Retention of Patients With Multiple Vulnerabilities in a Federally Qualified Health Center Buprenorphine Program: Pennsylvania, 2017–2018

Abstract

Objectives. To describe and report initial outcomes of a low-threshold, group-based primary care medication for opioid use disorder (OUD) program in a federally qualified health center.

Methods. We performed a retrospective chart review of patients enrolled in the program from October 4, 2017, to October 3, 2018, in Philadelphia, Pennsylvania. The main outcome measure was time retained in treatment, defined as time from treatment initiation to unplanned treatment termination. Secondary outcomes were the relationships between treatment retention and cocaine use or housing status. We analyzed retention in treatment using Kaplan-Meier survival estimates.

Results. The 3- and 6-month retention rates were 82% and 63%, respectively. The log-rank test showed no significant differences for comparisons between homeless versus not homeless (P = .25) and cocaine use versus no cocaine use (P = .12).

Conclusions. The medication for OUD program engaged a large number of patients from marginalized groups. Three- and 6-month retention rates were comparable with those reported of other federally qualified health center populations.

Public Health Implications. Integrating treatment of OUD into primary care shows promise for increasing access to and retention in medication for OUD services. The federally qualified health center payment structure supports the sustainability of the group visit model.

Seventy thousand people in Philadelphia, Pennsylvania, use heroin, and another 50 000 misuse prescription opioids.¹ Philadelphia is disproportionately affected with an opioid death rate (1100 estimated in 2018) triple the city’s murder rate and double the national overdose death rate (44.3 per 100 000).² Philadelphia has increased its capacity for medication for opioid use disorder (MOUD; formerly known as medication-assisted treatment [MAT]) with buprenorphine from 100 treatment slots for patients in 2017 to 2906 as of December 2018, but clearly there is still a need for thousands of more treatment slots.³ Crucially, capacity does not imply that treatment is available for everyone who needs it, because there are still logistical, administrative, and stigma-related barriers to accessing treatment for some.⁴ Additionally, recent evaluations reveal that people from vulnerable populations, including individuals experiencing homelessness, non-White persons, and those with chronic medical diseases, have even more difficulty accessing MOUD.^5,6

Several programs across the United States are working to reverse these disparities through a low-threshold approach to treatment. Kourounis et al.⁷ define treatment threshold as barriers that patients face when initially accessing MOUD and during ongoing treatment. Treatment barriers include accessibility barriers (e.g., waiting lists, inflexible admission criteria, limited points of access, and office-based induction) and design treatment barriers (such as a 1-size-fits-all approach and zero tolerance for relapse or polysubstance use). In 2019, building on earlier work in this area, including methadone programs,^7–9 Jakubowski and Fox⁸ described 4 main criteria for treatment programs to be considered low threshold: same-day treatment entry, a harm-reduction approach, flexibility, and availability in nontraditional settings.

Although Philadelphia has the notorious distinction of being an epicenter of the opioid crisis, coordination at the local and state levels has provided a supportive environment for creating and evaluating innovative responses. Using a low-threshold approach, we developed a MOUD program to offer immediate and ongoing access to patient-centered OUD treatment. The program is embedded in holistic health care services and designed explicitly for marginalized populations. The 3 main objectives of our program are (1) increase access to MOUD of marginalized communities, (2) increase treatment retention of patients with multiple comorbidities, and (3) increase access to primary care of people with OUD. Therefore, we hypothesized that our model would increase marginalized populations’ access to and retention in MOUD and increase access to primary care of people with OUD. Figure 1 presents a logic model for our program showing the relationship among harm-reduction principles, program philosophy and structure, and anticipated outcomes.

FIGURE 1—

Relationship of Harm-Reduction Principles to Program Philosophy and Structure

METHODS

The Stephen Klein Wellness Center, a federally qualified health center (FQHC), provided data for this longitudinal retrospective study. The data set consisted of the records of 79 participants from the first year of a group-based MAT program for people with opioid use disorder from October 4, 2017, through October 3, 2018.

Intervention and Program Description

In response to the opioid crisis, 2 medical providers in partnership with several behavioral health consultants at the FQHC began a pilot individual appointment–based MOUD program in 2016. Within 3 months and without external advertising, demand outpaced availability for consistent access to individual appointments. Additionally, on January 1, 2017, we began receiving block funding for OUD treatment from the Pennsylvania Department of Human Services Center of Excellence. Requirements for participation in this program influenced the evolution of our care model. Selected relevant requirements were the following: (1) develop and deploy a community-based care management team that consisted of an MAT care coordinator and licensed and unlicensed professionals to support people in accessing treatment, (2) perform a standard Level of Care risk and needs assessment on all new participants, (3) refer participants to monthly mental health and drug and alcohol treatment, and (4) report participants’ monthly urine drug screens (UDS).

We reviewed several group models for MOUD services, including nurse-led care,¹⁰ shared medical appointments,¹¹ and behavioral health–led psychoeducational and psychosocial support¹² with individual medical provider contact. Incorporating some practices from each, and addressing the requirements for participation in the Pennsylvania Department of Human Services Center of Excellence, we primarily adopted the behavioral health–led group session model with brief individual medical provider check-in visits during the session. We determined target group size for patient access, provider sustainability, and fiscal responsibility to be 8 to 12 patients per group. Team members included an MAT care coordinator (who also provides the front office functions of scheduling appointments, verifying insurance, checking patients in, and supporting general case management needs), a licensed behavioral health provider, a waivered medical provider, and a medical assistant (to collect urine drug screens, both for point of care testing and off-site laboratory testing; responsibilities later expanded to other services, such as providing vaccines and laboratory draws).

Patients initially accessed MOUD services as walk-ins or via telephone. The MAT care coordinator performed a standard level of care risk and needs assessment. We offered patients for whom a higher level of care was appropriate assistance in obtaining this care, but they were not required to if they were medically stable with no immediate life-threatening problems. We scheduled initial buprenorphine intake appointments within 48 hours, with some walk-in availability.

During the initial appointment, patients met with 3 staff members: (1) the MAT care coordinator to assess substance use history and identify barriers to care, (2) an X-waivered primary care provider (one who has special Drug Enforcement Administration registration to prescribe buprenorphine products) to assess physical health needs and treatment options, and (3) a behavioral health consultant (licensed behavioral health provider) for a focused psychosocial and mental health assessment. Unless contraindicated by medical concerns, we provided patients with education and support for home induction of buprenorphine. Patients have access to an on-call provider, and we confirmed their next expected visit at group sessions. Team members were allocated 1 clinic session for MOUD group visits that included pregroup huddle to review the most recent UDS, to update the team about any relevant changes in patient status or needs, and to review the prescribing plan. A 1-hour group psychoeducational session followed, and then there was a concluding postgroup team huddle to review any departures from plans.

A team of MAT prescribing providers, behavioral health consultants, care coordinators, and certified recovery specialists developed this model to streamline the MOUD process for participants, providers, and staff while providing a low-threshold approach to care and maintaining flexibility. At the Stephen Klein Wellness Center site, after the initial individual intake, MOUD access is now solely through the group program. At the time of this evaluation, we held 2 weekly walk-in MOUD groups at this site; we have since expanded this to 4 days per week. This structure obviates the need for patients to make appointments, significantly reducing the administrative burden of patients and staff to schedule, cancel, and reschedule multiple appointments in person or by telephone. Prescriptions are generally written on a weekly basis, with extensions to 2-, 3-, and 4-week intervals based on clinical stability. At the time of the evaluation, we defined clinical stability as serial UDSs with consistent amounts of buprenorphine and no illegal or unprescribed substances.

The ongoing use of opioids was accepted as part of the recovery process, as long as appropriate levels of buprenorphine and norbuprenorphine were present in the UDS. In the case of UDS that was negative for buprenorphine, patients were offered a “split script,” meaning patients received a smaller amount of buprenorphine and had to return to the pharmacy later in the week to pick up the rest of the weekly prescription. In the case of multiple UDSs that were negative for buprenorphine, patients were offered the option of daily observed dosing in the pharmacy. Ongoing use of other substances, primarily cocaine, was tolerated, although patients were kept at weekly visits so that we could offer support for stimulant use disorder. Patients with multiple UDSs that were positive for benzodiazepines were referred for additional onsite behavioral health support and treatment of anxiety. Multiple access points per week allow us to hold firm boundaries related to timely arrival (with a 5-minute grace period). Patients who arrived after the group session had started or who missed the group session were given “bridge” prescriptions until the next group session; this was modified on a case-by-case basis.

We are able to offer onsite childcare through our colocation with a YMCA. The MAT care coordinator arranges transportation for Medicaid patients by purchasing transit passes for those who are otherwise ineligible for Medicaid transportation. Participants’ access to primary care services is the same as that of the rest of the clinic population: we use a same day–next day model whereby appointments can be accessed via telephone or as a walk-in.

The FQHC model and Pennsylvania Medicaid payment model support the sustainability of the group visit model. We bill the same fixed FQHC visit rate for each medical and behavioral health group visit for patients with Medicaid; Medicare does not pay for the group delivery method but does reimburse for the individual medical visit. At the time of this evaluation, the Pennsylvania Department of Human Services Center of Excellence funding covered the salary and benefits of the MOUD coordinator. We performed a basic cost–revenue analysis by examining OUD billing codes during the study period. This analysis showed that the MOUD program funds 92% of its costs. Pharmacy revenue and Health Resources and Services Administration funds covered the balance. This cost–revenue ratio is similar for the program as a whole.

Data Sources

We derived our data set from the FQHC electronic health record, Next Gen, which provided sociodemographic information as well as information on MOUD group visits and primary care visits (defined as any non-MOUD billable encounter with a primary care provider, i.e., nonbehavioral health, nondental provider), MOUD visit notes, UDS results, medical and psychiatric diagnoses, HIV testing results, hepatitis C virus testing results, and whether patients were classified as individuals experiencing homelessness. We determined whether participants were experiencing homelessness using the Health Resource and Service Administration definition for the FQHC Uniform Data Service Reporting; this includes people living in the following circumstances: on the street, in a shelter, in transitional housing, doubled-up, in a single room occupancy hotel, in permanent supportive housing, and in other housing programs that are targeted to individuals experiencing homelessness.¹³

Inclusion and Exclusion Criteria

We initially included 82 participants who entered the MOUD program after October 4, 2017, and before October 3, 2018. We then excluded from the analysis the 3 participants with only 1 MOUD group visit, representing 3.7% of the sample. After excluding the 3 participants with only 1 MOUD group visit, our final data set consisted of 79 individuals.

Our primary outcome was time retained in treatment, defined as the time from treatment initiation to first unplanned treatment termination. Following previous studies of MAT outcomes, we defined unplanned treatment termination as unplanned self-discharge from the program (i.e., no active prescription and no appointments for 60 days) or drug overdose death.^14–16 We did not consider patients leaving the program because of incarceration, planned transfer to another program, or death unrelated to drug overdose as unplanned treatment termination. Our secondary outcomes were the relationships between (1) cocaine use in the first month of program admission and time retained in treatment, and (2) housing status as reported at intake and time retained in treatment.

Statistical Methods

We calculated demographic variables using frequencies for categorical variables and means and SDs for continuous variables. We examined bivariate relationships between all study variables using the χ² test for categorical variables, Pearson correlation coefficients for continuous variables, and an independent-samples t test for categorical–continuous relationships, with a significance level of P less than .05.

For our primary outcome of time retained in treatment, we calculated retention in treatment using Kaplan-Meier survival analysis. We defined survival time as months retained in treatment. We removed patients not experiencing treatment failure (i.e., incarceration, planned transfer to another program, or death unrelated to buprenorphine or drug use) from the survival analysis (or censored them) on the date that the outcome occurred. We then created separate survival curves for participants who currently or formerly experienced homelessness versus those not experiencing homelessness and for participants whose UDS was positive for cocaine in days 2 through 31 of the program versus those whose UDS was negative for cocaine. We conducted the log-rank test to assess differences in median months retained in treatment of individuals currently or formerly experiencing homelessness versus individuals not experiencing homelessness and for participants with positive versus negative cocaine UDS in the first month, with a significance level of P less than .05.

We conducted all analysis using SPSS version 26 (IBM, Somers, NY).

RESULTS

Table 1 provides demographic information for participants. The average age of the participants (n = 79) was 47 years (SD = 10.0); 59% were male, and 58% were African American. Forty-three percent of participants were individuals currently or formerly experiencing homelessness. Thirty-four percent of participants had UDSs that were positive for cocaine in days 2 through 31 of the program. Top medical and psychiatric comorbidities coded in the electronic health record included nicotine dependence (57%), hypertension (30%), chronic hepatitis C virus (24%), anxiety (22%), major depressive disorder (13%), and posttraumatic stress disorder (11%). Insurance status included 81% using Medicaid, 11% using Medicare, and 8% uninsured. The majority of participants (56%) had 2 or more primary care visits over the time of the evaluation. No bivariate relationships between demographic variables and retention in treatment were statistically significant. Forty-eight patients experienced an unplanned treatment termination event, including 1 overdose death. Seven patients left the program for other reasons, including 5 planned transfers, 3 incarcerations, 1 hospitalization, and 1 death unrelated to opioid use; we did not categorize these 7 as experiencing an unplanned treatment termination event.

TABLE 1—

Demographics and Clinical Characteristics of Patients: Stephen Klein Wellness Center, Philadelphia, PA, October 2017–October 2018

Characteristic	No (%) or Mean ±SD
Age, y	46.6 ±10
Sex
Male	47 (59)
Female	32 (41)
Race
Black or African American	46 (58)
White	29 (37)
Declined to specify	4 (5)
Screened for HCV	28 (35)
Chronic HCV	19 (24)
Screened for HIV	53 (67)
Nicotine dependence	45 (57)
Hypertension	24 (30)
Psychiatric diagnoses
Anxiety	17 (22)
Major depressive disorder	10 (13)
Posttraumatic stress disorder	9 (11)
Clinic zip code catchment area	35 (44)
PCP visits
0	18 (20)
1	17 (24)
≥ 2	44 (56)
Baseline cocaine use	27 (34)
Individuals currently or formerly experiencing homelessness	34 (43)

Note. HCV = hepatitis C virus; PCP = primary care provider. Sample size was n = 79.

Kaplan-Meier analysis showed an overall 3-month retention rate of 82% and a 6-month retention rate of 63%. Months retained in treatment of individuals experiencing versus not experiencing homelessness were similar at 10.45 versus 10.75 months, respectively. The log-rank test for individuals experiencing versus not experiencing homelessness was not significant (P = .25). Months retained in treatment of cocaine use versus no cocaine use were lower at 5.33 versus 11.54 months, respectively. The log-rank test for cocaine use versus no cocaine use was not significant (P = .12; Figures 2 and 3).

FIGURE 2—

Cumulative Medication Assisted Treatment Retention by Housing Status: Pennsylvania, PA, October 2017–October 2018

Note. Censored individuals, defined as those with unplanned treatment termination, are represented with tick marks.

FIGURE 3—

Cumulative Medication Assisted Treatment Retention by Baseline Cocaine Use: Pennsylvania, PA, October 2017–October 2018

Note. Censored individuals, defined as those with unplanned treatment termination, are represented with tick marks.

DISCUSSION

Based on our initial findings, our program shows promise in addressing its main objectives: to (1) increase access to MOUD of marginalized populations, (2) increase treatment retention of patients with multiple comorbidities, and (3) increase access to primary care of people with OUD. Our MOUD program engaged a larger percentage of African American patients and patients 35 years and older compared with Philadelphia MOUD programs citywide (57% vs 29% and 87.34% vs 65.11%, respectively).¹⁷

Although exact comparisons cannot be made because of wide variation in the definition of treatment retention, our retention rates at 3 months (82%) and 6 months (63%) are comparable with those reported in other FQHC populations and in programs specifically focused on highly marginalized populations. Haddad et al.¹⁸ examined treatment retention in a MOUD program in 2 urban FQHC network sites in Connecticut; the majority of the management team members of this program were primary care providers. The authors reported retention in MOUD at 3 and 6 months as 71.8% and 56.8%, respectively, in a cohort of 266 people. Importantly, this study classified participants who returned to treatment after a treatment gap as retained in treatment. Sixty-seven percent of participants in this cohort experienced a treatment gap, with a mean gap of 116 days.¹⁸ A study of an innovative Philadelphia MOUD program through Prevention Point Philadelphia reported retention rates at 3 months (77%) and 6 months (65%), similar to our program.¹⁶ Similar to our approach, this study defined treatment failure as unplanned self-discharge from the program, no active prescription, and failure to attend scheduled appointments for 60 days, and it did not include as treatment failures leaving the program because of incarceration, a planned transfer to another program, or death unrelated to buprenorphine or other drugs.¹⁶

Few studies have examined retention in MOUD specifically of populations experiencing homelessness or those using cocaine. Carter et al.¹⁹ reported retention in MOUD at 3 and 6 months as 37% and 27%, respectively, of 95 people currently experiencing homelessness in a street medicine program in San Francisco, California. This study used a broad definition of treatment retention and categorized patients as retained on buprenorphine during the month in question if they had an active prescription for buprenorphine for more than 2 weeks of the 4-week period. Patients did not have to have active prescriptions in the previous months to be considered retained on buprenorphine during the month in question if they met these criteria. Additionally, this study included patients who transferred to another program as retained in treatment.¹⁹

Stancliff et al.²⁰ reported active retention in MOUD at 3 and 6 months as 42% and 31%, respectively, in 100 people from marginalized populations (marginally housed, low income) attending the Harm Reduction Coalition Buprenorphine Maintenance Harm Reduction Program in New York City. In this study, participants were given a 30-day prescription for buprenorphine and were considered active in the program up to 45 days from the last prescription.²⁰ Other promising low-threshold buprenorphine programs designed for specific populations in nontraditional settings include the Project Connections at Re-Entry program, which is delivered in a mobile van outside the Baltimore, Maryland, City Jail,²¹ and the Recovery Community Center Office-Based Opioid Treatment model, which is delivered in a community site with multiple resources.²²

There are limited data on the relationship between cocaine use and MOUD treatment retention, possibly because cocaine use has historically been regarded as a reason to restrict access to or continuation of MOUD.²³ One outpatient community health center study reported no difference in treatment retention or reported opioid use among cocaine users versus nonusers over a 6-month period. However, this study relied only on self-report of cocaine use, not UDS.^24,25 In our study, our comparison of months retained in treatment of cocaine versus noncocaine users had a nonsignificant log-rank test; however, this may reflect our small sample size. The retention trends showed shorter times in treatment of cocaine versus noncocaine users (5.33 vs 11.54 months). Therefore, further investigation into possible mechanisms that support treatment retention in these subpopulations is warranted.

Eighty percent of our patients accessed primary care at our FQHC during the study period. By nature of the group-based integrated model, 100% of patients received group-based behavioral health and substance use services. In comparison, the FQHC sites in Haddad et al.¹⁸ reported similar results in primary care utilization of 90% of patients over the 18-month study period. However, in this non–group-based integrated model, 56% of patients accessed onsite behavioral health services, and 53% accessed substance use services.

Limitations

We acknowledge several limitations of our study, including the retrospective design, relatively short follow-up period, and lack of a comparison group. We did not classify incarceration or planned treatment transfers as treatment failures, and this may mask later negative outcomes. Three patients left the program after their initial visit, but we did not track the reasons for these dropouts, which is important information for continuing to improve our program. The reported number of participants with cooccurring psychiatric diagnoses may be an underestimate as the medical and psychiatric conditions were pulled from billing diagnoses—the majority of which are generated by primary care. Generalizability is restricted by these issues and may not translate to FQHC settings with different demographic populations, those that lack the structure for group appointments, and those with different payment structures for medical and behavioral health services. Other supportive services unique to our program, such as help with transportation and childcare, may not be accessible to other FQHCs. Ongoing evaluation of the program will provide outcome results with a more extended period and with an expanded population.

There are some notable limitations in the structure of our program. Our main clinic site continues to offer MOUD only through the group-based model because of limitations in medical provider availability both for MOUD and primary care. Use of the same day–next day model for primary care means that appointment times are often filled by midmorning so patients cannot schedule a primary care appointment in advance of 1 day. In our smaller satellite clinics with limited patient populations, we are able to offer individual MOUD appointments for patients who are unable to access group sessions because of medical and social complexities that preclude a regular schedule. The philosophy of our program is inherently patient centered, and we recently completed a series of in-depth focus groups with patients and staff. In response, we plan to develop a patient advisory board to provide an ongoing structural method to include patients in ongoing program expansion and evaluation.

Public Health Implications

Integrating treatment of OUD into primary care shows promise in increasing access to and retention in MOUD services. Our program provides early evidence of an efficient, transferable model to support the expansion of MOUD to other FQHCs and primary care sites both locally and at a national level. From the onset, our program embraced low-threshold concepts of harm reduction and flexibility, which may improve access to MOUD for marginalized populations. In continued efforts to implement a low-threshold program, we are now approaching same-day treatment entry for all patients and are tending our services to nontraditional locations, including a large drop-in center for people experiencing homelessnessness.⁸

We believe our logic model (Figure 1) can serve as a potential framework for the expansion of low-threshold MOUD models more broadly. Further plans for the development of our program model include comprehensive evaluation focusing on health care service outcomes, patient outcomes, and implementation outcomes. Our multilevel approach will include quantitative and qualitative data from the individual and organizational levels and will incorporate perspectives from patients, providers, learners, staff, key stakeholders, and policymakers. These results will be used to inform the development of an enhanced, definitive intervention ready for larger-scale testing in a variety of settings.