The treatment setting is a large FQHC near downtown Durham, NC that treats a diverse population including the local community and patients from surrounding rural areas. The FQHC has provided buprenorphine treatment since 2016 and offered extended-release naltrexone (XR-NTX) for OUD from January 2019 to June 2020 through a state-funded grant. During the traditional office-based model, patients seeking treatment were scheduled for either a group information session and then a behavioral health assessment, or only a behavioral health assessment, prior to an initial visit with a medical provider to start MOUD.
In November 2019, we implemented low barrier treatment by eliminating the requirement for an assessment or information session prior to the medical visit. We established protected same-day MOUD visits on medical provider schedules for new appointments, such that any patient who walked in or was seen in the clinic for another reason could establish care on-demand. Patients who called requesting initial appointments could often be accommodated same-day or within a few days. The low barrier model centered harm reduction, with patients continuing to be treated who had ongoing opioid and other substance use. Patients were encouraged to consider higher levels of care and supportive services when appropriate and available, but access to MOUD was not conditional on this engagement.
In both models of care, patients were connected with one of several medical providers at the clinic who prescribe MOUD, and this provider typically became their primary care provider and addressed primary care needs during MOUD visits. All buprenorphine inductions were home or non-facility-based inductions. Patients had appointments weekly until their buprenorphine dose was stable and then followed up less frequently at intervals between every two weeks and 3 months depending on their clinical stability. All patients met briefly with both a behavioral health counselor and a medical provider at every visit, per clinic leadership preferences and to facilitate team-based care. Although patients had scheduled appointments, our behavioral health team could accommodate patients who needed flexibility and meet on a walk-in basis to facilitate consistent access to MOUD. During the COVID-19 pandemic, the clinic transitioned to mostly telehealth visits for several months and then resumed in-person visits with the option of telehealth if there were barriers to clinic attendance.
Uninsured patients typically paid a sliding scale fee of $10 for their appointments and could receive vouchers to cover the cost of buprenorphine at the FQHC’s on-site pharmacy. During low barrier implementation, grant funding covered the costs of the sliding scale fees for the first ten visits for patients on buprenorphine. For patients on XR-NTX, all visit and medication costs were paid by the state grant.
We extracted electronic health records (EHR) data from Duke’s Epic/Clarity database, including patient demographics, encounter dates, visit types, diagnosis and procedure codes, prescription order records, and death dates. The study was approved by the Duke Institutional Review Board.
We included patients who had at least one office visit with an FQHC medical or behavioral health provider seeking MOUD, selecting the record with the earliest date the patient met with a member of the MOUD team as the patient’s index visit. We excluded patients with a buprenorphine or XR-NTX prescription from an FQHC provider in the 3 months preceding the index visit. We restricted the cohort to patients with index visits during: a) the low barrier intervention period (11/1/2019—7/31/2020) and b) the historical control period (3/1/2018—3/31/2019).
Intervention and historical control groups
The intervention and historical control groups were defined by index visits within the low barrier intervention and historical control periods. We selected the historical control group to be as close in time to the intervention while allowing for 7 months of follow-up for outcome ascertainment without any overlap with the intervention period.
Information on age, gender, race, ethnicity, insurance status and tobacco use as of the index visit were extracted from the EHR. Gender was defined as male or female based on the patient’s gender identity listed in the EHR; there were no non-binary participants. Information on race was condensed into the categories of white and Black/other due to the Duke EHR cell suppression policy, which did not allow reporting of frequencies between one and ten nor data that could allow back calculation of those small cells. We searched EHR data between the index visit and 1 year prior for diagnosis of comorbid physical and mental health conditions, including diabetes, HCV infection, HIV infection, MRSA or MSSA infection, chronic pain, depression, schizophrenia spectrum or other psychotic disorder, and bipolar disorder based on validated coding algorithms (Additional file 1: Appendix: Table S1).
Outcomes and measures
The primary outcomes were any MOUD prescription within 6 months of the index visit and 3- and 6-month retention in treatment without care gap [33,34,35]. Any MOUD prescription within 6 months was defined as any prescription of buprenorphine or XR-NTX within 6 months of index visit. Retention in treatment at 3 and 6 months without care gap was defined as an office or telemedicine visit with an MOUD medical or behavioral provider or MOUD prescription in the 3rd or 4th month following index visit (day 61–120 after index), or in the 6th or 7th month following index visit (day 151–210 after index), respectively, without a care gap. Care gap was defined as 60 consecutive days without an office/telemedicine visit or MOUD prescription.
Secondary outcomes were all-cause hospitalization and all-cause emergency department (ED) visit within 6 months of index visit. Hospitalization was defined by any inpatient claim including an ED visit that transferred to inpatient, while ED visit was defined as any ED encounter that did not result in an inpatient hospitalization. We included these utilization outcomes to provide information about rates of opioid-related harms and changes in health status.
Cells containing a value of 1 to 10 were suppressed in accordance with Duke EHR cell suppression policy.
We summarized baseline patient characteristics and tested for differences by intervention status using chi-square tests for categorical variables and Wilcoxon rank sum tests for continuous variables.
We calculated percentages of 6-month MOUD prescription, 3- and 6-month retention without care gap, all-cause hospitalization, and all-cause ED visit, and tested for group differences using chi-square tests. We described the number of all-cause hospitalizations and ED visits within 6 months using means with standard deviations and medians with interquartile ranges and tested for differences by group using Wilcoxon rank sum tests. Logistic regression models were used to evaluate the multivariable adjusted associations between intervention group and primary and secondary outcomes. For multivariable modeling, we adjusted for age, sex, race, ethnicity, insurance status, and comorbid conditions listed in the baseline characteristics table (Additional file 1: Appendix: Table S1).
We descriptively calculated rates of MOUD prescriptions and office/telemedicine visits in each of the 1st through 6th months following index visit. We also descriptively calculated rates of 3- and 6-month retention in care without excluding those with a care gap as a secondary analysis to explore the extent to which treatment interruptions affected retention in care. We described the rates of 6-month all-cause mortality, determined by death dates in the EHR, and 6-month unintentional opioid overdose, based on validated coding algorithms (Additional file 1: Appendix: Table S2).