Discontinuation of Buprenorphine Treatment for Opioid Use Disorder During the Coronavirus Disease-2019 Pandemic

Background:

The coronavirus disease-2019 pandemic has been associated with large increases in opioid-related mortality, yet it is unclear whether specific subpopulations were especially likely to discontinue buprenorphine treatment for opioid use disorder as the pandemic ensued.

Objective:

The aim was to assess predictors of buprenorphine discontinuation in the early months of the coronavirus disease-2019 pandemic (April–July 2020) compared with a prepandemic period (April–July 2019).

Design:

In each time period, we estimated a multilevel regression models to assess risk of discontinuation in April–July for people who started buprenorphine in January–February. Models included person-level, prescriber-level, and area-level covariates.

Subjects:

Individuals age 18 years or older in the all-payer IQVIA Longitudinal Prescription Claims.

Measures:

The primary outcome was buprenorphine discontinuation (ie, no filled prescriptions during the follow-up periods).

Results:

Overall, 13.98% of patients discontinued buprenorphine in April–July 2020, less than the 15.71% in 2019 (P<0.001). In 2020, patient-level factors associated with discontinuation included younger age, male sex, shorter baseline possession ratio, and payment by cash. Compared with patients with a primary care physician prescriber, specialties most associated with discontinuation were pain medicine and physician assistant/nurse practitioner. Compared with the South Atlantic region, discontinuation risk was lowest in New England and highest in the West South Central States. The association between patient, prescriber, and geographic variables to risk of discontinuation was very similar in 2019 and 2020.

Conclusions:

While clinical and policy interventions may have mitigated opioid use disorder treatment discontinuation following the pandemic, such discontinuation is nevertheless common and varies by identifiable patient, provider and geographic factors.

More than 93,000 individuals died from drug overdose in the United States in 2020, and an estimated three-quarters of these deaths involved opioids.1 Overdose deaths accelerated during the coronavirus disease-2019 (COVID-19) pandemic.2 Likely contributors include an increasingly unpredictable illicit drug supply with spreading fentanyl and social isolation.3,4 Other disruptions may have been related to access to medications for opioid use disorder.5 Buprenorphine, the most widely used medication treatment, normally requires in-person visits to a prescriber, which was a challenge early in the pandemic.5

To mitigate such impacts, policy changes were instituted allowing for buprenorphine initiation using telehealth and loosening regulations to allow for telephonic visits without video.6 Federal regulations waived requirements for face-to-face visits, and some state laws further supported flexibility during the pandemic.7,8 Several studies suggest these policies achieved their goal, as national buprenorphine volume remained stable during the early pandemic,9–12 and following the public health emergency, the total volume of buprenorphine plateaued, whereas other prescription medications declined.9 The onset of the pandemic led to a substantial shift in mental health and substance use treatment to telehealth, which is a likely factor explaining better continuity of care.13

The literature offers important insights, but much remains unknown about patient retention in buprenorphine treatment, an important predictor of recovery and reduced overdose risk.14 First, while many prior studies are focused on specific payers and geographic areas, there is a dearth of research focused on large, multipayer databases, which can provide a more complete picture of the national patient population. Second, it is unclear whether the pandemic had a differential impact on patient subgroups, such as older people who may have been less likely to adopt telehealth,15 otherwise accommodate to the changing structure of health care delivery, or who resided in areas with an especially high COVID-19 burden, which would be expected to be associated with greater disruptions in treatment. Third, it is uncertain how much buprenorphine treatment retention is fundamentally because of patient-level factors, versus factors that arise at the level of the clinic or geographic area. Retention differs widely across age and sex, insurance coverage, clinician-level factors such as specialist type and size of patient panel (as higher-volume prescribers are unique practices16,17), and geographic areas,14,18–20 but prior literature has not attempted to quantify the relative contribution of different levels in a single framework.

We examined these questions using a multilevel framework with a dataset tracking a near census of adults in the United States filling prescriptions for buprenorphine from retail and mail order pharmacies. A separate study addresses buprenorphine prescriptions for adolescents.21 We hypothesized that 2020 discontinuation rates would be similar to 2019 overall, but that discontinuation would increase in 2020 in geographic areas that were initially most affected by COVID-19 and among groups that had less stability in treatment in 2019 (such as people with cash payment). We also expected that prescriber-level and area-level factors would explain a substantial amount of the overall variation in discontinuation.

METHODS

We used January 1, 2019 through July 31, 2020 IQVIA Longitudinal Prescription Claims (LRx), which captures more than 90% of prescriptions filled at retail and mail order pharmacies in the United States. We restricted our sample to individuals with one or more prescription fills for buprenorphine during January or February of a given year. We limited our analysis to buprenorphine products with a US Food and Drug Administration (FDA) indication for the treatment of opioid use disorder, which we identified using a comprehensive list of National Drug Codes (product names are in the appendix, Supplemental Digital Content 1, http://links.lww.com/MLR/C570). Each buprenorphine record included patient demographics (age and sex), prescription information (date of prescription, days of supply, and payment type) and prescriber information (prescriber ID, specialty code, ZIP, and state). After restricting to patients age 18 years or above and dropping patients with missing sex, or prescriber specialty or ZIP code, the final analytic sample contained 789,519 patients in 2019 and 789,275 patients in 2020 (eTable 1, Supplemental Digital Content 1, http://links.lww.com/MLR/C570).

Patient-Level Variables

To create patient-level predictors during the baseline period, we focused on their prescriptions in January or February of each year (whichever month they most recently appeared in). We aggregated the record-level information in that month to construct independent variables. Thus, for a given individual during a given month, we quantified: (1) the percentage of days with buprenorphine at hand, which we constructed using date of prescription and days of supply; and (2) 4 indicators of having any buprenorphine prescription for cash, Medicaid/Medicaid managed care, Medicare, and commercial payment.

Prescriber Variables

We assigned the most frequent buprenorphine prescriber to each individual during the baseline month (the primary prescriber). If a patient filled a prescription in both January and February, we use the latter value (ie, February). If 2 or more prescribers had the same number of buprenorphine prescriptions for an individual, we chose the earliest buprenorphine prescriber in that month as the primary prescriber. Through the assigned prescriber specialty code, we identified the specialty group, and adopted a parsimonious taxonomy of specialty codes: physician assistant/nurse practitioner (PA/NP), addiction/psychiatry, pain, adult primary care physician (PCP), and other.20 We categorized prescriber buprenorphine volume as low, medium, or high based on the terciles of total buprenorphine prescriptions across all buprenorphine prescribers in the month.

Geographic Variables

We proxied for the patient’s geographic area based on the ZIP code of the prescriber. At a more aggregate level, we also identified whether the patient’s residence was in an urban area (metropolitan area)22 and across the nine Census divisions:23 Pacific, Mountain, West North Central, West South Central, East North Central, East South Central, New England, Mid-Atlantic, and South Atlantic.

Prevalence of Coronavirus Disease-2019

In the 2020 cohort, we calculated the monthly reported prevalence of COVID-19 at the ZIP level in April and July using county-level data from the JHU CSSE Coronavirus tracking project.24 We cross-walked county rates to ZIP code rates by imputing the rate of the county to the ZIP code (where the ZIP was entirely in a county), or, for ZIP codes spanning more than one county, by assigning a proportionate rate based on share of the ZIP population in each county. We used these data to calculate two measures: the COVID-19 reported prevalence in April 2020 and the change in the COVID-19 prevalence from April to July.

Outcome

For each study subject, we determined whether they filled a buprenorphine prescription between April and July; because the first major lockdowns occurred in mid-March, 2020, and thus the month included prepandemic and postpandemic time, we excluded March from the sample as a “washout” period. We classified an individual as discontinuing buprenorphine if there were no filled prescriptions during the April–July period, and as a buprenorphine continuer if there was 1 or more buprenorphine prescription during this time. In a sensitivity analysis, we examined individuals who discontinued during April and May of a given year and then resumed prescriptions during June and July. The results and interpretation of this analysis were substantively similar to our main analysis and are reported in the appendix, Supplemental Digital Content 1, http://links.lww.com/MLR/C570.

Statistical Method

We compared the characteristics of the 2019 and 2020 cohorts. For categorical or binary variables, we performed χ² tests to evaluate whether differences existed between these 2 groups; for continuous variables we performed Wilcoxon rank-sum tests. We built similar models in both cohorts with the only exception being the COVID-19 measures for the 2020 cohort. To examine whether stopping buprenorphine was highly concentrated in a small number of prescribers in each year (2019 or 2020), we calculated for each primary prescriber the percent of their patients stopping buprenorphine in the April–July period. We ranked prescribers by percent of patients stopping buprenorphine and plotted the percentages of patients who stopped buprenorphine in each percentile of the prescriber distribution for the year.

We next estimated regression models for the probability of stopping buprenorphine. Given the hierarchical structure of the data—multiple patients could be assigned to the same prescriber while multiple prescribers could come from the same ZIP code—we applied a multilevel modeling approach. We adopted mixed linear models with random intercepts at the prescriber and ZIP level. We first examined the proportion of variance explained at each level with and without independent variables; then, we compared the coefficients associated with independent variables between two cohorts to evaluate whether COVID-19 had an impact on the predictors of becoming a buprenorphine discontinuer. The proportion of variance explained at each level was derived from dividing the variance at a given level by the sum of variance across all 3 levels. In addition, we also tested a logistic regression using generalized estimating equations to adjust for the correlation between observations as a sensitivity analysis. Our study was determined to be exempt by the Johns Hopkins School of Public Health IRB.

RESULTS

Crude Treatment Discontinuation Rates and Cohort Characteristics

The mean discontinuation rate was 13.98% in 2020 as compared with 15.71% in 2019. Table 1 displays the characteristics for the 2019 versus 2020 samples. In both years, slightly more than half (54.82% in 2020) the sample was male. The mean age in 2019 was 42.51 and in 2020 was 43.44. The most common source of payment was third party payment (primarily private insurance), followed by Medicaid, Medicare, and cash (in 2020: 47.95%, 34.69%, 13.03%, 6.87%). The majority of patients (>90%) were served by prescribers in the high-volume group. Adult PCP were the most commonly visited specialty for buprenorphine prescriptions followed by PA/NP, addiction/psychiatry, and pain medicine (in 2020: 49.74%, 18.07%, 15.48%, 10.19%). About 85% of patients resided in urban areas. The most common census divisions were South Atlantic, Mid-Atlantic, and East North Central (in 2020: 20.36%, 16.41%, 14.55%). The mean COVID-19 case rate in the ZIP code in April 2020 was 0.2% (48.48%) and the mean April to July change in COVID-19 case prevalence was 0.79 percentage points.

TABLE 1.

Summary Statistics of Study Sample

Variable	2020 (%)	2019 (%)
N	789,275	789,519
% discontinuing buprenorphine	13.98	15.71
With 1+ buprenorphine prescription in February	13.08	14.34
Male	54.82	54.90
Age in years	43.44	42.51
Age group
18–30	10.93	13.26
30–40	35.15	36.11
40–50	24.60	23.37
50–60	16.23	15.75
60+	13.09	11.51
% of days with buprenorphine on hand	82.72	81.43
Any buprenorphine fill paid by
Cash	6.87	7.74
Medicaid	34.69	34.78
Medicare	13.03	11.98
Third party	47.95	48.52
Provider: level of total buprenorphine volume
Low	1.20	1.35
Medium	6.49	6.52
High	92.31	92.13
Provider specialty
Addiction/psychiatry	15.48	17.06
PA/NP	18.07	14.05
Other	6.51	6.74
Pain	10.19	10.18
PCP adult	49.74	51.97
Urban area	84.41	84.96
Census division
Pacific	9.76	10.04
Mountain	5.58	5.64
West North Central	3.23	3.09
West South Central	7.98	7.87
East North Central	14.55	15.10
East South Central	12.38	12.24
New England	9.75	9.89
Mid-Atlantic	16.41	16.38
South Atlantic	20.36	19.76
Zip: % of population with COVID-19 in April 2020	0.20
Zip: level of COVID-19% in April
Low	15.81
Medium	35.71
High	48.48
Zip: change in % of COVID-19 between April and July	0.79

COVID-19 indicates coronavirus disease-2019; PA/NP, physician assistant/nurse practitioner; PCP, primary care physician.

Source: IQVIA Longitudinal Prescription Claims (LRx), 2019–2020.

Adjusted Treatment Discontinuation Rates

Table 2 displays the regression coefficients from the hierarchical models separately predicting discontinuation risk in 2019 and 2020. The point estimates across the 2019 and 2020 models were very similar for most of the predictors, and the focus here is mainly on 2020 estimates. In both years, the largest predictors of discontinuation were patterns of buprenorphine filling in the baseline period. Having any buprenorphine prescription in February, versus exclusively in January, was associated with 28.0 percentage point higher discontinuation rate (95% CI: 0.277–0.282). A 10 percentile higher possession ratio for buprenorphine was associated with a 3.3 percentage point lower discontinuation rate (95% CI:−0.033 to −0.032). Male sex (vs. female) was very modestly associated with higher discontinuation (0.008; 95% CI: 0.006–0.009). Greater age was associated with a lower discontinuation rate compared with age 18–29, for example, age 30–40 was 2.8 percentage points lower (95% CI: −0.030 to −0.025). Having cash as a source of payment was associated with a 1.2 percentage point higher discontinuation rate (95% CI: 0.007–0.016), while Medicare rates (−0.01; 95% CI:−0.015 to −0.005) and commercial insurance (−0.002; 95% CI:−0.026 to −0.017) were associated with lower discontinuation, and the coefficient for Medicaid was close to zero and not significant (−0.002; 95% CI:−0.006 to 0.003).

TABLE 2.

Linear Mixed Effect Models for Risk of Discontinuation

	2020			2019
Variable	Coefficients	P	95% CI		Coefficients	P	95% CI
Intercept	0.409	<0.0001	0.400	0.418	0.451	<0.0001	0.443	0.458
≥1 bup prescription in February (ref=January only)	0.280	<0.0001	0.277	0.282	0.275	<0.0001	0.272	0.277
Male (ref=female)	0.008	<0.0001	0.006	0.009	0.008	<0.0001	0.006	0.009
Age (ref=18–29)
30–40 years	−0.028	<0.0001	−0.030	−0.025	−0.031	<0.0001	−0.034	−0.029
40–50 years	−0.035	<0.0001	−0.037	−0.032	−0.038	<0.0001	−0.040	−0.035
50–60 years	−0.038	<0.0001	−0.040	−0.035	−0.041	<0.0001	−0.044	−0.039
60+ years	−0.036	<0.0001	−0.039	−0.033	−0.042	<0.0001	−0.045	−0.038
% days with bupe on hand (10 ptile increase)	−0.033	<0.0001	−0.033	−0.032	−0.034	<0.0001	−0.034	−0.034
Payment source
Cash	0.012	<0.0001	0.007	0.016	0.002	0.294	−0.002	0.007
Medicaid	−0.002	0.500	−0.006	0.003	−0.010	<0.0001	−0.015	−0.005
Medicare	−0.010	0.000	−0.015	−0.005	−0.019	<0.0001	−0.024	−0.014
Third party	−0.022	<0.0001	−0.026	−0.017	−0.037	<0.0001	−0.042	−0.033
Provider total buprenorphine volume (ref=high)
Low	0.028	<0.0001	0.021	0.035	0.038	<0.0001	0.032	0.045
Medium	0.021	<0.0001	0.018	0.024	0.027	<0.0001	0.023	0.030
Prescriber specialty (ref=PCP)
Addiction/psychiatry	0.000	0.853	−0.003	0.004	−0.004	0.045	−0.008	0.000
PA/NP	0.026	<0.0001	0.023	0.030	0.032	<0.0001	0.029	0.036
Other	0.015	<0.0001	0.010	0.021	0.008	0.010	0.002	0.013
Pain	0.038	<0.0001	0.034	0.043	0.039	<0.0001	0.035	0.044
Urban area	−0.003	0.210	−0.007	0.002	−0.002	0.317	−0.006	0.002
Census division (ref=South Atlantic)
Pacific	0.000	0.884	−0.006	0.005	−0.014	<0.0001	−0.019	−0.009
Mountain	0.005	0.090	−0.001	0.012	−0.005	0.118	−0.011	0.001
West North Central	−0.006	0.130	−0.014	0.002	−0.006	0.128	−0.014	0.002
West South Central	0.000	0.963	−0.006	0.006	0.002	0.435	−0.004	0.008
East North Central	−0.005	0.045	−0.011	0.000	−0.007	0.008	−0.012	−0.002
East South Central	−0.007	0.024	−0.013	−0.001	−0.003	0.257	−0.009	0.003
New England	−0.022	<0.0001	−0.028	−0.016	−0.032	<0.0001	−0.038	−0.026
Mid-Atlantic	−0.015	<0.0001	−0.021	−0.010	−0.021	<0.0001	−0.026	−0.016
COVID incidence in April 2020 (ref=high)
Low	0.005	0.074	0.000	0.010
Medium	0.002	0.334	−0.002	0.005
10 percentile increase in monthly COVID-19 incidence from April to July	−0.007	0.683	−0.039	0.025

COVID-19 indicates coronavirus disease-2019; PA/NP, physician assistant/nurse practitioner; PCP, primary care physician.

Estimates represent coefficients from hierarchical linear models predicting risk of buprenorphine discontinuation in 2020 or 2019 with random effects for prescriber and ZIP code.

Source: IQVIA Longitudinal Prescription Claims (LRx), 2019–2020.

Compared with patients of high-volume prescribers, patients of low (0.028; 95% CI: 0.021–0.035) and medium-volume (0.021; 95% CI: 0.018–0.024) prescribers had significantly higher rates of discontinuation. Compared with patients whose primary prescriber was a PCP, the rate of discontinuation was higher for PA/NP (0.026; 95% CI: 0.023–0.030), pain (0.038; 95% CI: 0.034–0.042), and “other” (0.015; 95% CI: 0.009–0.021), and not significantly different for addiction/psychiatry (0.000; 95% CI: −0.003 to 0.004). Compared with the South Atlantic region, adjusted differences across Census divisions were relatively modest except that discontinuation rates were notably lower in the New England (−0.022; 95% CI:−0.028 to −0.016) and mid-Atlantic regions (−0.015; 95% CI:−0.021 to −0.01). The area-level COVID-19-related measures were not substantially predictive of discontinuation risk, with the exception of residing in an area with low (compared with high) COVID-19 case rates in April, which was associated with slightly higher discontinuation risk (0.006; 95% CI: −0.001 to 0.012).

The coefficients for the 2019 model were very similar to 2020. Differences that were found for 2019 included that, in contrast to 2020, having cash as a source of payment was not significantly associated with discontinuation (0.002; 95% CI: −0.002 to 0.007) and having Medicaid was associated with discontinuation (−0.01; 95% CI:−0.015 to −0.005). There were also slightly different patterns of geographic associations in 2019. For example, compared with the South Atlantic region, the Pacific region had significantly lower discontinuation risk (−0.014; 95% CI:−0.019 to −0.009).

The parameters to assess the variance components in a hierarchical model without covariates was 89.91% patient-level, 8.28% for provider-level, and 1.80% provider ZIP in 2020. When covariates were added, the components were 96.05% patient-level, 3.34% provider-level, and 0.60% provider ZIP. Models were virtually identical for 2019 (Table 3).

TABLE 3.

Variance Parameters from Hierarchical Linear Model Predicting Risk of Buprenorphine Discontinuation

Level	2020 (%)	2019 (%)
With predictors
ZIP	0.60	0.37
Prescriber	3.34	3.47
Patient level	96.05	96.16
Without predictors
ZIP	1.80	1.56
Prescriber	8.28	8.54
Patient level	89.91	89.90

Estimates represent variance components from hierarchical linear models predicting risk of buprenorphine discontinuation (see Table 2 for coefficients).

Source: IQVIA Longitudinal Prescription Claims (LRx), 2019–2020.

Opioid Use Disorder Treatment Discontinuation Across Providers

Figure 1 visualizes the variation in mean discontinuation rates in 2019 and 2020 for individual prescribers. This figure was constructed by calculating the mean discontinuation rate for all primary patients (patients who received most/all of their prescriptions from the prescriber) of each prescriber with at least 10 patients in their panel and then ranking each prescriber by the share of patients who discontinued in the follow-up period. In both 2019 and 2020, there were 5% of prescribers whose patients had 0% discontinuation. The median discontinuation rate in 2020 was 12.08% with an interquartile range from 7.40% to 20.00%. Prescribers with the highest discontinuation rates were still generally below 40%. In 2019, the prescriber discontinuation curve was largely similar, but the median and interquartile range were slightly higher.

FIGURE 1.

Buprenorphine discontinuation at the level of the prescriber in 2019 and 2020. The percentage of patients who discontinue buprenorphine during the follow-up period is calculated for each prescriber with at least 10 primary patients in the January–February period of each year: 742,601 patients were included from 13,314 prescribers in 2019 and 745,807 from 13,062 prescribers in 2020. Percentage is shown on the y-axis and prescribers are arrayed by percentile (eg, the 50th percentile represents the prescribers with the median rate of patients stopping buprenorphine). Source: IQVIA Longitudinal Prescription Claims (LRx), 2019–2020.

DISCUSSION

This study examined buprenorphine discontinuation among patients early in the COVID-19 pandemic (April–July 2020) who had filled prescriptions earlier in the year. Predictors were compared with patients observed to the same months from 2019. Finding were generally consistent with prior research that found stability of buprenorphine prescription fills early in the COVID-19 pandemic.9–11 While we hypothesized that discontinuity might increase among some subgroups, we actually found that discontinuation during early 2020 was lower than early 2019. The discontinuation rate fell from 15.71% in 2019 to 13.98% in 2020, a roughly 11% improvement. We did not find evidence that patients in ZIP codes with higher COVID-19 burden in 2020 were more likely to discontinue treatment, which is unexpected given the impact of the pandemic on access to medical care generally.13

The decrease in discontinuation between 2019 and 2020 is noteworthy in its own regard, and could reflect a national trend toward better retention in buprenorphine treatment. While the workforce treating patients with buprenorphine has grown in recent years,25 relatively little is known about recent trends in continuity of treatment. Policy initiatives, such as the State Opioid Response grants and Medicaid continuum of care waivers, have been launched in recent years to increase access and quality of buprenorphine treatment.26 Another factor may be changes in the composition of patients filling buprenorphine prescriptions. Patients in the 2020 cohort had slightly better adherence heading into the pandemic than their 2019 counterparts, which may in turn reflect some demographic factors in the composition of buprenorphine. For example, in 2020, a higher proportion of patients were 60 or older, a group with higher adherence and generally better health insurance coverage.

We also extended prior research by assessing the contribution of patient, prescriber, and geographic factors with discontinuation of buprenorphine in a comprehensive national sample during and before the pandemic. Consistent with existing literature, we find that discontinuation rates were lowest in both 2019 and 2020 among older and commercially insured patients, among those receiving care from specialists in addiction/psychiatry compared with primary care, and in northeastern states known to have strong treatment infrastructure (eg, New York and Massachusetts).14,18,20 Although 2019 and 2020 were generally similar, one notable difference is that people with treatment covered by Medicaid had relative improvement in discontinuation, while this relatively worsened for people with cash payment. One possible explanation is that Medicaid enrollees had better continuity of treatment during the pandemic due to emergency provisions that stabilized enrollment and access to care for Medicaid enrollees,27 whereas those who paid cash may have experienced greater hardship from out-of-pocket cost. In general, insurance may have promoted lower discontinuation, since we also observed similar direction changes for commercial insurance and Medicare.

In both years, we found that most variation in discontinuation (about 90%) resided at the patient-level rather than the prescriber or ZIP code level. While perhaps surprising, this finding must be interpreted carefully because it only measures the direct influence of each level on overall variation. For example, the patient-level variables will capture factors such as insurance status that arise indirectly from area-level influences. Further, states varied in their implementation of treatment policies during COVID-19, but we do not attempt to measure these complex policy changes.8 Another surprising finding is that prescriber-level variation around discontinuation remained stable in the 2 years, with roughly similar distributions of average discontinuation rates for patients across prescribers. These differences likely reflect a combination of case-mix across prescribers (as some prescribers may serve more stable patients) and also prescriber or practice-level efforts to retain patients in care.

While the study is unique in its comprehensiveness and multilevel approach, it has limitations. First, the study lacks diagnosis codes and other clinical details. It also not possible to assess markers of clinical progress such as remission of symptoms and reduced overdose risk. Second, the primary measure of geographic aggregation (ZIP code) was measured at the prescriber level, and may be an imperfect proxy for the environment in which patients live. The measure of COVID-19 incidence relies on reported case rates, and is likely to be low in communities that had poor testing early in the pandemic. Third, the data period reflects issues arising early in the pandemic, which may not reflect current challenges facing patients. Fourth, we excluded patients with missing data on covariates, who tended to be older and female (Appendix, Supplemental Digital Content 1, http://links.lww.com/MLR/C570). Fifth, the data only included 1 medication treatment, excluding methadone and naltrexone. Finally, the study data end in July 2020, and therefore do not account for evolving trends in the pandemic.

CONCLUSION

Discontinuation among existing buprenorphine patients decreased in the early months of the COVID-19 pandemic compared with the same months from 2019. The emergency federal regulations created under the public health emergency likely have sustained better access to treatment. Despite the improvement relative to 2019, the study also highlighted notable differences across patients, prescribers, and geographic areas that persisted into the COVID-19 pandemic. With historically high overdose rates in 2020,28 it is important to develop further policy strategies to reduce buprenorphine discontinuation. The current study identifies multiple possible avenues that could be pursued through policies that target, for example, patients who pay cash for treatment or different prescriber specialties. Turning the tide on the overdose crisis depends on increasing access and quality of buprenorphine treatment for more Americans.

Supplementary Material

mlr-61-95-s001.docx ^{(34.8KB, docx)}