Sociodemographic Differences in Quality of Treatment to Medicaid Enrollees Receiving Buprenorphine

Abstract

Background.

Buprenorphine is a key medication to treat opioid use disorder, but little is known about how treatment quality varies across sociodemographic groups.

Objective.

We examined measures of treatment quality and explored variation by sociodemographic factors.

Methods.

We used Medicaid MAX data from 50 states from 2006 to 2014 to identify buprenorphine treatment episodes (N=317,494). We used multivariable logistic regression to examine the quality of buprenorphine treatment along four dimensions: (1) sufficient duration, (2) effective dosage, and concurrent prescribing of (3) opioid analgesics and (4) benzodiazepines. We explored how quality varied by race/ethnicity, age, sex, and urbanicity.

Results.

In adjusted models, compared to non-Hispanic White individuals, non-Hispanic Black and Hispanic individuals had lower odds of receiving effective dosage (OR=0.79) and sufficient duration (OR=0.64), and lower odds of concurrent prescribing of opioid analgesics (OR=0.86) and benzodiazepines (OR=0.51). Older individuals had higher odds of sufficient duration (ORs from 1.21-1.33), but also had higher odds of concurrent opioid analgesics prescribing (ORs from 1.29-1.56) and benzodiazepines (ORs from 1.44-1.99). Females had higher odds of sufficient duration (OR=1.12), but lower odds of effective dosage (OR=0.77) and higher odds of concurrent prescribing of opioid analgesics (OR=1.25) and benzodiazepines (OR=1.16). Compared to individuals living in metropolitan areas, individuals living in non-metropolitan areas had higher odds of sufficient duration (ORs=1.11 and 1.24) and effective dosage (ORs=1.06 and 1.33), and lower odds of concurrent prescribing (ORs from 0.81-0.98).

Conclusions.

Black and Hispanic individuals were less likely to receive effective buprenorphine dosage and sufficient duration. Quality results were mixed for older and female individuals; although these individuals were more likely to receive treatment of sufficient duration, they were also more likely to be concurrently prescribed potentially contraindicated medications, and females were less likely to receive effective dosage. Findings raise concerns about adequacy of care for minority and other at-risk populations.

Introduction

Opioid use disorder (OUD) represents an acute public health crisis in the United States.¹ In 2018, the estimated OUD-related cost to society totaled $786.8 billion.² More than 2.1 million Americans are affected by OUD,³ and fatal opioid overdoses, a major cause of death,⁴ increased 480% over the period 1999 to 2018.⁵

Medications for OUD (MOUD), such as methadone, buprenorphine, and naltrexone, are gold standard treatments for OUD. MOUD improves patient outcomes compared to counseling alone.^6–8 State and federal policies and initiatives have expanded the availability and utilization of MOUD, including buprenorphine.^9–16 Buprenorphine is particularly advantageous: unlike methadone, which can only be dispensed in opioid treatment programs (OTPs), it can be prescribed in office-based settings. Use of buprenorphine has significantly improved access to MOUD, especially where geographic and patient preferences prohibit access to MOUD within OTPs.^{9,10,15,17,18}

Although buprenorphine treatment for opioid use disorder has improved access, the quality of buprenorphine treatment may be suboptimal.^19–25 Quality of care metrics are not well-established for buprenorphine treatment; ^26–28 however, indicators of poor quality can include buprenorphine treatment episodes of insufficient duration, often considered to be episodes less than 180 days,^{7,15,24,25,29} and buprenorphine dosages less than 8mg, commonly considered the lowest effective dose.³⁰ Furthermore, concurrent prescribing of several other medications to individuals being treated for OUD should be undertaken cautiously. These medications include benzodiazepines, which are associated with an elevated risk of overdose among individuals misusing opioids,³¹ and opioid analgesics, which can facilitate relapse.^24,31–34 However, these medications are not always contraindicated for individuals with OUD and may be a clinically meaningful part of individual treatment plans.

Little is known, however, regarding how these metrics vary across populations. For example, given differences in the receipt and maintenance of buprenorphine treatment among historically underserved and high-risk populations, including racial and ethnic minorities,^35–37 it is essential to understand the degree to which buprenorphine treatment quality varies across these groups. We examine an array of sociodemographic factors and focus specifically on race/ethnicity given the growing recognition of racial/ethnic disparities in access to care generally^38,39 and substance use treatment specifically.^35,40 We examined buprenorphine treatment among Medicaid enrollees in multiple states over the period 2006 to 2014. This time period represents an era when buprenorphine treatment began scaling up rapidly among Medicaid enrollees.^{19,35,41–44} While the policy environment and epidemiology of opioid use was different during this time period, any variation that existed across sociodemographic groups is likely to be informative to current challenges addressing lingering, long-term disparities in quality of care for individuals with OUD. Specifically, we assessed the quality of buprenorphine treatment over time along four dimensions: (1) sufficient duration, (2) effective dosage, (3) concurrent prescribing of opioid analgesics, and (4) concurrent prescribing of benzodiazepines, and we explored the extent to which quality varied by patient sociodemographic characteristics, including race/ethnicity, age, sex and urbanicity.

Methods

Data and Sample.

We used Medicaid administrative data from 2006-2014 Medicaid Analytic eXtract (MAX) files to identify individuals 14 to 64 years of age who filled buprenorphine formulations approved for OUD treatment. We excluded individuals who were not Medicaid-enrolled for the month of the first buprenorphine fill and for the subsequent three months. We also excluded treatment episodes that began before March 1, 2006, and after October 1, 2014, to ensure a sufficient gap in buprenorphine treatment before the start of the episode and to allow for sufficient follow-up observation, following the standard approach for constructing episodes of care within claims data.^22,45 Medicaid MAX data are not available for all states in each year we examined, and data for fewer states were available in the later years we examined. A table showing states and years of data availability is provided in Appendix Table 4.

We constructed a buprenorphine treatment episode using National Drug Codes (NDCs) to identify buprenorphine in pharmacy claims;³⁵ we included only buprenorphine formulations that the Food and Drug Administration (FDA) had approved for OUD treatment, and we excluded buprenorphine formulations that the FDA had approved solely for pain management. A list of NDCs used to identify buprenorphine prescription fills is provided in Appendix Table 5. Buprenorphine episodes started on the date of the first filled buprenorphine prescription following at least 60 days of the individual not having any buprenorphine from a prior prescription; the episode ended when the supply of buprenorphine had been exhausted, followed by at least 60 days with no filled buprenorphine prescription. The RAND institutional review board determined the research to be exempt.

Variables.

We used four previously used indicators to examine quality of buprenorphine treatment for each episode.^{21,24,32,46,47} We identified individuals in continuous receipt of buprenorphine for at least 180 days, consistent with other studies, the National Quality Forum, and recommendations from the American Society of Addiction Medicine, all of which have used this measure of sufficient duration.^{7,15,24,25,29} We identified an average daily dose of at least 8 milligrams as effective dosage (i.e., above a low dose), as dosage below 8 milligrams is unlikely to be clinically effective, and higher doses of 8-16 milligrams have demonstrated better retention.³⁰ We also identified buprenorphine treatment episodes with concurrent fills of opioid analgesics and concurrent fills of benzodiazepines, defined as episodes in which there were filled opioid analgesic or benzodiazepine prescriptions during the episode, respectively.³³ Although concurrent use of opioid analgesics or benzodiazepines is not always contraindicated in individuals being treated with buprenorphine for OUD,²⁴ concurrent use substantially increases the risk of abuse and overdose death. ^31,32

Independent Variables.

Using the MAX eligibility files, we determined an individual’s sex; age (14-25 years, 26-35, 36-45, 46-54, and 55-64); and race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, or other/unknown race/ethnicity). Using the MAX outpatient and inpatient services files, we determined whether an individual had a concurrent mental health or chronic condition comorbidity (see Appendix Table 6 for the diagnosis codes included, based on the International Classifications of Disease, Version 9 (ICD-9) codes).^48,49 We defined the buprenorphine episode year as the calendar year in which the first buprenorphine fill of the treatment episode occurred.

Analytic Approach.

We first conducted univariate and bivariate analyses, describing characteristics of Medicaid enrollees with buprenorphine treatment episodes. We then conducted multivariable logistic regressions. The outcomes were our dichotomous quality measures; the independent variable were individual-level race/ethnicity, age group, sex, and mental health and chronic condition comorbidity status, and county-level urbanicity (categorized as metropolitan, non-metropolitan adjacent to a metropolitan area, and non-metropolitan not adjacent to a metropolitan area, using 2013 Rural-Urban Continuum Codes).⁵⁰ We structured our analysis as a pooled cross-sectional analysis but allowed individuals to contribute multiple treatment episodes in the data. As such, we clustered our standard errors at the patient level, which is our primary unit of analysis. Each observation in our dataset is a patient episode linked to state of residence and year of the data. More than 75% of our analytic sample is made up of patients with a single episode. We included year and state fixed effects to account for state- and time-invariant characteristics that may influence the predictor or outcome variables. P-values less than 0.05 were significant for all tests. Analyses were conducted using STATA 16.0 (College Station, TX).

Results

We identified 317,494 buprenorphine treatment episodes for 240,871 Medicaid enrollees (mean episodes 1.65, SD=0.95); the mean and median episode duration was 182 days (SD=134.9) and 158 days, respectively. Medicaid enrollees with buprenorphine treatment episodes were primarily female (57.8%), non-Hispanic White (77.3%), and residents of metro counties (78.2%) (Table 1).

Table 1.

Characteristics of Buprenorphine Treatment Episodes, 2006-2014

	N (%)
Total	317,494 (100)
Age Cohort
14-25	53,771 (16.9)
26-35	140,168 (44.1)
36-45	69,687 (21.9)
46-55	41,781 (13.2)
56-64	12,087 (3.8)
Sex
Male	134,086 (42.2)
Female	183,408 (57.8)
Race/Ethnicity
Non-Hispanic White	245,344 (77.3)
Non-Hispanic Black	28,566 (9.0)
Hispanic	25,603 (8.1)
Other/Unknown Race/Ethnicity	17,981 (5.7)
Mental Health Comorbidity
Yes	160,988 (50.7)
No	156,506 (49.3)
Chronic Condition Comorbidity
Yes	81,556 (25.7)
No	235,938 (74.3)
Sufficient Duration (≥ 180 days)
Yes	146,149 (46.0)
No	171,345 (54.0)
Appropriate Dosage (average daily dose ≥ 8 mg)
Yes	279,275 (88.0)
No	38,219 (12.0)
Concurrent Opioid Analgesics
Yes	75,770 (23.9)
No	241,724 (76.1)
Concurrent Benzodiazepines
Yes	79,225 (25.0)
No	238,269 (75.0)
High Quality for All Metrics
Yes	68,647 (21.6)
No	248,847 (78.4)
Urbanicity
Metro	248,292 (78.2)
Non-metro, adjacent to urban area	45,715 (14.4)
Non-metro, not adjacent to urban area	23,487 (7.4)

The most common age category was 26-35 years old (44.1%), followed by age groups 36-45 (21.9%), 14-25 (16.9%), 45-54 (13.2%), and 55-64 (3.8%). Slightly more than half (50.7%) of episodes had a concurrent mental health condition, and slightly over one quarter (25.7%) had a concurrent chronic condition. Slightly fewer than half (46.0%) of episodes had sufficient duration (at least 180 days), and 88.0% of episodes had an average daily dose of at least 8 milligrams. Approximately one-quarter of episodes had concurrent opioid analgesic fills (23.9%) and concurrent benzodiazepine fills (25.0%). Only 21.6% of episodes met all criteria for high quality buprenorphine treatment (sufficient duration, effective dosage, and no concurrent filling of either benzodiazepines or opioid analgesics prescriptions).

Multivariable logistic regression results. (Table 2)

Table 2.

Characteristics Associated with Quality of Buprenorphine Treatment, 2006-2014

	N of episodes	% of Episodes with Sufficient Duration	Sufficient Duration aOR (95% CI)	% of Episodes with Effective Dosage	Effective Dosage aOR (95% CI)	% of Episodes with Concurrent Opioid Analgesics	Concurrent Opioid Analgesics aOR (95% CI)	% of Episodes with Concurrent Benzodiazepines	Concurrent Benzodiazepines aOR (95% CI)
Total	317,494	46.0	-	88.0	-	23.9	-	25.0	-
Age
14-25 (omitted)	53,771	40.7	REF	86.5	REF	18.6	REF	17.9	REF
26-35	140,168	47.8	1.33 (1.30 1.36)	88.6	1.18 (1.15, 1.22)	22.6	1.29 (1.25, 1.32)	24.5	1.44 (1.40, 1.48)
36-45	69,687	46.8	1.29 (1.25, 1.32)	88.8	1.17 (1.13, 1.21)	26.1	1.47 (1.42, 1.51)	27.8	1.70 (1.65, 1.76)
46-55	41,781	45.7	1.21 (1.17, 1.24)	87.5	1.01 (0.97, 1.06)	29.3	1.56 (1.51, 1.62)	29.2	1.78 (1.72, 1.85)
56-64	12,087	45.4	1.24 (1.19, 1.30)	83.8	0.75 (0.70, 0.80)	29.9	1.55 (1.48, 1.63)	30.4	1.99 (1.88, 2.10)
Sex
Male (omitted)	134,086	43.9	REF	89.0	REF	21.1	REF	22.2	REF
Female	183,408	47.6	1.12 (1.10, 1.13)	87.2	0.77 (0.75, 0.79)	25.9	1.25 (1.22, 1.27)	26.9	1.16 (1.14, 1.18)
Race/Ethnicity
Non-Hispanic White (omitted)	245,344	47.5	REF	88.3	REF	23.6	REF	25.9	REF
Non-Hispanic Black	28,566	38.1	0.64 (0.62, 0.66)	85.9	0.79 (0.75, 0.82)	27.3	0.86 (0.83, 0.89)	21.3	0.51 (0.49, 0.54)
Hispanic	25,603	40.6	0.71 (0.69, 0.73)	88	0.89 (0.85, 0.93)	23.6	0.85 (0.82, 0.89)	21.3	0.59 (0.57, 0.62)
Other/Unknown Race/Ethnicity	17,981	46.4	0.95 (0.91, 0.98)	87	0.91 (0.87, 0.96)	21.8	0.93 (0.89, 0.97)	23.1	0.79 (0.76, 0.83)
Mental Health Comorbidity
No (omitted)	156,506	38.6	REF	87.6	REF	18.4	REF	13.7	REF
Yes	160,988	53.3	1.71 (1.69, 1.74)	88.3	1.13 (1.10, 1.15)	29.2	1.59 (1.56, 1.61)	35.9	3.38 (3.32, 3.45)
Chronic Condition
No (omitted)	235,938	43.3	REF	87.8	REF	20.3	REF	22.3	REF
Yes	81,556	53.8	1.56 (1.54, 1.59)	88.3	1.05 (1.02, 1.08)	34.1	1.76 (1.72, 1.79)	32.8	1.35 (1.32, 1.38)
Urbanicity
Metro (omitted)	248,292	44.9	REF	88.0	REF	23.8	REF	25.6	REF
Non-metro, adjacent to urban area	45,715	48.8	1.11 (1.09, 1.14)	86.8	1.06 (1.02, 1.10)	23.3	0.98 (0.95, 1.00)	21.5	0.83 (0.81, 0.86)
Non-metro, not adjacent to urban area	23,487	52.8	1.24 (1.20, 1.29)	89.6	1.33 (1.26, 1.41)	25.7	0.93 (0.90, 0.97)	24.6	0.81 (0.77, 0.84)

^*Note: All percentages shown in Table 2 represent unadjusted percentages of episodes in each covariate group that met criteria for the dichotomous quality measure. All adjusted models include dichotomous year covariates and state fixed effects, and standard errors clustered at the individual level.

Episode duration.

In the full sample, 46% of buprenorphine treatment episodes were at least 180 days. Among race/ethnicity groups, we found that compared to 47.5% of non-Hispanic White individuals, all other race/ethnicity groups had significantly lower odds of buprenorphine treatment episodes of at least 180 days. Just 38.1% of non-Hispanic Black individuals (aOR= 0.64, 95% CI 0.62 to 0.66), 40.6% of Hispanic individuals (aOR = 0.71, 95% CI 0.69 to 0.73), and 46.4% of individuals with other/unknown race/ethnicity (aOR = 0.95, 95% CI 0.91 to 0.98) had buprenorphine treatment episodes of sufficient duration.

Less than half (40.7%) of individuals 14-25 years old had buprenorphine treatment episodes of at least 180 days. Compared to this group, all older cohorts had significantly higher odds of sufficient duration, with unadjusted sufficient duration percentages ranging from 45.4% to 47.8% and aORs ranging from 1.21 (95% CI 1.17 to 1.24) to 1.33 (95% CI 1.30 to 1.36).

Females had significantly higher odds of having an episode of at least 180 days compared to males (47.6% of females versus 43.9% of males, aOR = 1.12, 95% CI 1.10 to 1.13). Forty-four point nine percent (44.9%) of individuals living in metro areas had episodes of sufficient duration. Compared to this group, those living in non-metro areas adjacent to metro areas (48.8% with sufficient duration; aOR = 1.11, 95% CI 1.09 to 1.14) and in non-metro areas not adjacent to metro areas (52.8% with sufficient duration; aOR = 1.24, 95% CI 1.20 to 1.29), had significantly higher odds of sufficient duration.

Effective dosage.

For the full sample, 88% of buprenorphine treatment episodes had an average dose of at least 8 mg. Among race/ethnicity groups, we found that compared to 88.3% of non-Hispanic White individuals, all other race/ethnicity groups had significantly lower odds of effective dosage. Just 85.9% of non-Hispanic Black individuals (aOR= 0.79, 95% CI 0.75 to 0.82), and 87.0% of individuals with other/unknown race/ethnicity (aOR = 0.91, 95% CI 0.87 to 0.96) had buprenorphine treatment episodes with an average daily dose of at least 8mg. Eighty-eight percent (88.0%) of Hispanic individuals had episodes with an average dose of at least 8mg; however, when controlling for other factors, odds were significantly lower (aOR = 0.89, 95% CI 0.85 to 0.93).

The majority (86.5%) of individuals 14-25 years old had buprenorphine treatment episodes with effective dosage. Compared to that group, individuals in cohorts 26-35 and 36-45 had significantly higher odds of effective dosage (88.6% of individuals 26-35, aOR = 1.18, 95% CI 1.15 to 1.22; 88.8% of individuals 36-45, aOR = 1.17, 95% CI 1.13 to 1.21). Individuals 56-64 had significantly lower odds of effective dosage (83.8%; aOR = 0.75, 95% CI 0.70 to 0.80).

Females had significantly lower odds of effective dosage compared to males (87.2% of females versus 89.0% of males, aOR = 0.77, 95% CI 0.77 to 0.79). The majority (88.0%) of individuals living in metro areas had effective dosage. Compared to that group, those living in non-metro areas adjacent to metro areas (86.8% with effective dosage; aOR = 1.06, 95% CI 1.02 to 1.10) and in non-metro areas not adjacent to metro areas (89.6% with effective dosage; aOR = 1.33, 95% CI 1.26 to 1.41) had significantly higher odds of effective dosage.

Concurrent Prescribing of Opioid Analgesics.

For the full sample, 23.9% of buprenorphine treatment episodes had concurrent prescribing of opioid analgesics. Compared to 23.6% of non-Hispanic White individuals, all other race/ethnicity groups had significantly lower odds of concurrent prescribing of opioid analgesics, after controlling for other factors. Approximately a quarter (27.3%) of non-Hispanic Black individuals (aOR = 0.86, 95% CI 0.83 to 0.89), 23.6% of Hispanic individuals (aOR = 0.85, 95% CI 0.82 to 0.89) and 21.8% of individuals with other/unknown race/ethnicity (aOR = 0.93, 95% CI 0.89 to 0.97) had buprenorphine treatment episodes with concurrent prescribing of opioid analgesics.

Among age groups, we found that compared to 18.6% of individuals 14-25 years old with concurrent prescribing of opioid analgesics, all older cohorts had significantly higher odds of concurrent prescribing of opioid analgesics; unadjusted percentages ranged from 22.6% to 29.9% and aORs ranging from 1.29 (95% CI 1.25 to 1.32) to 1.56 (95% CI 1.51 to 1.62).

Females had significantly higher odds of concurrent prescribing of opioid analgesics compared to males (25.9% of females versus 21.1% of males, aOR = 1.25, 95% CI 1.22 to 1.27). We found that 23.8% of individuals living in metro areas experienced concurrent prescribing of opioid analgesics. Twenty-three point three percent (23.3%) of those living in non-metro areas adjacent to metro areas and 25.7% of those living in non-metro areas not adjacent to metro areas experienced concurrent prescribing of opioid analgesics; however, when we controlled for other factors, odds of concurrent prescribing were significantly lower only for those living in non-metro areas not adjacent to metro areas (aOR = 0.93, 95% CI 0.90 to 0.97).

Concurrent Prescribing of Benzodiazepines.

In the full sample, 25% of buprenorphine treatment episodes had concurrent prescribing of benzodiazepines. Compared to 25.9% of non-Hispanic White individuals, all other race/ethnicity groups had significantly lower odds of concurrent prescribing of benzodiazepines. Slightly less than a quarter (21.3%) of non-Hispanic Black individuals (aOR = 0.51, 95% CI 0.49 to 0.54), 21.3% of Hispanic individuals (aOR = 0.59, 95% CI 0.57 to 0.62), and 23.1% of individuals with other/unknown race/ethnicity (aOR = 0.79, 95% CI 0.76 to 0.83) had buprenorphine treatment episodes with concurrent prescribing of benzodiazepines.

Individuals 14-25 years old had the lowest rate (17.9%) of concurrent prescribing of benzodiazepines; all older cohorts had significantly higher odds of concurrent prescribing of benzodiazepines, with unadjusted percentages ranging from 24.5% to 30.4% and aORs ranging from 1.44 (95% CI 1.40 to 1.48) to 1.99 (95% CI 1.88 to 2.10).

Females had significantly higher odds of concurrent prescribing of benzodiazepines compared to males (26.9% of females versus 22.2% of males, aOR = 1.16, 95% CI 1.14 to 1.18). Compared to 25.6% of individuals living in metro areas with concurrent prescribing of benzodiazepines, those living in non-metro areas adjacent to metro areas (21.5% with concurrent prescribing; aOR = 0.83, 95% CI 0.81 to 0.86) and in non-metro areas not adjacent to metro areas (24.6% with concurrent prescribing; aOR = 0.81, 95% CI 0.77 to 0.84) had significantly lower odds of concurrent prescribing of benzodiazepines.

We performed several sensitivity analyses to assess the robustness of our results. First, we excluded individuals who were not Medicaid-enrolled for the month of the first buprenorphine fill and for the subsequent six months, in an effort to account for differential follow-up. In the second sensitivity analysis, we removed 2014 data from our sample due to the potential confounding of Medicaid expansion in that year. In the third sensitivity analysis, we implemented both sample restrictions. None of the sensitivity analyses produced meaningfully different results. We show the analysis with the largest sample (including 2014 data and individuals with shorter periods of Medicaid enrollment), and provide the other analyses in Tables 1–3 of the Appendix.

Discussion

In our analysis of quality of care for individuals treated with buprenorphine for opioid use disorder, we found that non-Hispanic Black and Hispanic individuals experienced lower dosage and duration of buprenorphine treatment, characteristics of treatment likely to be directly associated with positive outcomes, compared to non-Hispanic White individuals, controlling for other factors. We have further considered measures of treatment quality related to concurrent prescribing of opioid analgesics and benzodiazepines, characteristics of treatment that potentially suggest insufficient coordination of care and may raise the risk of relapse or overdose among individuals who have relapsed, but we found less consistent relationships. The finding for non-Hispanic Black and Hispanic individuals is consistent with research indicating disparities in the use of medication treatment among historically underserved and at-risk populations, including racial and ethnic minorities.^35,40 Our findings demonstrate meaningful and troubling disparities in buprenorphine treatment quality for these groups. Given the increasing odds of opioid related problems and overdoses in racial/ethnic minority groups,^51–53 our work underscores the need to increase access and enhance the quality of MOUD treatment to non-Hispanic Black and Hispanic individuals.

Black and Hispanic Medicaid enrollees often have less access to care generally,^38,39 and poor access to treatment for pain has been a long-standing issue.^54–62 Poorer access more generally could help explain our finding that non-Hispanic Black and Hispanic individuals were less likely to be concurrently prescribed benzodiazepines and opioid analgesics. The difference in prescribing patterns could be amplified by clinicians if they are less likely to address pain in Black and Hispanic patients as they do in White patients.⁶³ Our paper draws attention to the potential differences through which non-White patients may encounter treatment; for example, Black and Hispanic patients are often treated for substance use disorder through their contact with the criminal justice system.^64,65 Furthermore, although Medicaid expansion increased access to care and health care affordability among low-income and minority groups,³⁹ widespread implementation of culturally-sensitive policy strategies are essential for meaningfully reducing disparities in opioid treatment. Strategies like faith-based community initiatives, rehabilitative versus punitive treatment approaches, and the use of racial and ethnic impact assessments^52,53 have the potential to improve treatment quality and health outcomes for minority individuals. Expanding efforts to integrate MOUD for justice-involved individuals,^66–68 where racial/ethnic minorities are disproportionately represented,^69,70 may also improve access and quality of care for minority individuals struggling with opioid misuse. Less than 1% of U.S. prisons and jails allow the use of MOUD in the treatment of OUD.^71,72 Policy initiatives are underway to address disparities in OUD treatment for justice-involved individuals: the National Institute on Drug Abuse (NIDA) has established the Justice Community Opioid Innovation Network (JCOIN), an initiative to study and generate evidence-based treatment strategies for individuals struggling with opioid misuse in justice settings.^73,74 This issue is especially salient given that more than half of individuals with OUD have reported criminal justice system involvement.^75,76 These differences we observe in care are likely to reflect structural inequalities in the healthcare system, and such inequities must be addressed using a range of approaches for both clinical and ethical reasons.

Our finding that females were more likely than males to have sufficient duration of buprenorphine treatment as well as concurrent fills of both opioid analgesics and benzodiazepines can probably be explained by the well-established literature⁷⁷ documenting both higher health care utilization generally^78–83 and increased opioid⁸⁴ and benzodiazepine⁸⁵ use specifically among women. Older individuals’ greater use of opioid analgesics and benzodiazepines generally likely also explains the higher rates of concurrent fills of both opioid analgesics and benzodiazepines in our older cohorts.⁸⁶ Our finding that females who received buprenorphine MOUD have lower buprenorphine doses is a novel finding; however, we note that there may be substantial variation in treatment patterns among women, as illustrated by the emerging literature on pregnant and postpartum buprenorphine MOUD access and retention outcomes.^87,88 Examining unique features of buprenorphine MOUD and how it varies for women may require additional research.

Compared to individuals living in metro areas, those living in non-metro areas adjacent to metro areas as well as those living in rural areas had higher odds of adequate dosage and sufficient duration of treatment and lower odds of concurrent prescribing of benzodiazepines and opioid analgesics (all indicating higher quality of buprenorphine treatment). Given the general challenges in accessing care in rural areas and the relative lack of providers,⁸⁹ we speculate that the higher quality of care for this group may reflect characteristics of providers who choose to practice in rural communities, and/or the characteristics of patients who are successful in accessing treatment in rural communities, where there is often less treatment available than in more urban communities. However, additional research should examine differences in treatment patterns between rural and more urban communities.

Our findings must be viewed in the context of the study limitations. We do not have all states for all years in our study period, so the results may not be generalizable to those states and years not examined. Further, the data are older, and our study period ends before Medicaid expansion occurred in many states; thus, it is possible that some of the treatment disparities observed were mitigated over time or with increased access to Medicaid coverage.^90,91 Race in the MAX data is subject to individual reporting practices from the states; a known issue is that some states have higher rates of unreported race.⁹² While there is an insufficient literature on the validity of Medicaid data, in general, race/ethnicity reported in federal administrative data have higher validity for Black, Hispanic and White individuals, and lower validity for other groups (e.g., Native Americans). Our study focuses on the distinct quality domain of prescribing quality and thus does not consider other important measures of OUD treatment quality, including screening and monitoring, and because our study focused on only one kind of medication for OUD, we are limited in our complete understanding of individuals who may have transitioned to methadone or long-acting naltrexone. While the quality indicators we used are appropriate for population-level monitoring, there may be unique cases where dosage, duration and coprescribing need to be tailored to individual patients’ clinical circumstances. We do not have detailed patient clinical information or data on severity of illness, which are likely related to our outcomes; therefore, we are unable to say anything about the appropriateness of treatment or associated clinical outcomes. Additionally, individuals may disenroll from Medicaid and not be observable in our data; while our requirement that patients must be enrolled in Medicaid for four continuous months surrounding the episode start reduces the likelihood of missing patients who have very short spells in Medicaid, we cannot fully account for the dynamics of patients switching in and out of Medicaid over time. Finally, patient status as fee-for-service or Managed care is not always accurately coded in the MAX data,⁹³ so we did not feel comfortable stratifying by or adjusting for this variable.

Conclusion

Our study is among the first to examine sociodemographic variation in buprenorphine quality of care in the United States. Since the study period, the impact of the overdose crisis has grown, particularly for Black and Hispanic individuals, populations experiencing the fastest rise in opioid overdose deaths.⁵³ The fact that differences vary by quality measure and subpopulation suggests that specific mechanisms must be considered in future interventions. For example, an emphasis on justice-involved individuals could disproportionately benefit minority individuals enrolling in Medicaid, while interventions focused on chronic pain may benefit populations experiencing concurrent opioid and buprenorphine prescribing. With more individuals able to access Medicaid through the continual expansion of the program, tailoring quality of care interventions to subgroups most at-risk could beneficially reduce disparities at a societal level.

APPENDIX

Appendix Table 1.

Characteristics Associated with Quality of Buprenorphine Treatment, 2006-2014: Sensitivity Analysis 1

	N of episodes	% of Episodes with Sufficient Duration	Sufficient Duration aOR (95% CI)	% of Episodes with Effective Dosage	Effective Dosage aOR (95% CI)	% of Episodes with Concurrent Opioid Analgesics	Concurrent Opioid Analgesics aOR (95% CI)	% of Episodes with Concurrent Benzodiazepines	Concurrent Benzodiazepines aOR (95% CI)
Total	258,788	54.3	-	88.3	-	25.4	-	26.0	-
Age
14-25 (omitted)	42,680	49.1	REF	86.9	REF	20.3	REF	18.9	REF
26-35	113,565	56.6	1.33 (1.29, 1.36)	88.9	1.18 (1.14, 1.22)	24.1	1.26 (1.22, 1.30)	25.5	1.43 (1.38, 1.47)
36-45	57,270	54.9	1.26 (1.22, 1.29)	89.2	1.16 (1.11, 1.21)	27.7	1.43 (1.38, 1.47)	28.7	1.68 (1.62, 1.74)
46-55	35,115	52.6	1.14 (1.11, 1.18)	87.9	1.00 (0.96, 1.06)	30.5	1.49 (1.44, 1.55)	30.1	1.76 (1.69, 1.83)
56-64	10,158	52.5	1.18 (1.12, 1.23)	84.3	0.75 (0.70, 0.80)	30.8	1.46 (1.38, 1.54)	31.1	1.95 (1.84, 2.07)
Sex
Male (omitted)	108,237	52.2	REF	89.3	REF	22.6	REF	23.3	REF
Female	150,551	55.8	1.07 (1.05, 1.09)	87.6	0.77 (0.75, 0.79)	27.4	1.23 (1.21, 1.26)	27.9	1.15 (1.12, 1.17)
Race/Ethnicity
Non-Hispanic White (omitted)	199,475	56.1	REF	88.6	REF	25.2	REF	27.0	REF
Non-Hispanic Black	23,931	44.1	0.61 (0.59, 0.63)	86.4	0.78 (0.74, 0.82)	28.8	0.85 (0.82, 0.88)	22.2	0.51 (0.49, 0.53)
Hispanic	21,089	47.5	0.68 (0.66, 0.71)	88.2	0.88 (0.84, 0.93)	25.4	0.87 (0.83, 0.90)	22.0	0.59 (0.56, 0.61)
Other/Unknown Race/Ethnicity	14,293	55.6	0.96 (0.93, 1.00)	87.5	0.91 (0.85, 0.96)	22.7	0.9 (0.86, 0.94)	23.9	0.79 (0.75, 0.83)
Mental Health Comorbidity
No (omitted)	124,175	46.5	REF	87.8	REF	19.6	REF	14.1	REF
Yes	134,613	61.5	1.75 (1.72, 1.77)	88.8	1.16 (1.13, 1.19)	30.8	1.58 (1.55, 1.61)	36.9	3.42 (3.35, 3.50)
Chronic Condition
No (omitted)	189,665	51.6	REF	88.2	REF	21.6	REF	23.1	REF
Yes	69,123	61.6	1.6 (1.57, 1.64)	88.7	1.07 (1.03, 1.10)	35.7	1.77 (1.73, 1.80)	33.7	1.35 (1.32, 1.38)
Urbanicity
Metro (omitted)	202,912	52.8	REF	88.4	REF	25.3	REF	26.6	REF
Non-metro, adjacent to urban area	36,846	58.1	1.16 (1.13, 1.19)	87.3	1.09 (1.04, 1.13)	24.9	0.98 (0.95, 1.01)	22.4	0.83 (0.81, 0.86)
Non-metro, not adjacent to urban area	19,030	62.3	1.27 (1.23, 1.32)	89.6	1.33 (1.25, 1.41)	27.3	0.93 (0.89, 0.97)	25.7	0.81 (0.77, 0.84)

^*Note: All percentages shown in Table 2 represent unadjusted percentages of episodes in each covariate group that met criteria for the dichotomous quality measure. All adjusted models include dichotomous year covariates and state fixed effects, and standard errors clustered at the individual level.

Appendix Table 2.

Characteristics Associated with Quality of Buprenorphine Treatment, 2006-2014: Sensitivity Analysis 2

	N of episodes	% of Episodes with Sufficient Duration	Sufficient Duration aOR (95% CI)	% of Episodes with Effective Dosage	Effective Dosage aOR (95% CI)	% of Episodes with Concurrent Opioid Analgesics	Concurrent Opioid Analgesics aOR (95% CI)	% of Episodes with Concurrent Benzodiazepines	Concurrent Benzodiazepines aOR (95% CI)
Total	290,376	46.5	-	88.1	-	24.6	-	25.2	-
Age
14-25 (omitted)	50,116	41.4	REF	86.7	REF	19.1	REF	18.2	REF
26-35	127,211	48.4	1.32 (1.29, 1.35)	88.7	1.18 (1.14, 1.22)	23.4	1.30 (1.26, 1.33)	24.9	1.44 (1.39, 1.48)
36-45	63,573	47.3	1.27 (1.24, 1.31)	88.9	1.16 (1.12, 1.21)	26.9	1.48 (1.43, 1.52)	28.0	1.69 (1.64, 1.75)
46-55	38,651	46.0	1.19 (1.16, 1.23)	87.6	1.01 (0.97, 1.06)	29.9	1.57 (1.51, 1.62)	29.2	1.78 (1.71, 1.85)
56-64	10,825	45.7	1.23 (1.17, 1.28)	84.0	0.75 (0.70, 0.80)	30.3	1.53 (1.45, 1.62)	30.3	1.99 (1.88, 2.11)
Sex
Male (omitted)	123,927	44.2	REF	89.2	REF	21.6	REF	22.3	REF
Female	166,449	48.2	1.12 (1.10, 1.14)	87.2	0.77 (0.75, 0.79)	26.7	1.24 (1.21, 1.26)	27.4	1.17 (1.14, 1.19)
Race/Ethnicity
Non-Hispanic White (omitted)	222,784	48.1	REF	88.4	REF	24.5	REF	26.3	REF
Non-Hispanic Black	26,736	38.4	0.65 (0.63, 0.67)	86.0	0.79 (0.75, 0.83)	27.7	0.85 (0.82, 0.88)	21.1	0.51 (0.49, 0.53)
Hispanic	24,789	40.6	0.71 (0.69, 0.73)	88.1	0.89 (0.84, 0.93)	23.7	0.85 (0.82, 0.88)	21.1	0.59 (0.56, 0.61)
Other/Unknown Race/Ethnicity	16,067	47.0	0.94 (0.91, 0.98)	87.5	0.91 (0.86, 0.97)	21.8	0.91 (0.87, 0.95)	22.8	0.76 (0.73, 0.80)
Mental Health Comorbidity
No (omitted)	143,273	39.0	REF	87.7	REF	18.9	REF	13.6	REF
Yes	147,103	53.9	1.73 (1.70, 1.75)	88.5	1.13 (1.10, 1.15)	30.0	1.59 (1.56, 1.62)	36.4	3.44 (3.38, 3.51)
Chronic Condition
No (omitted)	214,924	43.8	REF	88.0	REF	21.0	REF	22.5	REF
Yes	75,452	54.2	1.57 (1.54, 1.60)	88.4	1.05 (1.02, 1.08)	34.7	1.75 (1.71, 1.79)	32.9	1.35 (1.32, 1.38)
Urbanicity
Metro (omitted)	227,497	45.3	REF	88.1	REF	24.4	REF	25.7	REF
Non-metro, adjacent to urban area	41,545	49.4	1.10 (1.08, 1.13)	87.0	1.05 (1.02, 1.09)	24.1	0.97 (0.94, 1.00)	22.1	0.84 (0.82, 0.87)
Non-metro, not adjacent to urban area	21,334	53.8	1.25 (1.2, 1.29)	90.0	1.28 (1.21, 1.36)	26.5	0.94 (0.90, 0.98)	25.4	0.82 (0.78, 0.86)

^*Note: All percentages shown in Table 2 represent unadjusted percentages of episodes in each covariate group that met criteria for the dichotomous quality measure. All adjusted models include dichotomous year covariates and state fixed effects, and standard errors clustered at the individual level.

Appendix Table 3.

Characteristics Associated with Quality of Buprenorphine Treatment, 2006-2014: Sensitivity Analysis 3

	N of episodes	% of Episodes with Sufficient Duration	Sufficient Duration aOR (95% CI)	% of Episodes with Effective Dosage	Effective Dosage aOR (95% CI)	% of Episodes with Concurrent Opioid Analgesics	Concurrent Opioid Analgesics aOR (95% CI)	% of Episodes with Concurrent Benzodiazepines	Concurrent Benzodiazepines aOR (95% CI)
Total	236,365	54.3	-	88.4	-	26.1	-	26.2	-
Age
14-25 (omitted)	39,688	49.5	REF	87.1	REF	20.7	REF	19.2	REF
26-35	102,806	56.7	1.31 (1.28, 1.35)	89.0	1.18 (1.13, 1.22)	24.9	1.27 (1.23, 1.31)	25.8	1.43 (1.38, 1.47)
36-45	52,228	54.8	1.24 (1.20, 1.27)	89.3	1.16 (1.11, 1.21)	28.5	1.45 (1.40, 1.50)	28.9	1.67 (1.61, 1.73)
46-55	32,527	52.5	1.13 (1.09, 1.17)	88.0	1.01 (0.96, 1.06)	31.1	1.50 (1.45, 1.56)	30.0	1.75 (1.68, 1.83)
56-64	9,116	52.2	1.16 (1.10, 1.22)	84.4	0.75 (0.69, 0.80)	31.5	1.47 (1.39, 1.56)	30.9	1.94 (1.82, 2.07)
Sex
Male (omitted)	100,099	52.2	REF	89.6	REF	23.1	REF	23.3	REF
Female	136,266	55.9	1.08 (1.05, 1.10)	87.6	0.76 (0.74, 0.79)	28.2	1.23 (1.20, 1.26)	28.3	1.15 (1.12, 1.18)
Race/Ethnicity
Non-Hispanic White (omitted)	180,668	56.3	REF	88.8	REF	26.0	REF	27.4	REF
Non-Hispanic Black	22,430	44.1	0.61 (0.59, 0.64)	86.5	0.79 (0.75, 0.83)	29.2	0.84 (0.80, 0.87)	21.9	0.50 (0.48, 0.52)
Hispanic	20,425	47.4	0.68 (0.66, 0.71)	88.2	0.88 (0.84, 0.93)	25.5	0.86 (0.82, 0.89)	21.8	0.58 (0.55, 0.61)
Other/Unknown Race/Ethnicity	12,842	55.5	0.95 (0.92, 0.99)	87.9	0.90 (0.84, 0.95)	22.8	0.89 (0.85, 0.93)	23.7	0.76 (0.73, 0.80)
Mental Health Comorbidity
No (omitted)	113,177	46.6	REF	87.9	REF	20.2	REF	14.0	REF
Yes	123,188	61.4	1.73 (1.70, 1.76)	88.9	1.16 (1.12, 1.19)	31.5	1.58 (1.55, 1.61)	37.4	3.49 (3.41, 3.57)
Chronic Condition
No (omitted)	172,419	51.7	REF	88.3	REF	22.3	REF	23.4	REF
Yes	63,946	61.3	1.59 (1.56, 1.63)	88.8	1.05 (1.02, 1.09)	36.3	1.76 (1.72, 1.80)	33.8	1.35 (1.31, 1.38)
Urbanicity
Metro (omitted)	185,796	52.8	REF	88.4	REF	25.9	REF	26.7	REF
Non-metro, adjacent to urban area	33,356	58.2	1.14 (1.11, 1.17)	87.5	1.07 (1.03, 1.12)	25.7	0.97 (0.94, 1.00)	23.1	0.85 (0.82, 0.88)
Non-metro, not adjacent to urban area	17,213	62.9	1.28 (1.23, 1.33)	90.1	1.29 (1.21, 1.37)	28.3	0.94 (0.90, 0.98)	26.6	0.82 (0.78, 0.86)

^*Note: All percentages shown in Table 2 represent unadjusted percentages of episodes in each covariate group that met criteria for the dichotomous quality measure. All adjusted models include dichotomous year covariates and state fixed effects, and standard errors clustered at the individual level.

Appendix Table 4.

States and Years Available in the MAX Data

State	2006	2007	2008	2009	2010	2011	2012	2013	2014
AK	X	X	X	X	X	X	X
AL	X	X	X	X	X	X	X
AR	X	X	X	X	X	X	X	X
AZ	X	X	X	X	X	X	X	X
CA	X	X	X	X	X	X	X	X	X
CO	X	X	X	X	X	X	X
CT	X	X	X	X	X	X	X	X
DC	X	X	X	X	X	X	X
DE	X	X	X	X	X	X	X
FL	X	X	X	X	X	X	X
GA	X	X	X	X	X	X	X	X	X
HI	X	X	X	X	X	X	X	X
IA	X	X	X	X	X	X	X	X	X
ID	X	X	X	X	X		X	X	X
IL	X	X	X	X	X	X	X
IN	X	X	X	X	X	X	X	X
KS	X	X	X	X		X	X
KY	X	X	X	X	X	X	X
LA	X	X	X	X	X	X	X	X	X
MA	X	X	X	X	X	X	X	X
MD	X	X	X	X	X	X	X
ME						X	X
MI	X	X	X	X	X	X	X	X	X
MN	X	X	X	X	X	X	X	X	X
MO	X	X	X	X	X	X	X	X	X
MS	X	X	X	X		X	X	X	X
MT	X	X	X	X	X	X	X
NC	X	X	X	X	X	X	X
ND	X	X	X	X	X	X	X
NE	X	X	X	X	X	X	X
NH	X	X	X	X	X	X	X
NJ	X	X	X	X	X	X	X	X	X
NM	X	X	X	X	X	X	X
NV	X	X	X	X	X	X	X
NY	X	X	X	X	X	X	X	X
OH	X	X	X	X	X	X	X	X
OK	X	X	X	X	X	X	X	X
OR	X	X	X	X	X	X	X	X
PA	X	X	X	X	X	X	X		X
RI	X	X	X	X	X	X	X
SC	X	X	X	X	X	X	X
SD	X	X	X	X	X	X	X		X
TN	X	X	X	X	X	X	X		X
TX	X	X	X	X	X	X	X
UT	X	X	X	X	X	X	X		X
VA	X	X	X	X	X	X	X
VT	X	X	X	X	X	X	X		X
WA	X	X	X	X	X	X	X
WI	X	X	X	X	X	X	X
WV	X	X	X	X	X	X	X		X
WY	X	X	X	X	X	X	X		X

Appendix Table 5.

Buprenorphine NDCs

00054017613
00054017713
00054018813
00054018913
00093360021
00093360040
00093360121
00093360140
00093360221
00093360240
00093360321
00093360340
00093365640
00093365721
00093365740
00093365821
00093365840
00093365921
00093365940
00093537856
00093537956
00093572056
00093572156
00228315303
00228315403
00228315473
00228315503
00228315567
00228315573
00228315603
00378092393
00378092493
00406192303
00406192403
00406800503
00406802003
00409201203
00409201232
00490005100
00490005130
12496010001
12496012832
12496013062
12496030001
12496075701
12496075705
12496120201
12496120203
12496120401
12496120403
12496120801
12496120803
12496121201
12496121203
12496127802
12496128302
12496130602
12496131002
16590066630
16590066730
23490927003
35356000407
35356000430
35356055530
35356055630
35356060504
35356060604
35356060704
40042001001
42023017901
42023017905
42291017430
42291017530
42858035340
42858049340
42858050103
42858050203
42858058640
42858075040
42858083940
43063018430
43063066706
43063075306
43598058201
43598058230
49999039515
49999039530
49999063830
49999063930
50268014415
50268014511
50268014515
50383028793
50383029493
50383092493
50383093093
52959030430
52959074930
53217013830
53217024630
54123011430
54123090730
54123091430
54123092930
54123095730
54123098630
54569549600
54569573900
54569573901
54569573902
54569632500
54569632600
54569639900
54569640800
54569657800
54868570700
54868570701
54868570702
54868570704
54868575000
55045378403
55390010010
55700014730
55700018430
55700030230
55700030330
55700057904
58284010014
59011075004
59011075104
59011075204
59011075704
59011075804
59385001201
59385001230
59385001401
59385001430
59385001601
59385001630
59385002160
59385002260
59385002360
59385002460
59385002501
59385002560
59385002601
59385002660
59385002760
60429058630
60429058633
60429058730
60429058733
62175045232
62175045832
62756045983
62756046083
62756096983
62756097083
63481016160
63481020760
63481034860
63481051960
63481068560
63481082060
63481095260
63629403402
63629403403
63874108403
63874117303
65162041503
65162041603
66336001630
68071138003
68071151003
68258299103
68258299903
68308020230
68308020830
00490005160
00490005190
16590066605
16590066705
16590066790
23490927006
23490927009
43063018407
49999039507
54868570703
55887031204
55887031215
63629402801
63629403401
63629409201
63874108503
63874117403
66336001530

Appendix Table 6.

Comorbidities ICD-9 Codes

Mental Health Conditions 1	All	295.X, 296.X, 311.XX,648.4X, 300.0X, 309.8X
Chronic Conditions 2	Chronic Respiratory Disease	491.X – 496.X
	Chronic Hypertension	401.X-405.X, 642.0X, 642.1X, 642.2X, 6427.X
	Diabetes	249.X, 250.X, 648.0X
	Chronic Heart Disease	412.X-414.X, 394.X-397.X, 424.X, 428.22, 428.23, 428.32, 428.33, 428.42, 428.43, 745.0X-747.4X, 648.5X
	Chronic Renal Disease	581.X-583.X, 585.X, 587.X, 588.X, 646.2X
	Chronic Liver Disease	571.X, 572.X
	HIV	042.X, V08.X

Model Specification.

The multivariable logistic regression equation that we used in our analysis is provided below:

$$\ln \left(Y\right)={\beta }_0+{\beta }_{1}Race+{\beta }_{2}Age+{\beta }_{3}Sex+{\beta }_{4}MHC+{\beta }_{5}CC+{\beta }_{6}Urbanicity+{\beta }_{7}State+{\beta }_{8}Year$$

In this model, ln(Y) is the outcome variable of interest (natural log of the odds of sufficient duration, efficient dosage, concurrent prescribing of opioid analgesics, and concurrent prescribing of benzodiazepines). Race is a vector of variables indicating individual-level race/ethnicity, Age is a vector of variables indicating individual-level age group, Sex is an indicator of individual-level sex, MHC is an indicator of individual-level mental health comorbidities, CC is an indicator of individual-level chronic conditions, Urbanicity is a vector of variables indicating county-level urbanicity, State is a state fixed effect, and Year is a year fixed effect. Standard errors are clustered at the individual (patient) level to account for multiple treatment episodes for the same individual.