriers. Some of that is education literacy, patients figuring out, "How do I even use MyChart?" [web-based patient portal]. I have a number of patients who still haven't been able to figure out how to access MyChart. That doesn't even include internet complications and all that stuff.

[D8; BHC]

Patient environment and preference also informed these choices and, according to clinical teams, patients most often chose to have telephone visits.

How do we pick between phone or video? We typically leave it up to the patient. A lot of our patients are marginally housed or don't have a lot of privacy when they are housed. It's much easier to have a quiet phone conversation than have a video chat on their phone. And some patients thrive on it…That's mostly tech and patient preference.

[D1; RN]

DISCUSSION

New (vs. established) patients, those with ≥1 psychiatric diagnosis (vs. none) and those seen in a rural (vs. urban) clinic were less likely to have the majority of their visits via telehealth than in-person. Patterns of visit type over time revealed continuation of telehealth visits after in-person restrictions were lifted but that, toward the end of our study period, in-person visits were more common in the rural clinic. Qualitative data suggest that decisions regarding visit type are often dependent on the patient and their circumstances, which included how new the patient was to the clinic, how much support the patient needed, technological savviness and patient preference.

Consistent with ASAM recommendations that telehealth be offered to patients considered stable in their recovery, new patients were significantly less likely to have the majority of their visits via telehealth. Similar to sentiments of clinical team members in our study, interviews from other studies with OUD care teams reported challenges in connecting and establishing rapport with new patients¹⁷ and not being as able to ascertain certain information (both tangible and intangible) via telehealth visits, especially with patients for whom they were less familiar.¹⁸

We also found that patients with ≥1 psychiatric diagnosis were less likely to have the majority of their visits via telehealth. This is interesting given telehealth has been used to provide care for mental health conditions pre-COVID-19 and is effective.^19-21 One reason identified in our study for a preference for in-person visits was the intangible benefits gained from face-to-face sessions, including being more conducive to providing more whole-person, supportive care compared to telehealth. This resonates with a study among Veterans Administration providers who highlighted the importance of in-person visits for veterans with OUD especially if they have comorbid mental health conditions.²²

An often-cited advantage of telehealth is that it can reach rural patients who often face barriers to accessing care (transportation, distance to clinic, shortage of MOUD prescribers)^3,23 We found that patients seen in the rural clinic were less likely to have the majority of visits via telehealth compared with those seen in the urban clinic. Findings from studies that examined rural/urban differences in mode of delivery of OUD-related care are inconsistent, with some reporting urban/rural differences similar to what we found,^10,24 and others reporting no signficant differences.^9,25 Our qualitative data suggest that fewer telehealth visits among this population are due to a combination of patient and clinic preference and limited telecommunications infrastructure. It should be noted that the differences between clinics were not due solely to factors associated with clinic location. Variation in COVID-related protocols for each clinic (e.g., limits on number of persons allowed in clinic; resources available to support telehealth) likely contributed to these differences as well. Future research, including the patient perspective, is needed to understand the relationships between rurality and telehealth use.

In both clinics, the overwhelming majority of telehealth visits were conducted by telephone throughout most of the study period. As noted by our clinical team members and those from other studies that have examined use of telephone vs video for health care visits,^26,27 reasons for higher rates of telephone visits included technical challenges with video visits such as connection issues and lack of access to, or experience using, video-capable devices. It was simply easier and sometimes more private to have a phone visit. This difference in rates of telehealth modality is important given that real-time video, but not telephone visits, have historically been reimbursed by many state Medicaid programs.¹ While many states have implemented payment parity policies requiring telehealth visits, including audio-only visits, to be reimbursed at the same rate as in-person visits during the pandemic, the future state of these policies beyond the COVID-19 public health emergency is unknown. Importantly, disparities in use of telephone versus video for OUD care have also been reported. One VA study found that among patients who used telehealth for accessing MOUD, those who were older, male, Black, non–service connected, or experiencing homelessness and/or housing instability were less likely to have video visits.⁹ Given the high rate of telephone visits in our study and in others,^9,25 as well as differences in use by demographics, larger studies are warranted to assess the effectiveness of this delivery modality for OUD treatment delivery and the patient characteristics associated with its utilization.

Limitations

This study has some limitations. This study included two family medicine clinics with established OUD treatment programs and that primarily see either rural patients and/or those of lower SES; thus, our findings may not generalize to other primary care settings. Further, our sample was primarily non-Hispanic White. Given the need to collapse all categories other than non-Hispanic White into one category for the adjusted analysis, we were unable to fully assess racial/ethnic disparities in telehealth use. The sample was also too small to quantitatively address adjusted differences stratified by clinic. Additionally, while we found differences in visit delivery type between our rural- and urban-based clinics and our qualitative interviews provided insight into potential barriers to telehealth due to rurality, patterns of care in these clinics also could have been a result of differences in clinic-specific protocols related to care throughout the COVID-19 public health emergency. We also note that urine drug screenings (UDS) were not routinely performed during the study period, but if one was requested, this did require an in-person visit. We were unable to examine OUD care delivery modes by medication management versus behavioral health visits. Some studies (not specific to OUD care) found differences in rates of telehealth and in-person visits by type of visit (e.g., primary care, behavioral health, speciality care);^28-30 thus, this should be explored in future studies focused on OUD care, which commonly includes both medical and behavioral health visits. Finally, and importantly, our qualitative interviews did not capture patients’ experiences and perceptions of OUD care delivered via telehealth, nor were we able to assess associations of delivery mode with patient level outcomes. Thus, future studies should be designed to understand the impact of having the majority of visits via telehealth on patient outcomes such as treatment retention or overdose.

Conclusions

To date, literature suggests that OUD care delivered via telehealth is at least as effective as care provided in-person.³¹ Given that OUD treatment is associated with reduced opioid overdose risk,^32-34 it is important that options for delivery of OUD care are convenient and accessible to persons seeking treatment. We found that certain characteristics are associated with use of primarily telehealth vs. in-person visits among patients receiving treatment for an OUD. Continued research is needed to further our understanding of how post-COVID-19 OUD treatment is delivered in primary care clinics that care for underserved populations (e.g., rural, lower socioeconomic status) and its impact on long-term OUD treatment outcomes. These findings are important for informing OUD-related policy decisions and for ensuring access to vital care for all patients.

Appendix Table 1. Patient characteristics and adjusted odds ratios of telehealth vs. in-person primary care visits for opioid use disorder (OUD) among patients with an OUD diagnosis and ≥1 medication prescription for OUD from 3/8/2020-08/31/2021 (N=781)

Characteristic	Total sample (N=781)	No Majority (ratio of in-person- to-telehealth visits relatively equal; not in model) (N=104)	Majority (>60%) of PC visits for OUD In-person (N=131)	Majority (>60%) of PC visits for OUD telehealth (N=546)	Odds Ratio (95% CI) Ref (in-person) (N=677)
Sex
Female	445 (57.0)	55 (52.9)	68 (51.9)	322 (59.0)	Ref
Male	336 (43.0)	49 (47.1)	63 (48.1)	224 (41.0)	1.04 (0.63-1.72)
Age
18-29	135 (17.3)	24 (23.1)	26 (19.9)	85 (15.7)	Ref
30-49	460 (58.9)	52 (50.0)	86 (65.7)	322 (59.0)	0.96 (0.50-1.84)
50 and older	186 (23.8)	28 (26.9)	19 (14.5)	139 (25.5)	1.60 (0.68-3.76)
Race/Ethnicity
NH White	573 (73.4)	81 (77.9)	98 (74.8)	394 (72.2)	Ref
Hispanic or not NH White					1.00 (0.41-2.43)*
NH American Indian/Alaskan Native	28 (3.6)	<10 (2.9)	<10 (3.1)	21 (3.9)
NH Asian or Pacific Islander	<10 (0.6)	<10 (1.0)	<10 (1.5)	<10 (0.4)
NH Black	34 (4.4)	<10 (1.9)	<10 (0.8)	31 (5.7)
Hispanic	23 (2.9)	<10 (1.0)	<10 (2.3)	19 (3.5)
Unknown/Missing	118 (15.1)	16 (15.4)	23 (17.6)	79 (14.5)	0.54 (0.28-1.05)
Clinic Location
Urban	503 (64.4)	42 (40.4)	21 (16.0)	440 (80.6)	Ref
Rural	278 (35.6)	62 (59.6)	110 (84.0)	106 (19.4)	0.04 (0.02-0.07)
Insurance
Private insurance	166 (21.3)	27 (26.0)	20 (15.3)	119 (21.8)	Ref
Public (Medicaid/Medicare/Tricare)	592 (75.8)	75 (72.1)	109 (83.2)	408 (74.7)	0.86 (0.44-1.67)
Self-pay/Unknown	23 (2.9)	<10 (1.9)	<10 (1.5)	19 (3.4)	0.70 (0.11-4.37)
Current smoking
No	312 (40.0)	50 (48.1)	48 (36.6)	214 (39.2)	Ref
Yes	469 (60.1)	54 (51.9)	83 (63.4)	332 (60.8)	0.74 (0.44-1.22)
≥1 medical diagnosis ±
No	438 (56.1)	67 (64.4)	82 (62.6)	289 (52.9)	Ref
Yes	343 (43.9)	37 (35.6)	49 (37.4)	257 (47.1)	0.97 (0.54-1.74)
≥1 psychiatric diagnosis ‡
No	290 (37.1)	47 (45.2)	49 (37.4)	194 (35.5)	Ref
Yes	491 (62.9)	57 (54.8)	82 (62.6)	353 (64.5)	0.46 (0.25-0.82)
New patient
No	679 (86.9)	89 (85.6)	94 (71.8)	496 (90.8)	Ref
Yes	104 (13.1)	15 (14.4)	37 (28.2)	50 (9.2)	0.28 (0.15-0.54)

Note. OUD=opioid use disorder; NH=Non-Hispanic, CI=confidence interval; Ref=reference

^*Categories other than non-Hispanic White and Unknown/Missing were collapsed in adjusted analysis to account for the low numbers of racial/ethnic minority patients

^±Includes chronic viral hepatitis C, HIV, chronic pain, fibromyalgia, migraine, low backpain, temporomandibular joint disorder, chronic fatigue syndrome, liver disease, cardiovascular diseases, diabetes, hypertension, COVID-related diagnoses, insomnia/sleep disorders, sexually transmitted diseases, non-skin malignancy

^‡Includes substance use disorders (other than OUD), depressive disorders, anxiety disorders, bipolar disorder, post-traumatic stress disorder, schizophrenia, schizoaffective disorder, attention deficit hyperactivity disorder