Medication for opioid use disorder treatment and specialty outpatient substance use treatment outcomes: Differences in retention and completion among opioid-related discharges in 2016

Abstract

Opioid medication treatment access is a public health priority aimed to improve opioid use treatment outcomes. However, Medicaid does not cover all forms of MOUD, particularly methadone, in many states. We examined associations between medication for opioid use disorder (MOUD) plans and substance use treatment discharge reason (e.g., completed treatment, dropped out of treatment) as well as treatment retention (i.e., length of stay), and estimated whether these relationships were modified by state Medicaid methadone coverage. Data from the 2016 Treatment Episode Data Set for Discharges (TEDS-D) included 152,196 opioid-related treatment episodes from 47 states using relative risk regression with state clustering. Discharges involving MOUD had higher treatment retention for 180+ days (aRR: 1.60, 95% CI: 1.29, 1.99) and 365+ days (aRR: 2.64, 95% CI: 2.00, 3.49) but lower treatment completion (aRR: 0.46, 95% CI: 0.38, 0.57). There was no evidence that state Medicaid methadone coverage modified any of these relationships. Focusing on treatment completion alone may obscure health benefits associated with longer MOUD treatment retention.

1. Introduction

1.1. Medication for opioid use disorder

Medication for opioid use disorder (MOUD) is considered the gold standard for opioid use disorder (OUD) treatment and has consistently been found to be more effective than behavioral treatments alone in treating OUD (Connery, 2015; Simoens, Matheson, Bond, Inkster, & Ludbrook, 2005). People with OUD who receive MOUD have lower risk of mortality, enhanced ability to seek stable employment opportunities, and decreased number of criminal arrests (SAMHSA, 2015; Deck et al., 2009; Larochelle et al., 2018). Despite this evidence of improved health and social outcomes, MOUD is underutilized (Vestal, 2016). Of individuals with OUD, 19.4% received any type of treatment based on a nationally representative survey sample from 2005–2013 (Wu, Zhu, & Swartz, 2016) and less than 10% received MOUD treatment specifically (Nosyk et al., 2013). As only about 36% of specialty substance use treatment facilities provide MOUD (Mojtabai, Mauro, Wall, Barry, & Olfson, 2019), studies should estimate whether structural factors, such as Medicaid coverage of MOUD, influence not only access but also outcomes of MOUD.

1.2. State-level Medicaid coverage of MOUD

Medicaid provides funding for one-third of all substance use treatment programs in the United States (Clark et al., 2015). However, more state and federal funding for MOUD is necessary to fill the gap in OUD treatment need (Jones, Campopiano, Baldwin, & McCance-Katz, 2015). Medicaid programs cover buprenorphine in all 50 states (MACPAC, 2016; Kaiser Family Foundation, 2019; Grogan et al., 2016), which can be prescribed in private offices (SAMHSA, 2015). In contrast, as of September 2015 there were 20 states (MACPAC, 2016) in which Medicaid did not cover methadone—one of the most widely used forms of MOUD—which can be administered only through highly regulated certified opioid treatment programs. Indeed, nearly all (84%) people recorded as using MOUD were in facilities dispensing methadone over other forms of MOUD in the 2016 National Survey of Substance Abuse Treatment Services, excluding the Department of Veterans Affairs or Department of Defense typically excluded from the Treatment Episode Data-Set (TEDS) (SAMHSA, 2017; Batts et al., 2014). Thus, understanding whether there is an impact of Medicaid methadone coverage on treatment outcomes in light of the overdose epidemic is a public health priority.

According to two studies of specialty substance use treatment episodes, admissions from states with Medicaid coverage of methadone were more likely to include MOUD in their treatment plans compared to those from states without Medicaid methadone coverage (Bachhuber, Mehta, Faherty, & Saloner, 2017; Saloner, Stoller, & Barry, 2016). Even among individuals who did not have Medicaid insurance, nonintensive outpatient opioid specialty treatment admissions from states with Medicaid methadone coverage were more likely to have MOUD included in the treatment plans compared to those from states without coverage (69.5% versus 48.6%) (Bachhuber et al., 2017). In a recent study, opioid treatment admissions in states with higher criminal justice treatment referrals were less likely to include MOUD in the treatment plan, even after accounting for an individual-level criminal justice referral source (Mantha et al., 2019), indicating that state-level system factors affect MOUD treatment plans. While studies have reported the positive relationship between state Medicaid methadone coverage and MOUD treatment plans (Bachhuber et al., 2017; Saloner et al., 2016; Stein et al., 2018), the impact that Medicaid methadone coverage may have on treatment outcomes is understudied. Previous literature has shown that Medicaid expansions do not just impact care access, and can have spillover effects on health outcomes ranging from improved overall mortality rates and cardiovascular disease mortality rates (Khatana et al., 2019; Miller S, Altekruse S, Johnson N, & Wherry LR, 2019), cigarette smoking cessation treatment outcomes (Koma, Donohue, Barry, Huskamp, & Jarlenski, 2017), to lower opioid overdose mortality (Kravitz-Wirtz et al., 2020).

1.3. Evaluating MOUD treatment outcomes

The National Institute on Drug Abuse recommends methadone treatment maintenance for at least 12 months, but OUD treatment plans may vary based on multiple clinical factors (e.g., psychiatric comorbidities or OUD severity) (NIDA, 2018). Retention in care is an important goal for individuals receiving substance use treatment, including those receiving any type of MOUD (i.e., methadone, buprenorphine, or naloxone); therefore, dropout before treatment completion is of concern to researchers (Marcovitz, McHugh, Volpe, Votaw, & Connery, 2016; Timko, Schultz, Cucciare, Vittorio, & Garrison-Diehn, 2016). Dropout from MOUD after only 12 weeks of treatment has been associated with up to a 90% likelihood of opioid use relapse (Weiss et al., 2011; Woody et al., 2008). Regardless of treatment completion, longer retention in MOUD is associated with decreased likelihood of opioid relapse (Clark et al., 2015). After 12 months of sustained methadone treatment, an observational study found the prevalence of any heroin use to decrease from 91% prior to treatment to 24% (Hubbard, Craddock, & Anderson, 2003). Previous studies have used 180 days (i.e., 6 months) as an indicator of premature discharge to this target goal of at least 12 months of methadone maintenance (Proctor et al., 2015).

Inconsistencies in measurement of MOUD treatment outcomes pose challenges to establishing adequate treatment effectiveness standards (Sanger et al., 2018). Various studies of substance use treatment admissions have examined treatment outcomes based on discharge reason (e.g., treatment completed, terminated by facility) (Garrison, Sahker, Yeung, Park, & Arndt, 2018; Krawczyk, Feder, Saloner, et al., 2017; Mennis & Stahler, 2016; Sahker, Toussaint, Ramirez, Ali, & Arndt, 2015; Stahler, Mennis, & DuCette, 2016). Other studies have focused on treatment retention (e.g., length of stay in treatment) as a means of operationalizing treatment outcomes (Hubbard et al., 2003; Proctor et al., 2015; Timko et al., 2016). To our knowledge only one other study measured these outcomes concurrently but did not examine MOUD treatment plans or focus on state Medicaid methadone coverage, as we do here (Mennis, Stahler, El Magd, & Baron, 2019).

By utilizing national substance use treatment facility discharge data, we aim to 1) describe the relationship between having MOUD in the individual treatment plan and long-term substance use treatment retention (≥ 180 days and ≥ 365 days), and to 2) estimate the relationship between having MOUD in the individual treatment plan and substance use treatment completion, based on discharge reason. In each aim, we also assessed whether these associations were modified by state Medicaid methadone coverage. As we were interested in observing potential spillover effects of state Medicaid methadone coverage on OUD treatment outcomes, we did not restrict analyses to Medicaid enrollees. We hypothesized that having MOUD in the individual treatment plan would increase the likelihood of both substance use treatment retention and treatment completion. We also expected that these relationships would be modified by Medicaid state methadone coverage, regardless of individual health insurance coverage type, with Medicaid-covered states having increased likelihood of treatment retention and completion.

2. Material and methods

2.1. Data sources

This cross-sectional study utilized public-use data from the Treatment Episode Data Set: Discharges (TEDS-D), a dataset that annually obtains substance use treatment discharges in the United States, including the District of Columbia and Puerto Rico. A study using 2007–2010 data estimated that 1.9 million people received treatment based on TEDS, and 2.5 million people received specialty treatment based on NSDUH, suggesting that TEDS could reflect 77% of people in specialty treatment. However, this same report found that the TEDS admissions count was 56% of admission counts based on the N-SSATS facility data (Batts et al., 2014). TEDS-D are administrative records that cannot be linked to individuals. Thus, an individual may appear multiple times within one dataset (SAMHSA, 2018a).

State-level Medicaid methadone coverage was based on data from the Medicaid and Children’s Health Insurance Program (CHIP) Payment and Access Commission (MACPAC) review of state Medicaid plans and methadone coverage as of September 2015 to ensure the policy data predated the 2016 TEDS-D information (MACPAC, 2016). State-level aggregate measures of demographic information were obtained from the 2010 United States Decennial Census (US Census Bureau, 2017).

2.2. Study population

Our study focused on the 2016 TEDS-D, which included 47 states, the District of Columbia, and Puerto Rico. Discharges for the 2016 calendar year that were received by March 16, 2018, from the continental United States were included (N=1,452,318 total discharges). Georgia, Oregon, and West Virginia did not report a sufficient amount of data to be included in the 2016 TEDS-D (SAMHSA, 2018b). Discharges were restricted to those that had heroin or opioids/synthetics reported as the primary, secondary, or tertiary substance at the time of admission (N=575,940 opioid discharges). From these, we restricted to ambulatory nonintensive outpatient service settings (e.g., individual or group services representing most of substance use disorder specialty services) (SAMHSA, 2017) to isolate discharges where MOUD is most likely to be administered (N=226,770 opioid-related outpatient discharges). Discharges were not restricted based on health insurance status since we aimed to examine potential spillover effects of state Medicaid coverage on substance use treatment outcomes.

After applying the inclusion criteria, we excluded 0.83% with missing data for our MOUD in the treatment plan exposure variable. We restricted all models assessing treatment completion and treatment retention at 180+ days and 365+ days to those with a value for our treatment outcome variable; thus, discharges with a discharge reason other than those included within our definition (i.e., transferred to another substance use treatment program, incarcerated, death, other) were excluded. Our final analytic sample included 152,196 opioid-related discharges from 47 states in 2016.

2.3. Measures of interest

The primary exposure of interest was individual MOUD in the treatment plan (i.e., engagement in MOUD treatment). Individual MOUD in a treatment plan was based on a TEDS-D field that stated “whether the use of opioid medications such as methadone or buprenorphine will be part of the client’s treatment plan.”

We examined treatment retention with two binary variables: (1) length of stay greater than or equal to 180 days versus length of stay fewer than 180 days, and (2) length of stay greater than or equal to 365 days versus length of stay fewer than 365 days. Treatment retention variables were based on the length of stay measure using the admission date and last date of communication with the facility.

We based substance use treatment completion on the documented treatment outcome or reason for discharge. Response options were (1) treatment completed, (2) left against professional advice, (3) terminated by facility, (4) transferred to another substance use treatment program or facility, (5) incarcerated, (6) death, or (7) other. Similar to other studies (Mennis & Stahler, 2016; Scalise, 2010), we created a dichotomous substance use treatment completion variable to distinguish between an explicit favorable and unfavorable outcome: (1) treatment completed or (0) dropped out of treatment/terminated by facility. Discharges were coded as “treatment completed” when all parts of the treatment plan were successfully finished, “dropped out of treatment” when patients did not complete the program at their own will regardless of advice from health professionals to continue treatment, and “terminated by facility” when patients exhibited nonadherent behavior (SAMHSA, 2018b). We also operationalized discharge reason in another way: (1) treatment completed/transferred to another treatment program and (0) dropped out of treatment/terminated by facility. Transferring to another treatment program could occur to maintain an appropriate level of services, such as to continue treatment at a higher level of care, but TEDS-D did not have details regarding reason for transferring to another facility.

We examined the following individual-level confounders. Age categories were: age 12 to 17, 18 to 20, 21 to 24, 25 to 34, 35 to 49, 50 to 64, or 65 and older. We defined sex as male, female, or missing. We analyzed a race/ethnicity variable with the following categories: non-Hispanic white, white Hispanic, non-Hispanic black, non-Hispanic, Hispanic non-white, Native Hawaiian/Native American/Alaskan Native/Asian or Pacific Islander, non-Hispanic multi-race, or missing. Health insurance was an optional reporting field (69.5% missing), so we did not include individual insurance status as a confounder; this field also did not specify whether insurance coverage included behavioral health services.

We used state Medicaid methadone coverage as the effect modification variable, coded as a dichotomous variable: (1) state Medicaid covered methadone as of September 2015 and (0) state Medicaid did not cover methadone as of September 2015. State-level aggregate measures of demographic information from the 2010 Census included state-level percentage male, percentage non-Hispanic white, percentage age 10 to 24 years old, percentage without high school education, and total state population.

2.4. Statistical analysis

We first described the prevalence of discharges with MOUD in the treatment plan by select demographic characteristics. To include the treatment episodes that would have otherwise been excluded using a list-wise deletion for individual covariates, missing data were included for covariates using a missingness indicator (e.g., for the sex variable there are female, male, and missing categories). We used generalized linear models with a log link, specified Poisson family, and robust standard errors to account for state clustering to regress substance use treatment completion and treatment retention on having MOUD listed in the treatment plan. We adjusted models for age, sex, race/ethnicity, and state-level demographics mentioned above. We then tested for effect modification by state Medicaid methadone coverage. A p-value less than 0.05 was determined to be statistically significant. Data were analyzed using STATA 15MP (StataCorp, 2017).

2.5. Sensitivity analyses

First, we conducted a sensitivity analysis on state-level Medicaid methadone coverage based on data from the Kaiser Family Foundation for fiscal year 2017 (which ran from July 1, 2016, through June 30, 2017) since some states may have gained coverage after September 2015. This data source included all states except Arkansas and Illinois, bringing the analytic sample to 149,231 discharges (Kaiser Family Foundation, 2019). Second, to assess whether opioid admission type changed our results, we analyzed the data restricting to heroin only reported as the primary, secondary, or tertiary substance at the time of admission (N=104,998). Another TEDS study found racial/ethnic disparities in MOUD treatment plans to be greater among people reporting heroin use compared to other opioids (Krawczyk, Feder, Fingerhood, & Saloner, 2017). Third, we examined whether findings would change if we restricted the inclusion criteria to those with opioids as the primary substance at time of admission, which was 83.4% (n=126,957) of opioid-related discharges (i.e., listing heroin/prescription opioids as primary, secondary, or tertiary substance). Fourth, since multiple discharges may appear in the dataset for one individual, we conducted analyses adjusting for prior admission within our models, coded as a binary variable with the following response options: (1) no prior admission and (2) one or more prior treatment episodes. Fifth, we also examined a model adding psychiatric comorbidity as a potential confounder, since psychiatric comorbidity can impact both complexity of the MOUD treatment plan and subsequent treatment success (Krawczyk, Feder, Saloner, et al., 2017). Psychiatric comorbidity was defined as whether the client experienced a psychiatric problem in addition to their substance use issue. Psychiatric comorbidity was not in the primary analyses because it was broad and nonspecific. Sixth, since another study using TEDS data excluded states with 0% prevalence of the exposure measure indicating a potential reporting issue (Krawczyk, Picher, Feder, & Saloner, 2017), we ran analyses excluding Kansas, Montana, North Dakota, and Oklahoma, which had 0% of treatment episodes with receiving MOUD, resulting in an analytic sample of 150,117 discharges. Since all four of these states did not have Medicaid methadone coverage, we found it more appropriate to include these states in our primary analyses.

3. Results

3.1. Sample population and demographics

Of the 152,196 opioid-related discharges included in our final analytic sample, 43.4% had MOUD listed in the treatment plan. As Table 1 shows, 42.0% of discharges were from people 25 to 34 years old. A higher percentage of overall discharges were men; however, fewer men had MOUD listed in their treatment plans compared to women (43.4% versus 56.6%, respectively). The majority of all discharges were from individuals who were recorded as non-Hispanic white race/ethnicity (69.6%). Of the states included within our study sample, 31 of the 47 states’ Medicaid covered methadone. Notably, 86% of all discharges were from states where Medicaid covered methadone. While 47.7% of discharges from states with Medicaid methadone coverage had MOUD in the treatment plans, only 16.4% of states without Medicaid methadone coverage had MOUD in the treatment plans.

Table 1:

Descriptive characteristics of discharges by medication for opioid use disorder (MOUD) treatment plan status, 2016 Treatment Episode Data Set- Discharges (TEDS-D)

Characteristic	Overall	MOUD in treatment plan	Did not have MOUD in treatment plan
	N= 152,196 (col. %)	N= 65,988 (43.4%) (row %)	N= 86,208 (56.6%) (row %)
Age
12-17 years	1,478 (1.0%)	48 (3.3%)	1,430 (96.8%)
18-20 years	4,237 (2.8%)	1,101 (26.0%)	3,136 (74.0%)
21-24 years	18,116 (11.9%)	6,038 (33.3%)	12,078 (66.7%)
25-34 years	63,856 (42.0%)	25,155 (39.4%)	38,701 (60.6%)
35-49 years	43,718 (28.7%)	21,076 (48.2%)	22,642 (51.8%)
50-64 years	19,615 (12.9%)	11,735 (59.8%)	7,880 (40.2%)
65+ years	1,176 (0.8%)	835 (71.0%)	341 (29.0%)
Sex
Male	90,556 (59.5%)	39,306 (43.4%)	51,250 (56.6%)
Female	61,613 (40.5%)	26,671 (43.3%)	34,942 (56.7%)
Missing	27 (0.02%)	11 (40.7%)	16 (59.3%)
Race/ethnicity^
NH White	105,954 (69.6%)	40,777 (38.5%)	65,177 (61.5%)
White Hispanic	7,207 (4.8%)	3,408 (47.3%)	3,799 (52.7%)
NH Black	14,276 (9.4%)	7,697 (53.9%)	6,579 (46.1%)
NH Native/Asian PI	3,016 (2.0%)	1,302 (43.2%)	1,714 (56.8%)
NH Multi-Race	3,999 (2.6%)	1,725 (43.1%)	2,274 (56.9%)
Hispanic Non-White	13,211 (8.7%)	8,577 (64.9%)	4,634 (35.1%)
Missing	4,533 (3.0%)	2,502 (55.2%)	2,031 (44.8%)
Methadone Covered by Medicaid
Yes	130,899 (86.0%)	62,491 (47.7%)	68,408 (52.3%)
No	21,297 (14.0%)	3,497 (16.4%)	17,800 (83.6%)

Note:

^{^}NH= Non-Hispanic; PI=Pacific Islander, includes Asian PI, Native Hawaiian, Native American, Alaskan Native Non-Hispanic; excludes Georgia, Oregon, West Virginia, Puerto Rico, and Washington District of Columbia; col.=column. All discharges were restricted to ambulatory non-intensive outpatient service setting admissions that had heroin or opioids/synthetics reported as the primary, secondary, or tertiary substance at the time of admission.

Median length of stay for opioid-related discharges was 34 days (interquartile range: 31 to 36 days); 31.2% had a length of stay greater than 180 days and 15.2% greater than 365 days. Overall, 29.5% completed treatment, while 46.7% either completed or were transferred to another treatment program. Treatment completion was 16.9% for discharges with MOUD listed in the treatment plan and 39.0% for those without MOUD plans. Similarly, treatment completion or transferring to another treatment program was lower in discharges receiving MOUD versus those without MOUD in the treatment plan (38.5% versus 53.2%, respectively). Of discharges retained in treatment for greater than 180 days and greater than 365 days, only 36.3% and 31.7%, respectively, completed overall substance use treatment (Figure 1). Of all discharges with a length of stay of at least 365 days, there were 15,903 discharges (68.3%) that did not complete substance use treatment.

Figure 1:

Substance use treatment outcome by treatment retention and medication for opioid use disorder receipt among ambulatory non-intensive outpatient service setting admissions that had heroin or opioids/synthetics reported as the primary, secondary, or tertiary substance at the time of admission, 2016 Treatment Episode Dataset- Discharges (TEDS-D) (N= 152,196)

Note: excludes Georgia, Oregon, West Virginia, Puerto Rico, and Washington District of Columbia

Limited to discharges with a value for treatment completion or dropped out/terminated by facility.

3.2. Multivariable regression results

Table 2 shows estimates of treatment completion and treatment retention at 180+ days and 365+ days by MOUD and state Medicaid methadone coverage. Listing MOUD in the treatment plan significantly decreased the likelihood of overall substance use treatment completion (adjusted risk ratio [aRR]: 0.46, 95% confidence interval [CI]: 0.38, 0.57), relative to dropped out of treatment/terminated by facility. On the contrary, discharges with MOUD in the treatment plan had a higher likelihood of a longer length of stay, both for retention longer than 180+ days (aRR: 1.60, 95% CI: 1.29, 1.99) and retention longer than 365+ days (aRR: 2.64, 95% CI: 2.00, 3.49), relative to shorter retention than each of these cutoffs. State-level Medicaid methadone coverage did not significantly modify the relationship between having MOUD listed in individual treatment plans and treatment completion, treatment retention at 180+ days, or 365+ days; moderation results are seen in models 1, 2, and 3 of appendix Table 1B (aRR for interaction: 0.82, 95% CI: 0.54, 1.27; aRR for interaction: 1.19, 95% CI: 0.88, 1.61; aRR for interaction: 1.22, 95% CI: 0.76, 1.95, respectively).

Table 2:

Associations between MOUD and substance use treatment completion/retention, 2016 Treatment Episode Data Set- Discharges (TEDS-D)

	Treatment Completion		Length of stay greater than 180 days~		Length of stay greater than 365 days~
Characteristic (N= 152,196)	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
	RR [95% CI]	aRR [95% CI]	RR [95% CI]	aRR [95% CI]	RR [95% CI]	aRR [95% CI]
MOUD in treatment plan
Yes	0.43 [0.35, 0.54]***	0.46 [0.38, 0.57]***	1.68 [1.35, 2.09]***	1.60 [1.29, 1.99]***	2.86 [2.12, 3.85]***	2.64 [2.00, 3.49]***
No	Ref	Ref	Ref	Ref	Ref	Ref
State Medicaid methadone coverage
Yes, within covered state	--	1.14 [0.83, 1.56]	--	1.02 [0.66, 1.56]	--	1.11 [0.65, 1.91]
No, not within covered state		Ref		Ref		Ref

Note: RR=risk ratio; aRR= adjusted risk ratio; MOUD= medication for opioid use disorder; excludes Georgia, Oregon, West Virginia, Puerto Rico, and Washington District of Columbia; col.=column;

^*p<0.05

^**p<0.01

^***p<0.001;

models adjusted for sex, race/ethnicity, age group, state-level percentage male, percentage NH White, percentage age 10 to 24, percentage without high school education, and standardized total population in addition to variables indicated in the table; accounted for state clustering.

^~Limited to discharges with a value for treatment completion or dropped out/terminated by facility. All discharges were restricted to ambulatory non-intensive outpatient service setting admissions that had heroin or opioids/synthetics reported as the primary, secondary, or tertiary substance at the time of admission.

Appendix Table 1 displays analyses with an alternative treatment outcome operationalized as treatment completion or transferred to another treatment program versus all other discharge outcomes. Thus, discharges with incarcerated, death, or other discharge reasons were excluded; the final sample was 201,527 (see appendix Table 1A). The direction and significance of associations were not altered with this alternative treatment outcome. None of the sensitivity analyses significantly changed the interpretation of our results (not shown; available upon request).

4. Discussion

4.1. Findings

The goal of this study was to examine the relationship between having MOUD listed in a treatment plan and substance use treatment outcomes among opioid-related discharges, focused on treatment retention and completion, with a comprehensive national dataset of treatment discharges in 2016. Treatment retention was longer in discharges with MOUD listed in the treatment plan, while treatment completion was lower, compared to opioid-related discharges without MOUD in the treatment plan. We also tested the role that state-level Medicaid methadone coverage plays in altering the associations between having MOUD as a component of the treatment plan and substance use treatment outcomes. Discharges from states with Medicaid methadone coverage were almost three times more likely to have MOUD in the treatment plan compared to those from states without Medicaid methadone coverage. While Medicaid increased MOUD treatment plans both in another study (Bachhuber et al., 2017) and in our study, we did not find evidence that Medicaid methadone coverage had spillover effects on altering the relationship between having MOUD in individual treatment plans and treatment completion or retention, after accounting for other individual- and state-level characteristics.

One major strength of our study was that it directly examined the relationship between having MOUD in individual treatment plans and substance use treatment outcomes with two ways of measuring outcomes: treatment completion and treatment retention. By utilizing these two measures of substance use treatment outcomes, our study showed that different ways of evaluating favorable results could lead to different conclusions. First, we hypothesized that having MOUD in individual treatment plans would increase likelihood of treatment completion; however, our findings did not support our hypothesis, and we found that having MOUD in individual treatment plans decreased likelihood of substance use treatment completion based on discharge reason. Second, independent of substance use treatment completion based on discharge reason, we hypothesized that MOUD treatment plans would increase likelihood of treatment retention at both 180+ days and 365+ days. Our hypotheses regarding treatment retention were supported in this study. Third, we hypothesized that state-level Medicaid coverage of methadone would modify the relationships between MOUD treatment plans and substance use treatment completion/retention. Our hypotheses were not supported and state-level Medicaid coverage of methadone did not modify any of these relationships. Nonetheless, our findings using the TEDS-D show that MOUD treatment plans were more likely in states with Medicaid coverage of methadone than in states without this coverage, consistent with a prior study using the TEDS-A (Bachhuber et al., 2017).

To our knowledge, no prior studies have examined whether state-level Medicaid coverage modifies the relationships between access to MOUD (i.e., having MOUD listed within the treatment plan) and treatment completion or treatment retention. We did not find evidence of spillover effects between state-level Medicaid methadone coverage and treatment retention or completion, beyond the effects of increasing access to MOUD. It is important to note that our findings support past studies that found that Medicaid coverage was associated with having MOUD included in treatment plans, highlighting the important role that state Medicaid plays in MOUD access. The proportion of discharges that included MOUD in the treatment plan were almost three times higher in states with Medicaid methadone coverage (47.7%) compared to states without methadone coverage (16.4%). In light of our findings that MOUD is associated with longer treatment retention, even after accounting for state and individual differences, state Medicaid agencies can play an important role in increasing treatment retention through increasing access to MOUD. MOUD treatment retention, regardless of treatment completion, has significant health benefits and is associated with lower overdose mortality (Krawczyk et al., 2020).

The consistencies that we found with our sensitivity analyses in the direction and magnitude of these associations improved our confidence in these findings. In another study using TEDS-D, Mennis and colleagues (2019) hypothesized that longer retention in substance use treatment would be associated with increased odds of treatment completion comparing opioids to alcohol as the primary substance; they found opioids to be associated with lower odds of treatment completion and higher odds of treatment retention at 365 days or more (Mennis et al., 2019). Our study found even starker differences in magnitude of associations observed between the two measures of substance use treatment outcomes, treatment completion and treatment retention, when examining having MOUD in individual treatment plans as the main exposure of interest. Mennis and colleagues note that omitting MOUD treatment plans from their models was a limitation of their study since MOUD treatment retention is known to approximate treatment progress (Mennis et al., 2019), which we were able to address in our study.

The TEDS-D did not include information about the intended time for MOUD treatment, and some MOUD treatment plans may be intended to be time limited while others may require long-term maintenance. TEDS-D excluded information about MOUD dose, type of MOUD, or severity of OUD. This information could have elucidated some differences between those who completed treatment and those who did not. Additionally, those with psychiatric comorbidities may have complex cases of OUD, which could result in longer treatment retention needed to complete treatment plans (Krawczyk, Feder, Saloner, et al., 2017). However, our sensitivity analyses did not show that psychiatric comorbidity could explain our findings. Stigma associated with taking MOUD could be one reason why likelihood of treatment completion is higher among people who do not take MOUD (Wakeman & Rich, 2018). Lack of continued access to treatment could be another reason why treatment completion was low. Limited access to MOUD due to prior authorization requirements or health insurance caps on the amount and time of treatment that is covered could inhibit people from completing recommended treatment (Davis & Carr, 2019). Only 6% of U.S. substance use treatment facilities provide all three types of MOUD (Mojtabai et al., 2019), and individuals in rural counties have even more difficulty accessing MOUD than those in urban counties (Jones, 2017). These findings call for structual interventions to reduce inequities in access to MOUD.

Previous studies have examined the relationship between MOUD treatment plans and substance use treatment outcomes based on length of stay in treatment (Hubbard et al., 2003; Timko et al., 2016). Consistent with the findings from our study, MOUD treatment plans were associated with increased treatment retention. To our knowledge only one other study measured these outcomes concurrently but did not adjust for whether MOUD was listed in the treatment plan (Mennis et al., 2019). Another substance use treatment effectiveness study focused broadly on multiple substances without examining the association of MOUD treatment plans on treatment outcomes (Bornstein, 2015). Other literature utilizing TEDS-D data and focusing on treatment completion removed treatment episodes with MOUD listed in the treatment plan altogether from analyses (Sahker et al., 2015).

Our study provides evidence that additional efforts may be needed to enhance treatment completion outcomes based on discharge reason, regardless of treatment retention, for those with MOUD treatment plans (e.g., programs that aid in care coordination for complex patients such as Medicaid health homes). Other factors, such as provider capacity, should be considered when trying to understand the mechanisms through which Medicaid coverage of methadone did not have spillover effects on treatment completion or retention. A study assessing the effects of Medicaid expansion for MOUD treatments found that provider capacity constraints may hinder benefits of coverage expansions. Specifically, Medicaid expansion effects on MOUD access were greater in states that had more buprenorphine-waivered providers, suggesting that access to MOUD is also impacted by capacity of providers (Gertner et al., 2020). While MOUD could be listed in the treatment plan, provider capacity issues could have limited benefits of Medicaid coverage, altering the relationship between MOUD access and treatment outcomes. Another study found that only 6.1% of specialty substance use facilities offered all forms of MOUD in 2016; however, 36.1% of facilities offered some form of MOUD at that time (Mojtabai et al., 2019). This could mean that other forms of MOUD that Medicaid covered were substituted for methadone in states without Medicaid coverage of methadone. Alternatively, it could be that other funding mechanisms, such as federal block grants or general revenue funding, could cover methadone treatments in the absence of Medicaid coverage. Future research should assess these other potential mechanisms that may modify the relationships between having MOUD listed in individual treatment plans and treatment completion/retention.

Based on our results, we would caution future researchers from utilizing the treatment completion measure in isolation when evaluating MOUD treatment success, and instead also incorporate retention and engagement as key outcomes. There is strong evidence that longer treatment episodes with high retention and maintenance of MOUD are associated with better clinical and social outcomes (SAMHSA, 2015; Deck et al., 2009; Larochelle et al., 2018). In that light, treatment completion could be considered a negative outcome rather than a positive outcome when long-term maintenance is the goal of the treatment plan. Treatment retention gives us a sense of the impact of MOUD on clinically relevant outcomes that researchers could miss when examining treatment completion alone. Future studies should explore if there are specific aspects of MOUD treatment plans that make them more likely to be completed, such as dose or type of MOUD. Other information from the MOUD treatment plan, such as severity of disorder, could help us understand the relationship between having MOUD in the individual treatment plan and treatment outcomes since people with more severe disorders may need more resources and time to complete treatment. Treatment seeking is higher among people with more years since initial diagnosis and more severe impairment from their disorder (Blanco et al., 2015), so TEDS may over-represent treatment episodes from those with more severe disorders that may be more difficult to treat compared to those with less severe disorders.

4.2. Limitations

TEDS included administrative discharge data, not individual people. Since we could not identify individuals, we could not determine whether a person was receiving multiple types of services because we did not have complete records for each individual (SAMHSA, 2009). Detailed information about the contents of MOUD treatment plans, such as severity of OUD or MOUD dose, was unavailable in the TEDS. Research has shown that MOUD dose has a significant impact on MOUD treatment outcomes (D’Aunno, Park, & Pollack, 2019). The language used in the TEDS for the MOUD treatment plan indicator cannot distinguish between forms of MOUD, such as buprenorphine and methadone, or whether MOUD was used or simply listed in the treatment plan for future use. Our findings may not generalize to office-based buprenorphine or naltrexone treatment, or people receiving opioid-related treatment in other settings that were not included TEDS (e.g., primary care or specialty psychiatric settings) (Chou R & D., 2016). Since methadone is dispensed through certified opioid treatment programs, we likely captured a large proportion of people who receive treatment at publicly funded facilities with this dataset. Completeness of TEDS records varies by state (SAMHSA, 2018a), so findings may not generalize to states that were not included in the TEDS. West Virginia, which had the highest opioid overdose death rate in the country at 43.4 per 100,000 in 2016 (Kaiser Family Foundation, 2016), was not included in the 2016 TEDS-D and therefore is not represented in these findings. While our analyses accounted for state-level differences, state discrepancies in reporting limit direct comparisons across states.

5. Conclusion and policy implications

Our findings suggest that there is a missed opportunity to count successes of MOUD treatment, which is an important part of addressing the opioid and overdose epidemic, if we only rely on a single indicator of treatment completion. There is still more work to do to increase retention in care for those who are engaged in MOUD treatment and redefine what treatment completion means, since this has implications for people with contingency plans based on their treatment success (e.g., legal or employment implications) (CSAT, 2005). Reducing MOUD treatment stigma remains an important goal for substance use treatment, not only for initiation but also maintenance, as stigma is a substantial barrier to initiating and engaging in MOUD treatment (Wakeman & Rich, 2018). MOUD treatment availability is still low and policymakers and public health professionals should address structural barriers to treatment, such as health insurance coverage and provider regulatory policies that limit treatment access. Increasing access to evidence-based treatment such as MOUD is an important strategy for addressing the overdose epidemic. Other population-level strategies, such as syringe exchange programs (CDC, 2019) and naloxone distribution (Doyon, Aks, & Schaeffer, 2014), can help to attenuate harms associated with OUD. Future research should assess how treatment outcomes may be impacted by other MOUD access policies that were beyond the scope of this study, such as prior authorization requirements or caps on duration of treatments (Davis & Carr, 2019; Haffajee, Bohnert, & Lagisetty, 2018). Medicaid efforts should invest resources in helping individuals to reach their own treatment targets, in addition to continuing to increase MOUD access and utilization. As we continue to invest resources in OUD treatments, it is essential to reframe our perspective on treatment outcomes for those receiving MOUD treatment long-term.

Highlights.

• Opioid medications were related to an increase in treatment retention

• Opioid medications were related to a decrease in treatment completion

• State-level Medicaid methadone coverage did not modify these relationships

• Our results can inform efforts to alter how we operationalize treatment outcomes

Appendix Table 1:

Associations between MOUD and substance use treatment completion or transferring to another treatment program and treatment retention with (A) and without (B) MOUD by state Medicaid methadone coverage interaction, 2016 Treatment Episode Data Set-Discharges (TEDS-D)

A)	Treatment Completion/Transfer	Treatment Completion/Transfer	Length of stay greater than 180 days^		Length of stay greater than 365 days^
Characteristic (N= 201,527)	Model 1 RR [95% CI]	Model 2 aRR [95% CI]	Model 3 RR [95% CI]	Model 4 aRR [95% CI]	Model 5 RR [95% CI]	Model 6 aRR [95% CI]
MOUD in treatment plan
Yes	0.72 [0.60, 0.88]**	0.80 [0.69, 0.93]**	1.97 [1.48, 2.62]***	1.73 [1.35, 2.21]***	3.40 [2.36, 4.88]***	2.91 [2.12, 3.99]***
No	Ref	Ref	Ref	Ref	Ref	Ref
State Medicaid methadone coverage
Yes, within covered state	--	1.08 [0.75, 1.56]	--	1.15 [0.67, 1.96]	--	1.33 [0.71, 2.49]
No, not within covered state		Ref		Ref		Ref
B)	Treatment Completion~	Length of Stay~		Treatment Completion/Transfer^	Length of Stay^
		180+ days	365+ days		180+ days	365+ days
Characteristic	Model 1 aRR [95% CI]	Model 2 aRR [95% CI]	Model 3 aRR [95% CI]	Model 4 aRR [95% CI]	Model 5 aRR [95% CI]	Model 6 aRR [95% CI]
MOUD in treatment plan
Yes	0.55 [0.38, 0.81]**	1.36 [1.10, 1.69]**	2.21 [1.54, 3.17]***	0.91 [0.82, 1.00]	1.42 [1.09, 1.85]*	2.39 [1.62, 3.51]***
No	Ref	Ref	Ref	Ref	Ref	Ref
State Medicaid methadone coverage
Yes, within covered state	1.16 [0.85, 1.59]	0.98 [0.61, 1.56]	1.03 [0.55, 1.95]	1.11 [0.77, 1.60]	1.08 [0.59, 1.97]	1.21 [0.55, 2.65]
No, not within covered state	Ref	Ref	Ref	Ref	Ref	Ref
Interaction state Medicaid methadone coverage*MOUD in treatment plan	0.82 [0.54, 1.27]	1.19 [0.88, 1.61]	1.22 [0.76, 1.95]	0.86 [0.70, 1.07]	1.24 [0.85, 1.80]	1.25 [0.74, 2.11]

Note: RR=risk ratio; aRR= adjusted risk ratio; MOUD= medication for opioid use disorder; excludes Georgia, Oregon, West Virginia, Puerto Rico, and Washington District of Columbia; col.=column;

^*p<0.05

^**p<0.01

^***p<0.001;

models adjusted for sex, race/ethnicity, age group, state-level percentage male, percentage NH White, percentage age 10 to 24, unemployment rate, standardized total population, and standardized median household income; accounted for state clustering.

^~Limited to discharges with a value for treatment completion or dropped out/terminated by facility (N= 152,196).

^{^}Limited to discharges with a value for treatment completion/ transferred to another treatment program or dropped out/terminated by facility (N=201,257). All discharges were restricted to ambulatory non-intensive outpatient service setting admissions that had heroin or opioids/synthetics reported as the primary, secondary, or tertiary substance at the time of admission.