Psychosocial and behavioral therapy in conjunction with medication for opioid use disorder: Patterns, predictors, and association with buprenorphine treatment outcomes

1. Introduction

Current evidence indicates that pharmacotherapy is the most effective treatment for opioid use disorder (OUD) (Mattick et al., 2014). The strength of the evidence base is reflected in the steady shift away from the term “medication-assisted treatment” (MAT) toward “medications for opioid use disorder” (MOUD). The robust clinical trial data demonstrating the effectiveness of buprenorphine and other medications are also reflected in evidence-based recommendations that emphasize MOUD as central and psychosocial services as complementary in OUD treatment (American Society of Addiction Medicine, 2020; National Academies of Sciences, Engineering, and Medicine & National Academies of Sciences, Engineering, 2019; Substance Abuse and Mental Health Services Administration, 2018b), including international Guidelines for the Psychosocially Assisted Pharmacological Treatment of Opioid Dependence (World Health Organization, 2009). Due to extensive evidence on the effectiveness of pharmacotherapy and high rates of medication discontinuation, retention on MOUD is often selected as a treatment outcome in clinical trials (Hser et al., 2014).

Trial data for psychosocial treatment delivered alongside MOUD have some limitations, but largely show no added benefit in treatment retention or outcomes beyond MOUD itself (Amato et al., 2011; Dugosh et al., 2016; Wyse et al., 2021). Although select studies of psychosocial interventions focusing specifically on treatment engagement and continuity have demonstrated a positive relationship with buprenorphine retention (Brigham et al., 2014; Katz et al., 2011), whether this relationship extends to observational settings and populations that allow for broader outcome measures over a longer timeframe is unclear, and this relationship may differ in terms of care delivery (i.e., usual treatment without the support services often provided in trials) and patient composition (e.g., more complex medical and behavioral health comorbidities) (Moran et al., 2019). Furthermore, evidence is lacking on the relationship between patients’ characteristics and psychosocial service use and outcomes (Blanco & Volkow, 2019).

The limitations of the literature may explain, in part, the discrepancy between the overall evidence base and current treatment practices and policies. Expert opinion supports psychosocial and behavioral therapy as particularly important for OUD treatment engagement and retention (Moran et al., 2019), a perspective that is reflected in clinical guidelines and recommendations (American Society of Addiction Medicine, 2020; Substance Abuse and Mental Health Services Administration, 2018b; World Health Organization, 2009). The perceived importance of therapy and counseling is also represented in third-party payer policies stipulating psychosocial treatment or referral as a prerequisite for MOUD, often including state Medicaid programs (Substance Abuse and Mental Health Services Administration, 2018a).

As the payer group covering the largest share of addiction treatment costs in the United States, Medicaid plays a key role in OUD treatment (Medicaid and CHIP Payment and Access Commission (MACPAC), 2017; Zur & Tolbert, 2018). Medicaid is particularly central to community-based buprenorphine treatment, which is covered by all state programs (Grogan et al., 2016) and accounts for the majority of annual MOUD expenditures (Clemans-Cope et al., 2019). Importantly, Medicaid is also a major financer of behavioral health care in the United States, including public services. In addition to covering a disproportionate share of adults with OUD (Medicaid and CHIP Payment and Access Commission (MACPAC), 2017), Medicaid is the primary source of insurance for low-income Americans, who have elevated rates of psychiatric disorders compared to those with higher incomes (Sareen et al., 2011). Identifying treatment patterns in this group is critical to increase understanding of current practices and facilitate improvements in patient outcomes.

One analysis of Medicaid administrative records of adolescents found lower risk of attrition for buprenorphine (with or without psychosocial services) compared to psychosocial treatment alone (Hadland et al., 2018), but this study did not compare combined treatment to buprenorphine alone. A second study of Missouri Medicaid data examined buprenorphine received across settings defined as either “medical” or “psychosocial”, with results suggesting that treatment in medical settings was associated with lower risk of both buprenorphine discontinuation and substance-related hospital visits (Presnall et al., 2019). However, the largest group with combined buprenorphine and psychosocial treatment was excluded from these analyses (Presnall et al., 2019). These two studies are important additions to the literature on variation in patient and treatment characteristics across modalities and settings, but more research needs to fill gaps on treatment predictors, course, and outcomes for patients receiving concurrent buprenorphine and psychosocial services.

To improve understanding of the role of psychosocial and behavioral therapy in OUD treatment, this study examined services received in the first 6 months after initiating buprenorphine. We focused on this particularly high-risk period for medication discontinuation (Samples et al., 2018; Timko et al., 2016), as it represents a critical window to identify factors associated with treatment retention and could facilitate improvements along the cascade of care (Williams et al., 2018). The main objectives were to: 1) define longitudinal patterns of therapy services and identify baseline patient characteristics associated with these patterns, 2) describe the course of treatment across patient subgroups, and 3) examine the association between therapy service patterns and buprenorphine treatment retention.

2. Material and methods

2.1. Study design and data

We conducted a retrospective cohort study using the MarketScan Multi-State Medicaid database, with data collected longitudinally from 2013 to 2018 on millions of U.S. enrollees, including enrollment information, demographic characteristics, diagnoses, and comprehensive health services and prescription data.

2.2. Sample

The study sample comprised adults aged 18–64 years at buprenorphine initiation, defined as an index claim following at least a 180-day baseline period with no observed buprenorphine prescriptions. We excluded buprenorphine episodes <7 days in duration, which could represent detoxification rather than maintenance treatment. For patients with multiple treatment episodes, we included the first qualifying episode in analyses. To capture new therapeutic episodes in addition to new medication treatment episodes, we also excluded individuals with any OUD-related psychosocial or behavioral therapy services during the 180-day baseline period prior to initiating buprenorphine. Finally, we excluded older individuals and those otherwise eligible for Medicare as well as individuals with behavioral health and prescription drug coverage not captured in the database (Appendix Figure A1).

2.3. Measures

2.3.1. OUD treatment outcomes

We identified buprenorphine treatment using National Drug Codes (NDC) from prescription drug claims. The start date for new treatment episodes was the date on the index buprenorphine claim. We measured treatment duration using the number of days’ supply listed on buprenorphine claims. Similar to prior studies of buprenorphine treatment episodes (Mark et al., 2013; Samples et al., 2020), we defined the discontinuation outcome as >30 days without buprenorphine supply, using the last date with supply as the end of the treatment episode.

We identified psychosocial and behavioral therapy services using Healthcare Common Procedure Coding System (HCPCS) procedure codes from outpatient claims (American Medical Association, 2014; Centers for Medicare & Medicaid Services, 2003). The study team selected services based on the Healthcare Effectiveness Data and Information Set (HEDIS) measures of initiation and engagement in substance use treatment, excluding services that were outside the scope of psychosocial and behavioral therapy (Appendix Table A1) (National Committee on Quality Assurance, 2019). For example, services included individual, group, and family psychotherapy or counseling services and are referred to hereafter simply as “therapy”. Trajectory analyses were limited to therapy services with an OUD diagnosis to capture treatment episodes specifically focused on opioid addiction (Busch et al., 2020). However, we did not require OUD to be the primary diagnosis with the aim of capturing comprehensive OUD treatment addressing common psychiatric comorbidities (Jones & McCance-Katz, 2019), including substance use disorders for which medication treatment is not available. In addition, we conducted a sensitivity analysis, including all services with a mental health or substance use disorder (MHSUD) diagnosis (i.e., not limited to OUD-related therapy). We included patients without documented OUD diagnoses in the baseline period prior to buprenorphine initiation because a substantial proportion of patients receive buprenorphine for opioid maintenance without an OUD diagnosis (Gordon et al., 2015; Mark et al., 2013).

2.3.2. Pre-treatment patient characteristics

We measured pre-treatment baseline sociodemographic and clinical characteristics in the 180-day period prior to buprenorphine initiation. We identified sociodemographic characteristics using enrollment data on sex (male, female), age (years), race/ethnicity (non-Hispanic white, non-Hispanic Black, Hispanic, other), and insurance plan type (capitation or fee-for-service). Clinical characteristics included baseline medical and behavioral health comorbidities and health service use. We identified comorbidities using primary and secondary diagnosis codes from all service claims.

Mental health indicators included diagnoses of depression, anxiety, bipolar disorder, schizophrenia, and any other mental health disorder. Substance use disorders included indicators for opioids, alcohol, cannabis, tobacco, cocaine, sedatives, amphetamines, and any other drug (Appendix Table A2). Medical indicators included hepatitis C and a weighted composite measure of overall medical burden. We calculated a modified Elixhauser Comorbidity Index score that excluded diagnoses of hepatitis C and behavioral health disorders from the final index, which accounted for 25 medical comorbidities (e.g. HIV, hypertension, obesity, diabetes, cancer) that are associated with hospital-based resource use and mortality (Moore et al., 2017).

Baseline health service use included indicators for all-cause inpatient and emergency department (ED) services, opioid overdose, non–opioid drug overdose, and any prescription opioids or psychotropic drugs (i.e., antidepressants, antipsychotics, stimulants, mood stabilizers, benzodiazepines, sedatives).

2.3.3. Treatment characteristics

Because previous research has shown an association with treatment discontinuation (Samples et al., 2018), we measured the initial buprenorphine dosage, defined as the dosage on the date of the index claim. We measured the proportion of days covered (PDC) by buprenorphine during the treatment period because PDC is a quality measure of the Centers for Medicare and Medicaid Services for other chronic conditions with outcomes that depend on treatment continuity (Centers for Medicare & Medicaid Services, 2019) and past research has shown an association between buprenorphine PDC and adverse opioid-related outcomes (Ronquest et al., 2018).

To better understand the course of treatment, we also measured adverse health care events observed during the first 6 months of buprenorphine treatment. In addition to all-cause inpatient and ED services, opioid and non–opioid drug overdose, we measured opioid-related events treated in inpatient or ED settings, defined according to methods from the Agency for Healthcare Research and Quality (Healthcare Cost and Utilization Project [HCUP], 2018). Opioid-related events include inpatient or ED services with diagnosis codes for opioid use disorder, poisoning, and adverse effects.

2.4. Analyses

2.4.1. Therapy patterns and treatment course

We used group-based trajectory modeling (GBTM) to identify patterns of OUD-related therapy services received in the first 6 months after initiating buprenorphine. GBTM is a flexible approach that captures variation in the population to identify longitudinal latent classes with distinctive trajectory paths (Nagin, 1999). The study based estimates on the number of days with therapy services in each 30-day period. The team used a zero-inflated Poisson distribution to account for the large number of individuals with no observed therapy services. The study team based the final model selection on established statistical criteria (Appendix Table A3) (Nagin & Odgers, 2010). Within the sample, separate analyses estimated patterns of MHSUD-related therapy services to identify groups for sensitivity analyses.

The team calculated unadjusted sample statistics to show baseline sociodemographic and clinical characteristics across trajectory groups. Adjusted multinomial logistic regression estimated the associations of pre-treatment patient characteristics with trajectory group membership, reported as relative risk ratios (RR).

To examine adverse health care events observed during buprenorphine treatment, we calculated unadjusted rates to determine the number of individuals with events per 1,000 person-months of buprenorphine treatment for each trajectory group. Adjusted Cox proportional hazards regressions separately estimated the time to each event. In addition to adjusting for all pretreatment patient characteristics, models adjusted for the initial buprenorphine dose and the proportion of days covered by buprenorphine from initiation to the time of the event (or the end of the episode if the study observed no event).

2.4.2. Buprenorphine treatment duration

To examine the association of therapy patterns with buprenorphine treatment duration, adjusted Cox proportional hazards models estimated time to discontinuation across trajectory groups. We used propensity score weighting to account for differences across groups in observed pre-treatment covariates, including baseline sociodemographic and clinical characteristics, with the goal of improving comparability across groups and reducing bias in the effect estimates of buprenorphine discontinuation across trajectory groups (Rosenbaum & Rubin, 1983). Using established methods for multiple treatments described in detail elsewhere, the study calculated multinomial propensity score weights using generalized boosted modeling (McCaffrey et al., 2013) in the R twang (Toolkit for Weighting and Analysis of Nonequivalent Groups) package (Griffin et al., 2014) using the mnps (multinomial propensity score) function (Burgette et al., 2021). This method iteratively fits regression trees to capture nonlinearities in the relationship of pretreatment covariates to treatment group membership and minimizes imbalance across therapy groups. To assess balance across groups, we calculated the absolute standardized difference in means for each pretreatment covariate before and after propensity score weighting (Appendix Figure A2). We used average treatment effect (ATE) weights to generate estimates representing the difference in buprenorphine treatment duration if all patients followed a given therapy trajectory compared to all patients receiving no therapy. We then examined the association between therapy patterns and buprenorphine discontinuation using propensity score weighted Cox models, fully adjusting for baseline characteristics to obtain doubly-robust estimates (Funk et al., 2011) with additional adjustments for the initial buprenorphine dose and PDC.

We conducted two sensitivity analyses of the association between therapy patterns and buprenorphine treatment duration. In the first, we assessed the robustness of Cox estimates of the association between therapy patterns and buprenorphine treatment duration to unmeasured confounding (Appendix Tables A4 and A5) (VanderWeele & Ding, 2017). In the second, we excluded comparisons across therapy trajectory groups, instead including a time-varying indicator for OUD therapy to identify the overall association of OUD therapy with buprenorphine treatment duration in the full sample (Appendix Table A6 and Figure A3).

3. Results

3.1. Patterns of psychosocial and behavioral therapy services

In trajectory analyses, a three-group specification for therapy service use patterns performed well on all tests of model adequacy (Appendix Table A3). Figure 1 shows therapy trajectories in the first 6 months after initiating buprenorphine. Most patients were categorized in the group with little to no services (n=45,720; 73.8%), averaging <1 day with therapy services per month (mean=0.07). Approximately one-fifth followed a low-intensity trajectory (n=10,689; 17.2%), averaging about 2 service days per month (mean=2.06). Nearly one-tenth followed a high-intensity trajectory (n=5,567; 9.0%), averaging about 7 therapy service days per month (mean=6.94), with higher use in the first month (mean=8.26) declining over the study period to about 5 services in the sixth and final month (mean=5.28). Most OUD services were for individual, group, or family therapy/counseling (65.6%) followed by therapy/counseling as part of a behavioral health program (30.3%) (Appendix Table A7).

Figure 1.

Patterns of psychosocial and behavioral therapy in the first 180 days after buprenorphine initiation

Note: Estimates were derived from group-based trajectory models based on the number of days with therapy services in each 30-day (one-month) period

Results from sensitivity analyses of all MHSUD-related therapy services were similar to the main findings (Appendix Figure B1, Appendix Tables B1–B2).

3.2. Patient characteristics associated with receipt of therapy services

The left panel of Table 1 shows unadjusted baseline characteristics across trajectory groups with no therapy, low-intensity, and high-intensity therapy patterns. Across all groups, the sample initiating buprenorphine treatment was mostly female, with average age of 35 years, and overwhelmingly white (90%). However, compared to the group with no therapy, the low- and high-intensity groups had a significantly smaller proportion of females and a higher proportion of whites (Table 1).

Table 1.

Unadjusted and adjusted baseline characteristics associated with patterns of psychosocial and behavioral therapy services

	Unadjusted			Adjusted
	No therapy (n=45,720)	Low-intensity (n=10,689)	High-intensity (n=5567)	Low-intensity vs. No therapy	High-intensity vs. No therapy	High-intensity vs. Low-intensity
	%	%	%	RR (95% CI)	RR (95% CI)	RR (95% CI)
Sociodemographic characteristics
Sex
Female	38.0	40.1	41.3	Reference	Reference	Reference
Male	62.0	59.9	58.7	1.02 (0.97, 1.06)	1.05 (0.99, 1.12)	1.04 (0.97, 1.11)
Age	34.6 (9.3)	34.9 (9.2)	34.4 (9.2)*	1.01 (1.00, 1.01)	1.00 (0.99, 1.00)	0.99 (0.99, 1.00)
Race/ethnicity
White	89.5	90.9	91.2	Reference	Reference	Reference
Black	6.0	5.8	5.7	0.93 (0.84, 1.02)	0.84 (0.74, 0.95)	0.90 (0.78, 1.04)
Hispanic	0.6	0.4	0.5	0.62 (0.45, 0.87)	0.81 (0.55, 1.21)	1.31 (0.81, 2.11)
Other/Unknown	3.9	2.9	2.5	0.76 (0.67, 0.86)	0.64 (0.54, 0.77)	0.84 (0.69, 1.03)
Plan type
FFS	45.2	28.8	26.6 *	Reference	Reference	Reference
Capitation	54.8	71.2	73.4 *	1.96 (1.87, 2.05)	2.20 (2.06, 2.34)	1.12 (1.04, 1.21)
Clinical characteristics
Mental Health Diagnoses
Depression	18.3	19.0	23.3 *	1.04 (0.98, 1.11)	1.15 (1.07, 1.25)	1.11 (1.01, 1.21)
Anxiety	24.0	23.9	24.3	1.08 (1.02, 1.14)	0.93 (0.86, 1.01)	0.86 (0.79, 0.94)
Bipolar disorder	9.1	9.6	10.8 *	1.10 (1.01, 1.19)	0.98 (0.88, 1.09)	0.90 (0.79, 1.01)
Schizophrenia	1.6	1.6	2.3 *	1.07 (0.89, 1.28)	1.21 (0.98, 1.49)	1.13 (0.88, 1.45)
Other mental illness	10.4	10.6	13.2 *	1.02 (0.94, 1.10)	1.05 (0.96, 1.15)	1.03 (0.93, 1.15)
Substance Use Disorders
Opioids	47.4	68.1	79.2 *	2.36 (2.25, 2.47)	3.69 (3.44, 3.95)	1.56 (1.45, 1.69)
Alcohol	5.4	6.0	11.0 *	1.00 (0.91, 1.10)	1.56 (1.41, 1.72)	1.55 (1.37, 1.75)
Cannabis	3.9	5.8	8.7 *	1.33 (1.21, 1.47)	1.69 (1.51, 1.89)	1.27 (1.12, 1.44)
Tobacco	28.5	27.4	30.2*	0.91 (0.86, 0.96)	0.90 (0.84, 0.96)	0.99 (0.92, 1.07)
Cocaine	3.1	3.8	6.7 *	0.99 (0.88, 1.11)	1.37 (1.20, 1.55)	1.39 (1.19, 1.61)
Sedatives	1.8	2.0	3.4 *	0.99 (0.85, 1.16)	1.25 (1.05, 1.48)	1.26 (1.03, 1.54)
Amphetamines	2.4	2.8	3.6 *	0.98 (0.85, 1.12)	0.97 (0.83, 1.14)	0.99 (0.82, 1.20)
Other drugs	16.9	17.5	23.1 *	0.93 (0.87, 0.98)	0.99 (0.92, 1.07)	1.07 (0.98, 1.17)
Medical Comorbidities
Elixhauser Index	1.0 (3.9)	1.1 (3.9)	1.3 (4.1) *	1.00 (1.00, 1.01)	0.99 (0.99, 1.00)	0.99 (0.98, 1.00)
Hepatitis C	32.7	32.2	32.9	1.01 (0.93, 1.09)	1.19 (1.08, 1.31)	1.18 (1.05, 1.32)
Health Services
All-cause inpatient	17.2	17.5	26.6 *	0.84 (0.79, 0.90)	1.11 (1.03, 1.20)	1.31 (1.20, 1.44)
All-cause emergency department	53.8	56.6	63.2 *	1.13 (1.07, 1.19)	1.21 (1.13, 1.30)	1.07 (0.99, 1.16)
Opioid overdose	3.0	4.6	8.3 *	1.22 (1.08, 1.38)	1.72 (1.50, 1.96)	1.41 (1.21, 1.64)
Non-opioid drug overdose	3.4	4.1	6.6 *	0.97 (0.85, 1.10)	1.01 (0.87, 1.16)	1.04 (0.88, 1.23)
Prescription Drugs
Antidepressants	33.7	32.7	35.7*	0.95 (0.90, 1.00)	0.99 (0.92, 1.07)	1.05 (0.96, 1.14)
Antipsychotics	11.5	11.2	14.1 *	0.98 (0.90, 1.06)	1.13 (1.02, 1.25)	1.15 (1.03, 1.30)
Stimulants	5.3	4.1	3.8	0.90 (0.81, 1.01)	0.83 (0.71, 0.97)	0.92 (0.77, 1.09)
Mood stabilizers	22.4	23.6	23.8	1.09 (1.02, 1.15)	0.96 (0.89, 1.04)	0.88 (0.81, 0.97)
Benzodiazepines	22.2	16.6	17.5	0.72 (0.67, 0.77)	0.76 (0.69, 0.83)	1.06 (0.95, 1.17)
Sedatives	4.2	2.7	2.7	0.76 (0.67, 0.87)	0.71 (0.60, 0.85)	0.94 (0.77, 1.15)
Opioids	44.4	41.2	42.8	0.92 (0.88, 0.97)	0.96 (0.90, 1.02)	1.04 (0.96, 1.12)
Treatment characteristics
Initial buprenorphine dose, mean (SD)	14.5 (5.9)	13.3 (5.0)	12.8 (5.3) *
Initial dose ≥16mg	61.8	52.0	48.4 *
PDC, mean (SD)	0.93 (0.12)	0.91 (0.12)	0.91 (0.13)
PDC in first 180 days, mean (SD)	88.9	85.5	86.6
Episode duration (days), median	84	145	136
At least 180 days	31.9	43.3	42.4

Abbreviations: RR, relative risk ratio; CI, confidence interval

Note: Therapy groups were derived from group-based trajectory models. For unadjusted statistics, bold font indicates significance at p<0.05 compared to the no therapy group, and

^*indicates significance at p<0.05 compared to the low-intensity therapy group. Adjusted estimates were derived from multinomial logistic regression models, fully adjusted for pretreatment patient characteristics.

The right panel of Table 1 shows fully adjusted multinomial logistic regression estimates of the association between baseline characteristics and trajectory group membership. Compared to the group with no therapy, low- and high-intensity therapy patterns were associated with capitated Medicaid plans (low-intensity RR=1.96; 95% CI, 1.87–2.05; high-intensity RR=2.20; 95% CI, 2.06–2.34), documented OUD (low-intensity RR=2.36; 95% CI, 2.25-–2.47; high-intensity RR=3.69; 95%CI, 3.44–3.95), cannabis use disorder (low-intensity RR=1.33;95% CI, 1.21–1.47; high-intensity RR=1.69; 95%CI, 1.51–1.89), all-cause ED services (low- intensity RR=1.13; 95% CI, 1.07–1.19; high-intensity RR=1.21; 95%CI, 1.13–1.30), and opioid overdose history (low-intensity RR=1.22; 95% CI, 1.08–1.38; high-intensity RR=1.72; 95%CI, 1.50–1.96) at baseline prior to treatment initiation. Patients who were less likely to receive therapy included those with “other” race/ethnicity (low-intensity RR=0.76; 95% CI, 0.67–0.86; high-intensity RR=0.64; 95% CI, 0.54–0.77), tobacco use disorder (low-intensity RR=0.91; 95% CI, 0.86–0.96; high-intensity RR=0.90; 95% CI, 0.84–0.96), and benzodiazepine (low-intensity RR=0.72; 95% CI, 0.67–0.77; high-intensity RR=0.76; 95% CI, 0.69–0.83) or sedative (e.g., zolpidem) (low-intensity RR=0.76; 95% CI, 0.67–0.87; high-intensity RR=0.71; 95% CI, 0.60–0.85) prescriptions at baseline (Table 1).

Results of sensitivity analyses across MHSUD-related therapy groups were consistent with the main findings (Appendix Table B3), with additional baseline behavioral health diagnoses significantly associated with receipt of therapy in adjusted models (e.g., depression, bipolar disorder, schizophrenia, other mental illness, alcohol use disorder, and amphetamine use disorder).

3.3. Patients’ characteristics associated with intensity of therapy services

In unadjusted analyses comparing baseline characteristics of the high-intensity therapy group to the low-intensity group, no significant differences existed in terms of sex or race/ethnicity, but the high-intensity therapy group was slightly younger on average and a greater proportion had capitated Medicaid plans (Table 1). The study observed a gradient for most clinical characteristics, whereby the high-intensity group had significantly higher rates of behavioral health diagnoses, medical comorbidities, all-cause inpatient and ED service use, medically treated overdoses, and prescriptions for antidepressants or antipsychotics.

The far-right panel of Table 1 shows fully adjusted multinomial logistic regression estimates of the association between baseline characteristics and intensity of therapy services. Compared to the low-intensity group, high-intensity therapy was associated with capitated Medicaid plans (RR=1.12; 95% CI, 1.04–1.21); depression (RR=1.11; 95% CI, 1.01–1.21); documented OUD (RR=1.56; 95% CI, 1.45–1.69); and substance use disorder diagnoses for alcohol (RR=1.55; 95% CI, 1.37–1.75), cannabis (RR=1.27; 95% CI, 1.12–1.44), cocaine (RR=1.39; 95% CI, 1.19–1.61), and sedatives (RR=1.26; 95% CI, 1.03–1.54); hepatitis C (RR=1.18; 95% CI, 1.05–1.32); all-cause inpatient service use (RR=1.31; 95% CI, 1.20–1.44); medically treated opioid overdose (RR=1.41; 95% CI, 1.21–1.64); and prescription antipsychotics (RR=1.15; 95% CI, 1.03–1.30).

Results from sensitivity analyses comparing MHSUD-related therapy groups were consistent with the main findings (Appendix Table B3), though some adjusted associations with high-intensity therapy were no longer significant (e.g., depression, sedative use disorder, hepatitis C, prescription antipsychotics), and additional associations emerged (e.g., amphetamine use disorder, other drug use disorders).

3.4. Association of therapy services with buprenorphine discontinuation

The left panel of Table 1 shows unadjusted characteristics of buprenorphine treatment duration. The median length of buprenorphine treatment episodes was shortest for the group with no therapy (84 days) but comparable for the low-intensity (145 days) and high-intensity (136 days) groups. Similarly, a lower proportion of patients in the group with no therapy (31.9%) reached the 180-day benchmark for minimum duration of pharmacotherapy than in the low-intensity (43.3%) and high-intensity (42.4%) groups. Overall, about one-third of the sample had 180-day or longer buprenorphine episodes (n=21,578; 34.8%) and two-thirds of those with minimum duration treatment were in the group with no therapy (n=14,591; 67.6%).

Propensity score weighting achieved good balance on all covariates across therapy groups (Appendix Figure A2). Table 2 shows fully adjusted results from Cox proportional hazard models estimating time to buprenorphine discontinuation in the first 180 days after treatment initiation. In unweighted models, both low-intensity (HR=0.55; 95% CI, 0.53–0.57) and high-intensity (HR=0.61; 95% CI, 0.58–0.63) therapy were associated with reduced risk of discontinuation. In propensity score weighted models, these associations remained statistically significant, though the hazard of buprenorphine discontinuation for high-intensity therapy was attenuated (HR=0.71; 95% CI, 0.67–0.74).

Table 2.

Adjusted hazard of buprenorphine discontinuation in the first 6 months of treatment

	Unadjusted			Adjusted
	HR	95% CI	p-value	HR	95% CI	p-value
No therapy	Reference			Reference
Low-intensity therapy	0.55	0.53, 0.57	<0.001	0.55	0.54, 0.57	<0.001
High-intensity therapy	0.61	0.58, 0.63	<0.001	0.71	0.67, 0.74	<0.001

Abbreviations: HR, hazard ratio; CI, confidence interval

Note: Estimates were derived from fully adjusted Cox proportional hazards regression models estimating time to buprenorphine discontinuation; weighted estimates are based on multinomial propensity scores and represent the average treatment effect if all patients in the sample had followed a given therapy trajectory compared to all patients receiving no therapy.

In sensitivity analyses examining time-varying receipt of OUD-related psychosocial and behavioral services in the full sample (i.e., no comparisons across therapy trajectory groups), therapy was also associated with significantly reduced risk of buprenorphine discontinuation (HR=0.40; 95% CI, 0.37–0.43; Appendix Table A6; Appendix Figure A3).

Results from sensitivity analyses of MHSUD-related therapy groups and services were similar to the main findings (Appendix Figures B2–B3, Appendix Tables B3–B6).

3.5. Additional characteristics of the treatment course across therapy service patterns

In unadjusted analyses (Table 1), the study observed a gradient in mean initial buprenorphine dosage (none=14.5mg, low-intensity=13.3mg; high-intensity=12.8mg), though differences were minimal and may not be clinically meaningful. Similarly, the proportion of patients with initial dosage ≥16mg was highest for the group with no therapy (61.8%), lower for the low-intensity therapy group (52.0%), and lowest for the high-intensity group (48.4%). On average, all groups had a high proportion of days covered (PDC) by buprenorphine throughout the treatment episode, including in the first 180 treatment days (Table 1). Though both the low-and high-intensity groups had significantly lower mean PDC compared to the group with no therapy, these differences were also minimal.

Table 3 shows unadjusted rates and adjusted estimates of adverse health care events observed during the first 6 months after initiating buprenorphine. Both the low- and high-intensity therapy groups had significantly higher unadjusted rates of all-cause and opioid-related ED service use as well as medical treatment for opioid overdose. The high-intensity therapy group also had higher rates of all-cause and opioid-related inpatient service use as well as medical treatment for opioid overdose. In adjusted analyses, only the increased risk of all-cause ED services remained significant for both the low-intensity (HR=1.05; 95% CI, 1.01–1.09) and high-intensity (HR=1.16; 95% CI, 1.10–1.22) groups. For the high-intensity group, the increased risk of opioid-related ED services (HR=1.27; 95%CI, 1.10–1.46) and medical treatment for opioid overdose (HR=1.29; 95% CI, 1.01–1.64) remained significant. Additionally, the low-intensity group had significantly lower risk of all-cause (HR=0.92; 95% CI, 0.85–0.99) and opioid-specific (HR=0.82; 95% CI, 0.71–0.94) inpatient services in adjusted analyses.

Table 3.

Unadjusted and adjusted adverse health care events observed during buprenorphine treatment

	Unadjusted						Adjusted
	No therapy (PM=146,939)		Low-intensity (PM=43,435)		High-intensity (PM=21,699)		Low-intensity vs. No therapy	High-intensity vs. No therapy	High-intensity vs. Low-intensity
	N	Rate per 1000 PM	N	Rate per 1000 PM	N	Rate per 1000 PM	RR (95% CI)	RR (95% CI)	RR (95% CI)
Inpatient
All-cause	2919	20.7	834	20.1	519	25.3 *	0.92 (0.85, 0.99)	0.96 (0.87, 1.06)	1.04 (0.94, 1.17)
Opioid-related	959	6.6	266	6.2	240	11.3 *	0.82 (0.71, 0.94)	1.15 (0.99, 1.34)	1.40 (1.18, 1.67)
Emergency department
All-cause	12,402	105.6	3681	110.0	2121	134.9 *	1.05 (1.01, 1.09)	1.16 (1.10, 1.22)	1.11 (1.05, 1.17)
Opioid-related	1071	7.4	390	9.2	280	13.3 *	1.10 (0.98, 1.24)	1.27 (1.10, 1.46)	1.15 (0.99, 1.35)
Overdose treatment
Opioids	333	2.3	122	2.8	98	4.6 *	1.03 (0.83, 1.27)	1.29 (1.01, 1.64)	1.25 (0.96, 1.64)
Non-opioid drugs	531	3.6	152	3.5	106	4.9 *	0.88 (0.73, 1.06)	0.99 (0.79, 1.23)	1.13 (0.88, 1.45)

Abbreviations: PM, person-months; HR, hazard ratio; CI, confidence interval

Note: For unadjusted rates, bold font indicates significance at p<0.05 compared to the no therapy group, and

^*indicates significance at p<0.05 compared to the low-intensity therapy group. Adjusted estimates were derived from Cox proportional hazards regression models for each outcome, adjusted for all pre-treatment patient characteristics, initial buprenorphine dosage, and proportion of days covered by buprenorphine from initiation until the event of interest or buprenorphine discontinuation if no event was observed.

In adjusted analyses compared to the low-intensity therapy group, rates of all adverse health care events observed during the first 6 months after buprenorphine initiation were significantly higher for the high-intensity group (Table 4). However, in adjusted analyses, only the increased risk of opioid-related inpatient services (HR=1.40; 95% CI, 1.18–1.67) and all-cause ED services (HR=1.11; 95% CI, 1.05–1.17) remained significant (Table 4).

Results of sensitivity analyses across MHSUD-related therapy groups were consistent with the main findings (Appendix Tables B3, B7), with fewer significant associations in adjusted analyses. Compared to the group with no therapy, low-intensity therapy was associated only with higher risk of all-cause ED services, while high-intensity therapy was associated only with higher risk of opioid-related inpatient and ED services. Compared to the low-intensity group, high-intensity therapy was associated only with higher risk of opioid-related inpatient services.

4. Discussion

In this study, we examined the association of concurrent OUD-related psychosocial and behavioral therapy services with buprenorphine treatment retention during a high-risk period for attrition. We found that patients initiating and remaining on buprenorphine for at least 7 days generally followed one of three therapy trajectories in the first six months after initiating medication. The majority of patients had little to no therapy (74%), while roughly 17% and 9% had low-intensity and high-intensity therapy trajectories, respectively. Overall rates of buprenorphine receipt with and without therapy are consistent with prior research of privately insured patients in which about 69% of patients received MOUD alone and 31% received MOUD with psychosocial treatment in a given year from 2013 to 2016 (Huskamp et al., 2020). In a study of Missouri Medicaid patients with OUD, about 25% of patients received buprenorphine in a psychosocial setting and 75% in a general clinical setting, where psychosocial services were less prevalent but not completely absent (Presnall et al., 2019).

In analyses of buprenorphine retention among patients with at least 7 days of treatment, receiving OUD-related psychosocial and behavioral therapy services was associated with lower risk of medication discontinuation in the first 180 days after treatment initiation, raising the possibility that concurrent therapy may help to reduce high attrition rates commonly observed early in the course of treatment (Samples et al., 2018; Timko et al., 2016). Notably, the association of therapy services with reduced risk of buprenorphine discontinuation was consistent across estimates that weighted patients with different service patterns to have similar pre-treatment characteristics and estimates that assessed therapy continuously over time regardless of overall therapy trajectories, strengthening the implications for concurrent psychosocial and behavioral therapy services as one potential approach for addressing premature buprenorphine discontinuation. These findings are consistent with expert opinion that psychosocial services are important for treatment engagement and continuity (Moran et al., 2019) and with some trial data showing a positive relationship between buprenorphine treatment duration and psychosocial services that focus explicitly on engagement and retention (Brigham et al., 2014; Katz et al., 2011). However, the observational design of this study limits causal inferences, and future prospective research should determine whether therapy is effective for extending buprenorphine retention.

While fewer individuals in the group with minimal or no therapy services reached the benchmark for minimum duration of pharmacotherapy, a substantial proportion had 180 days or more of buprenorphine treatment, indicating that many patients persist in treatment with medication alone. Thus, insurance policies requiring referral or receipt of psychosocial services to receive buprenorphine or other MOUD may create overly restrictive barriers to highly effective medication treatment. Furthermore, many patients have limited or no access to such services due to fragmentation of health care, workforce shortages, underfunding, overburdened providers and programs, and individual barriers to health care (e.g., transportation difficulties, employment/family obligations). Ensuring adequate availability of psychosocial support services is important to facilitate access for patients who could benefit from therapy, particularly with recent expansions in telehealth services that could address current barriers to care (Hughto et al., 2021). To improve alignment of treatment needs and services, more research should help us to understand the complex relationships among patient characteristics, care components, and outcomes.

We found that receipt of therapy services was associated with baseline behavioral health diagnoses, including opioid and cannabis use disorders and, for those receiving high-intensity therapy, alcohol, cocaine, and sedative use disorders. In addition, receipt of therapy was associated with medical treatment for opioid overdose at baseline. These findings are consistent with prior research showing higher rates of mental health and substance use problems among patients with combined medication and psychosocial treatment (Huskamp et al., 2020; Presnall et al., 2019), indicating that patients who receive therapy in conjunction with buprenorphine may have more complex behavioral health treatment needs. Furthermore, high-intensity service patterns were associated with comorbid addiction to substances for which no FDA-approved medications are currently available (e.g., cocaine), potentially reflecting the limited treatment options available to patients with multiple substance use problems.

Patients receiving high-intensity therapy also had higher risk of opioid-related health care events during buprenorphine treatment episodes, suggesting that these patients also have more complex treatment courses. Considering the elevated risk profile in this group, these findings could indicate that patients with more complex behavioral health problems require more intensive or longer duration treatment to stabilize. In addition, the increased risk of opioid overdose and opioid-related ED care during buprenorphine treatment may signal a need for emerging hospital-based interventions to include protocols for care coordination and communication with providers of patients currently receiving OUD treatment in addition to OUD treatment induction and referrals. While these analyses adjusted for a range of pretreatment (baseline) and buprenorphine treatment (initial dosage, PDC) characteristics, they were not propensity-score weighted. Thus, due to potential residual confounding across therapy groups, we cannot draw causal inferences concerning the effects of therapy on risks of these adverse outcomes.

This study and prior research demonstrating the benefits of psychosocial and behavioral therapy in observational settings with more generalizable patient populations may have implications in an era of increasingly personalized medicine, with a growing emphasis on tailoring services to meet individual needs of patients. Overall, patients with greater or more complex treatment needs received more intensive services, which may contribute to improved retention in these groups during a period with high risk of attrition. Future research should continue to examine variation across patient subgroups and therapy services in the relationship of therapy to buprenorphine retention. For example, patients who have comorbid substance use disorders that lack effective medication options could represent a priority group for psychosocial services. This work is critical to align patient needs with evidence-based treatment to improve retention and patient outcomes.

4.1. Limitations

We could not ascertain specific types of therapy because the data do not include codes indicating, for example, cognitive behavioral therapy or contingency management. Some prior studies have compared the effectiveness of various therapeutic approaches (Wyse et al., 2021), but more work is needed in this area. Furthermore, the data do not capture services that are not reimbursed by Medicaid and therefore do not generate a Medicaid claim, such as self-help or peer-support services, brief or limited services delivered as part of medical management, and services funded by block grants or other non-Medicaid sources. However, underestimating service use would have resulted in more conservative effect estimates. In addition, we could not distinguish buprenorphine treatment characteristics across settings because prescription drug data do not include inpatient medication dispensing or the specific outpatient setting (e.g., primary care vs. opioid treatment program). Although we adjusted for a wide range of individual characteristics and propensity score methods achieved good balance across groups, residual confounding could result from unobserved factors. To further address this concern, sensitivity analyses to determine the robustness of results (VanderWeele & Ding, 2017) indicated that unmeasured confounding was unlikely to explain away the associations between therapy patterns and buprenorphine discontinuation. Additionally, results may not generalize to patients who discontinue buprenorphine in the first week of treatment, as the sample was restricted to episodes at least 7 days in duration. Finally, findings from Medicaid data may not be generalizable to individuals with no insurance or private coverage and, although the study drew the data from a large multi-state sample, the findings may not be representative of all Medicaid enrollees.

5. Conclusion

This is the first known study characterizing longitudinal trends in psychosocial and behavioral therapy received alongside buprenorphine in the treatment of adults with opioid use disorder. Most patients received little or no therapy services. Among those with concurrent therapy, the intensity of services corresponded to indicators of treatment need, potentially signifying appropriate alignment between patient characteristics and clinical practices. In addition, receipt of psychosocial and behavioral therapy was associated with reduced risk of early buprenorphine discontinuation, suggesting a possible protective role of therapy in improving treatment retention and potentially other outcomes for patients.

HIGHLIGHTS.

• Most buprenorphine patients receive no psychosocial or behavioral therapy services

• Concurrent therapy is associated with indicators of clinical need and complexity

• Service patterns suggest therapy intensity aligns with behavioral health problems

• Therapy services may improve buprenorphine retention, but future research is needed