Journal of Substance Abuse Treatment 104 (2019) 15–21

Contents lists available at ScienceDirect

Journal of Substance Abuse Treatment

journal homepage: www.elsevier.com/locate/jsat

A comparison of adherence, outcomes, and costs among opioid use disorder T

Medicaid patients treated with buprenorphine and methadone: A view from
the payer perspective☆
⁎
Suzanne Kinskya, , Patricia R. Houckb, Kristin Mayesa, David Lovelandc, Dennis Daleyd,
James M. Schustere
a
UPMC Center for High-Value Health Care, US Steel Tower, 600 Grant Street, 40th Floor, Pittsburgh, PA 15219, USA
b
UPMC Health Plan, Department of Health Economics, , US Steel Tower, 600 Grant Street, 21st Floor, Pittsburgh, PA 15219, USA
c
Community Care Behavioral Health, 339 Sixth Avenue, Suite 1300, Pittsburgh, PA 15222, USA
d
UPMC Health Plan, US Steel Tower, 600 Grant Street, Pittsburgh, PA 15219, USA
e
UPMC Insurance Services Division, US Steel Tower, 600 Grant Street, 55th Floor, Pittsburgh, PA 15219, USA

A R T I C LE I N FO A B S T R A C T

Keywords: Medication-assisted treatment (MAT) with methadone or buprenorphine has been shown to be more effective at
Medication-assisted therapy reducing the use of illicit opioids, the risk of drug-related overdose, and overall healthcare costs, on average,
Buprenorphine compared to abstinence-based addiction treatments for individuals with an opioid use disorder (OUD).
Methadone Individuals who are adherent to MAT are more likely to experience positive outcomes.
Adherence
We used physical and behavioral Medicaid claims data of individuals newly treated with methadone
Cost
Opioid addiction
(n = 212) and buprenorphine (n = 972) to examine the overall predictors of adherence, differences in adherence
to each medication, the relationship between adherence and ED nonfatal drug-related overdose, and differences
in total cost of care between the two medications.
We found that older individuals and women had significantly lower risk of non-adherence. At six months,
only 3.6% of individuals who were adherent to either treatment experienced a nonfatal drug-related overdose in
the ED, compared to 13.2% of individuals who were non-adherent. We found no significant difference between
methadone and buprenorphine on nonfatal drug-related overdose. Non-adherence to methadone was associated
with a significant increase in total cost of care. Implications for how these results could be used to improve the
overall impact of MAT are discussed.

1. Introduction to MAT long-term demonstrate increased pharmacy costs relative to

The opioid epidemic in the United States is associated with high
costs to individuals, families, health systems, and society. Medication-
assisted treatment (MAT) is frequently used to treat opioid use disorder
(OUD) due to its clinical and cost effectiveness. Compared to non-
pharmacologic treatment interventions, such as behavioral therapy,
MAT is associated with lower rates of relapses and decreased health
care utilization (Clark et al., 2015; Wickizer, Krupski, Stark, Mancuso,
& Campbell, 2006). Several studies find that total cost of care for in-
dividuals with OUD receiving MAT is significantly lower than those not
on MAT (Clark et al., 2015; Clark, Samnaliev, Baxter, & Leung, 2011;
Lynch et al., 2014; McCarty et al., 2010). While individuals who adhere
their non-adherent counterparts, total cost of care remains lower
(Tkacz, Volpicelli, Un, & Ruetsch, 2014). Moreover, adherence to MAT
is associated with reduced mortality and reduced unplanned care uti-
lization (Lo Ciganic et al., 2016; Ma et al., 2018). Factors associated
with lower adherence to MAT include race/ethnicity, co-occurring
serious mental illness (SMI), and younger age (Clark et al., 2015; Hser
et al., 2014; Manhapra, Petrakis, & Rosenheck, 2017; Rieckmann,
Gideonse, Risser, DeVoe, & Abraham, 2017; Weinstein et al., 2017).
While MAT is often regarded as the best strategy to treat OUD, there
are important differences between the medications approved for MAT
use. Methadone and buprenorphine, the two most widely used MAT
medications, differ in terms of the underlying pharmacologic actions,

☆
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
⁎
Corresponding author.
E-mail addresses: kinskys@upmc.edu (S. Kinsky), houckpr@upmc.edu (P.R. Houck), mayeskl@upmc.edu (K. Mayes), lovelanddl@ccbh.com (D. Loveland),
daleydc@upmc.edu (D. Daley), schusterjm@ccbh.com (J.M. Schuster).

https://doi.org/10.1016/j.jsat.2019.05.015
Received 1 November 2018; Received in revised form 24 May 2019; Accepted 30 May 2019
0740-5472/ © 2019 Elsevier Inc. All rights reserved.
S. Kinsky, et al.
the dosing schedule, resources required to provide treatment, and cost
of providing treatment, making it important for all stakeholders to
understand the relative costs and benefits of each. Methadone is a
schedule II opioid dispensed only in licensed and accredited opioid
treatment programs (Chou et al., 2016). These restrictions make me-
thadone more resource-intensive and less accessible to rural patients
than buprenorphine, which is available in primary care physician of-
fices and can be provided by a waivered physician extender (Substance
Abuse and Mental Health Services Administration, 2006). Compared to
those on methadone, those on buprenorphine use fewer healthcare re-
sources and spend significantly fewer healthcare dollars (Barnett, 2009;
Baser, Chalk, Fiellin, & Gastfriend, 2011; Clark et al., 2011; Maas,
Barton, Maskrey, Pinto, & Holland, 2013). However, there is mixed
evidence suggesting methadone is more likely than buprenorphine to
retain individuals in treatment and/or prevent relapse. Some studies
have found that individuals on buprenorphine are more likely to relapse
or have shorter treatment episodes compared to methadone (Barnett,
2009; Clark et al., 2011; Maas et al., 2013). On the other hand, in-
dividuals initiating MAT with buprenorphine are slightly less likely to
experience a drug-related overdose compared to those taking metha-
done, particularly in the early stages of treatment (Kimber, Larney,
Hickman, Randall, & Degenhardt, 2015; Sordo et al., 2017).
Pennsylvania is particularly impacted by the opioid epidemic. The
rate of drug-related overdose deaths in the state was 43 per 100,000 in
2016 (DEA, 2018) compared to the national average of 22 per 100,000
(Hedegaard, Minino, & Warner, 2018). The vast majority (85%) of
overdose deaths in Pennsylvania are related to opioids (DEA, 2017).
The opioid epidemic in PA has led to a steady increase in MAT for
individuals with an OUD. Medicaid is the primary funding source for
MAT in western Pennsylvania as well as in the U.S. Nationwide, ap-
proximately 38% of adults with opioid addiction have Medicaid.
Moreover, individuals who have Medicaid are more likely to receive
MAT than in other insurance programs (Zur & Tolbert, 2018).
Given the association between initiation and retention on MAT and
risk of drug-related overdose, the aim of this analysis was to assess
adherence, costs, and emergency department (ED) nonfatal drug-re-
lated overdose outcomes among a sample of Medicaid members. Our
sample consisted of individuals with an OUD receiving MAT within
UPMC, which has 3.5 million members across all insurance products
and is the second largest managed Medicaid program in PA. We used
physical and behavioral claims data of individuals newly treated with
methadone and buprenorphine to examine the overall predictors of
adherence as well as differences in adherence to methadone versus
buprenorphine for six months. We also assessed the relationship be-
tween adherence and ED nonfatal drug-related overdose over six
months. Finally, we examined differences in total cost of care between
the two treatment groups.
2. Material and methods
2.1. Sample
Individuals targeted for this analysis were those on Medicaid re-
ceiving both physical and behavioral health insurance coverage
through the UPMC Insurance Services Division's managed care pro-
ducts. For physical health, Medicaid members are covered by UPMC for
You, and for behavioral health, members are covered by the
Community Care Behavioral Health Organization (Community Care) if
they live in one of the 39 counties served by Community Care.
Individuals were eligible for inclusion in the analysis sample if they
began receiving either methadone or buprenorphine during calendar
years 2015 or 2016 (Fig. A.1). We examined pharmacy and medical
visit claims data to identify new start members. New start members
were defined as those without a MAT prescription 12 months prior to
treatment initiation; however, we did not require that members be
continuously enrolled for 12 months prior to treatment. Members
16
Journal of Substance Abuse Treatment 104 (2019) 15–21
treated with naltrexone, and those who received both methadone and
buprenorphine during the observation period, were excluded due to
insufficient numbers. Data on medication adherence and cost were
pulled from the Medicaid Deidentified Data Source, which contains
both behavioral and physical health claims for members with UPMC for
You and Community Care coverage. ED nonfatal drug-related overdose
data came from UPMC for You claims data. Primary ED diagnosis was
alcohol or opioid use (F10 or F11) or poisoning by alcohol or drug (T39-
T43, T50, T51). Only nonfatal events were counted in this study.
2.2. Outcomes
Medication adherence was measured for six months after starting
treatment. Methadone was administered directly by clinic-based pro-
viders and insurance claims were filed for each clinic visit. Non-ad-
herence to methadone therapy was defined as a lapse in daily clinic visit
claims for more than seven days. Buprenorphine was prescribed as oral
medication to be taken daily. We examined pharmacy claims, pre-
scription fills, dispensing dates, and days supply to infer gaps in med-
ication coverage. Non-adherence to buprenorphine therapy was defined
as a medication coverage gap longer than ten consecutive days. We
measured time to non-adherence as well as whether or not members
were continuously adherent for six months after starting treatment. ED
nonfatal drug-related overdoses were defined as members' visits to an
ED within the observation period with a primary diagnosis of drug or
alcohol poisoning. Cost of care represents actual payments made by
UPMC and was calculated on a per member per month (PMPM) basis
for medical care, behavioral health care, and pharmacy costs, as well as
the total of all three categories over the year following MAT start.
2.3. Analysis
The sample available for the adherence analysis was limited to
members who were treatment naïve for at least three months of
membership prior to medication start and had at least six months of
membership available after medication start. Distribution of key cov-
ariates was compared between members receiving methadone and bu-
prenorphine using group t-test for continuous variables and chi-square
test for categorical variables. Covariates of interest included age (≤40
vs. 40+), county of residence, gender, race (white vs. other), number of
unique prescriptions filled in the previous three months as a proxy for
comorbidity (Veehof, Stewart, Haaijer-Ruskamp, & Jong, 2000), whe-
ther a member had a diagnosis of serious mental illness (SMI), and
whether the member was adherent for six months. Area Deprivation
Index (ADI), a composite of socioeconomic factors (Singh, 2003), was
used as a dichotomous variable, with low ADI defined as < 110 and
high ADI defined as ≥110. This cut-off is empirically derived from
UPMC Health Plan data, representing the top tertile of over one million
members (unpublished data). High ADI indicates an area of higher
socioeconomic deprivation. We explored adding alcohol use disorder as
a covariate but did not include it because of significant correlation with
other covariates and concerns about data incompleteness. Proportional
hazards regression was used to examine the association of covariates to
time to non-adherence. Next, in two separate steps, we added the in-
teraction terms of treatment group and SMI, and treatment group and
ADI, to assess differences by treatment group (using buprenorphine as
the reference).
Kaplan-Meier survival curves were constructed and compared
(using the log-rank statistic) for time to ED nonfatal drug-related
overdose for individuals who were vs. were not adherent for six
months. Then, a proportional hazards regression model was constructed
as before to examine the effects of covariates on time to ED nonfatal
drug-related overdose. We then added the interaction term of ad-
herence and treatment group.
To estimate cost differences, we used a difference-in-difference ap-
proach to compare the change in PMPM over one year between the two
S. Kinsky, et al. Journal of Substance Abuse Treatment 104 (2019) 15–21

treatment groups. In order to conduct a robust cost analysis, we ex- Table 2

tended the requirements for membership to obtain sufficient data upon Proportional hazards model results for predictors of time to non-adherence.
which to draw inference. For this analysis, members were required to Variable Adjusted HR (95% CI) p
have at least 9 months of membership before and after the treatment
start date; membership did not need to be continuous. Generalized es- Age 40+ 0.822 (0.691, 0.978) 0.027
Female 0.856 (0.735, 0.998) 0.047
timating equations (PROC GENMOD) were constructed with age,
White 0.822 (0.633, 1.068) 0.143
gender, race, ADI, SMI, and number of prescription drugs at treatment Area Deprivation Index = 110+ 1.093 (0.938, 1.273) 0.254
start as covariates. We included a three-way interaction term of treat- Serious mental illness 1.165 (0.998, 1.360) 0.053
ment group, time (pre/post), and adherence to assess the impact of 10+ current prescriptions 0.968 (0.824, 1.137) 0.689
adherence to each drug on PMPM. Methadone Treatment Group 0.890 (0.722, 1.098) 0.276

Significance was determined as p < 0.05. All analyses were con-

Bold text indicates statistical significance at the threshold of p < .05 or lower.
ducted using SAS/STAT Version 13.2 (SAS Institute Inc., 2014).

3. Results

3.1. Adherence

After applying exclusions for the adherence analysis, there were

1184 members available for the analysis of adherence, 212 of whom
received methadone and 972 of whom received buprenorphine.
Individuals treated with methadone were significantly older than those
treated with buprenorphine (mean age 38.8 vs. 35.3, respectively) and
were predominantly located in an urban region but were otherwise
similar to those treated with buprenorphine. A higher proportion of
methadone-treated individuals were adherent for six months compared
to those on buprenorphine (49.1% vs. 40.8%, respectively), though this
difference did not reach significance (χ2 = 2.77, p = 0.096) (Table 1).
In the proportional hazards model, older individuals (OR = 0.82, Fig. 1. Adherence to MAT by ADI and treatment group.
p = 0.027) and women (OR = 0.86, p = 0.047) had significantly lower
risk of non-adherence (Table 2). Individuals with SMI had a higher risk months, only 3.6% of individuals who were adherent to either treat-
of non-adherence, but significance was only marginal (OR = 1.17, ment experienced an ED nonfatal drug-related overdose, compared to
p = 0.053). MAT treatment group was not associated with non-ad- 13.2% of individuals who were non-adherent. As in the prior analysis,
herence, after controlling for other covariates (OR for methadone in the proportional hazards model predicting time to ED nonfatal drug-
compared to buprenorphine = 0.089, p = 0.28). The interaction term related overdose, older individuals and women were less likely to ex-
of treatment group and ADI was significant (p = 0.026). Fig. 1 shows perience an overdose (Table 3). Individuals living in an area with high
the distribution of treatment group and ADI on six-month adherence. socioeconomic deprivation were also less likely to experience a nonfatal
The proportion of individuals who were adherent for six months did not overdose (OR = 0.66, p = 0.028), as were individuals who were ad-
differ among those living in an area with high ADI, but in low ADI herent for six months (OR = 0.29, p < 0.0001). An interaction of ad-
areas, a significantly higher proportion of individuals on methadone herence and treatment group was not significant (p = 0.69, data not
were adherent for six months compared to individuals on buprenor- shown).
phine (55% vs. 40%, respectively). That is, living in an area with low
socioeconomic deprivation was associated with higher adherence
among individuals receiving methadone. The interaction of treatment 3.2. Cost
group and SMI status was not significant (p = 0.87, data not shown).
The Kaplan-Meier survival analysis of time to ED nonfatal drug-re- After applying exclusions for the cost analysis (at least 9 months of
lated overdose indicated that overdoses were higher among individuals membership in the year prior and year after treatment start), there were
who were non-adherent for six months (p = 0.001) (Fig. 2). At six 692 members in the sample (125 receiving methadone and 567 re-
ceiving buprenorphine). Individuals receiving methadone were sig-
Table 1 nificantly older than those receiving buprenorphine but there were no
Characteristics of members newly initiating methadone or buprenorphine, other significant differences between the groups. The difference-in-
2015–2016. difference analysis showed no difference in PMPM between the two
Variable Methadone Buprenorphine χ2 or t p treatments, although individuals receiving buprenorphine treatment
increased pharmacy cost by $219 while those receiving methadone
Total members 212 972
decreased pharmacy cost by $23 (p = 0.011) (data not shown). There
Age (mean, range) 38.8 [19,64] 35.5 [19,65] 4.32 0.0001
Age 40+ 43.9% 28.0% 20.59 0.0001
was a significant interaction effect between change in PMPM by treat-
Female 55.7% 54.5% 0.09 0.76 ment group and adherence (p = 0.014, Table A.1). Among individuals
White 88.2% 91.6% 1.97 0.16 receiving methadone, non-adherence was related to a significant in-
Allegheny county 79.3% 56.9% 36.52 0.0001 crease from baseline in PMPM: PMPM decreased by $13.27 among in-
Area Deprivation 52.4% 56.9% 1.45 0.23
dividuals who were adherent to methadone but increased by $1173.50
Index = 110+
Serious mental illness 41.0% 48.8% 4.17 0.041 among individuals who were non-adherent. The change in PMPM was
10+ current prescriptions 37.7% 38.8% 0.08 0.78 much smaller for buprenorphine: cost increased by $533.20 for in-
Adherent for six months 49.1% 40.8% 2.77 0.096 dividuals who were adherent and $475.94 for individuals who were
non-adherent (Fig. 3).
Sample limited to members with at least 3 months of data prior to and 6 months
of data after new start.
Bold text indicates statistical significance at the threshold of p < .05 or lower.

17
S. Kinsky, et al. Journal of Substance Abuse Treatment 104 (2019) 15–21

Time to ED Nonfatal Drug-Related Overdose

Adherent 6 months n=501 Non Adherent n=683
20%

15% 13.2%

% of members 10%

5% 3.6%

0%
0 60 120 180 240 300 360
Days aer starng treatment

Fig. 2. Kaplan-Meier survival curves for time to ED nonfatal drug-related overdose for methadone and buprenorphine.

Table 3 observational design of this analysis cannot rule out important sources
Proportional hazards model results for predictors of time to ED nonfatal drug- of bias (Shrank, Patrick, & Brookhart, 2011). For example, the re-
related overdose. lationship between adherence to MAT and ED nonfatal drug-related
Variable Adjusted HR (95% CI) p overdose could be confounded by unobserved factors that predispose
the individual to both remain in treatment and engage in safer opioid
Age 40+ 0.598 (0.373,0.960) 0.033 use practices. We found no significant difference in adherence over six
Female 0.676 (0.460,0.992) 0.045
months between individuals treated with methadone versus buprenor-
White 1.277 (0.588,2.772) 0.537
Area Deprivation Index = 110+ 0.655 (0.449,0.955) 0.028 phine, which is largely consistent with other studies (Mattick, Breen,
Serious mental illness 1.360 (0.917,2.016) 0.126 Kimber, & Davoli, 2014). However, unlike prior literature (Kimber
10+ current prescriptions 1.371 (0.925,2.034) 0.116 et al., 2015; Sordo et al., 2017), we did not observe a difference be-
Methadone Treatment Group 0.599 (0.320,1.124) 0.110 tween treatments on ED nonfatal drug-related overdose. Similar to
Adherent for six months 0.290 (0.176,0.477) < 0.0001
other findings, there is a significant relationship between adherence,
Bold text indicates statistical significance at the threshold of p < .05 or lower. treatment group, and change in total cost of care. We found a higher
pharmacy cost among individuals receiving buprenorphine. Non-ad-
4. Discussion herence to methadone was associated with a significant increase in total
cost of care.
Results provide additional insight into the relationships between We observed an association between living in an area with high
MAT treatment group, adherence, ED nonfatal drug-related overdose, socioeconomic deprivation and adherence to MAT, whereby a smaller
and cost among a sample of individuals on Medicaid in Pennsylvania. proportion of individuals taking buprenorphine were adherent if they
Adherence to MAT and ED nonfatal drug-related overdose were corre- lived in an area of high socioeconomic deprivation relative to in-
lated such that the proportion of individuals who experienced an dividuals taking methadone. Because Medicaid members in PA do not
overdose was over 3.5 times as high among those who were non-ad- have cost-sharing for medications, cost of medication cannot explain
herent compared to those who were adherent. However, the this finding. It is likely that the differences in age, county of residence,

Change in Total PMPM from Baseline

Adherent Not Adherent
$1,173.50
p = 0.014
$1,100

$900
Cha nge fr om pr e tr ea tment

$700
$553.20
$475.94
$500

$300

$100 $(13.27)
n=60 n=237 n=330
n=65
$(100)
Methadone N=125 Buprenorphine N=567

Results from GEE analysis. See Table A.1 in appendix for full results.

Fig. 3. Total per member per month cost change from baseline by adherence and treatment group, adjusted for age, gender, race, number of prescriptions, and ADI.

18
S. Kinsky, et al.
and comorbidities contribute to observed differences. However, we
found no association between low socioeconomic deprivation and ED
nonfatal drug-related overdose among individuals taking either MAT
medication. Members living in an area with high socioeconomic de-
privation showed a decreased chance of experiencing nonfatal over-
dose. These conflicting results highlight the challenge of measuring and
controlling for contextual factors related to opioid use (Campbell et al.,
2018). For example, the observed relationship between area depriva-
tion and overdose reflects unobserved confounders such as ability to
access an emergency department and the availability of overdose re-
versal medication. As this was a sample of Medicaid members, low
variability in individual-level socioeconomic status likely limited our
ability to elucidate the relationship between socioeconomics and MAT
outcomes. Additional research on the contribution of social and en-
vironmental contextual factors to MAT outcomes would be useful for
clinicians and policy makers.
This research highlights the nuances of perspective when comparing
the value of methadone to buprenorphine: from an overdose stand-
point, there was no difference between methadone and buprenorphine,
whereas from a cost standpoint, the consequences of non-adherence
were more salient for individuals receiving methadone. From either
perspective, adherence to MAT regardless of treatment group appeared
crucial. When considering opioid interventions, health care providers
and policy makers should consider factors such as age, gender, and
neighborhood that can influence adherence. How to effectively address
barriers to adherence is an area of needed exploration, as our findings
emphasize the profound and multi-level consequences of non-ad-
herence. In addition to behavioral and structural interventions, future
studies could examine long-acting forms of buprenorphine and their
effect on adherence, overdose, and costs.
There are several limitations to our analysis. This is an observa-
tional, non-randomized study. Many factors which were not examined
in this study contribute to a patient's adherence level and likeliness to
experience an ED nonfatal drug-related overdose. These factors may
impact a patient's total cost of care. Further, these factors are likely to
influence a physician's choice of prescribing clinic-based methadone or
prescription buprenorphine for individual patients. This was a rela-
tively small sample of individuals and was limited to individuals who
Appendix A
Table A.1
General estimating equations model results for total cost of care.
Variable Adjusted estimate (95% CI)
Age 40+ −440.247 (−708.088, −172.405)
Female
White
Area Deprivation Index = 110+
Serious mental illness −467.495 (−656.870, −278.120)
10+ current prescriptions −726.795 (−890.871, −562.719)
Methadone Treatment Group
Adherent for six months
Time (pre/post)
Methadone Treatment Group ∗ time
Time ∗ adherent −1186.76 (−2159.670, −213.842)
Methadone Treatment Group ∗ adherent
Methadone Treatment Group ∗ time ∗ adherent
Bold text indicates statistical significance at the threshold of p < .05 or lower.
19
Journal of Substance Abuse Treatment 104 (2019) 15–21
had both UPMC for You for physical health coverage and Community
Care for behavioral health coverage. We were not able to observe
outcomes for individuals who changed enrollment during the ob-
servation period. Although Community Care covers a diverse area of the
state in terms of demographics and geography, we cannot rule out
coverage area as a confounder of results. Use of claims data to define
comorbidities used in our modeling may not fully capture the burden of
disease (Klabunde, Warren, & Legler, 2002).
Data availability also presents limitations. For example, there is
consensus in the literature showing that there is little to no difference in
treatment retention between patients receiving medium or high dose
buprenorphine and those receiving medium or high dose methadone
(Mattick et al., 2014). In our sample, the majority (78.5%) of bupre-
norphine patients received the recommended maintenance dose of
buprenorphine (16 mg) (Greenwald, Comer, & Fiellin, 2014; Indivior
UK Ltd, n.d.). Unfortunately, methadone dosage information is not
available for this sample. Also, ethnicity data is not available on this
Medicaid population, so we were unable to explore its association with
outcomes. Overdoses defined as non-fatal visits to an ED likely under-
counts the actual number of overdoses, as not all are treated in an ED
and some visits end in fatality. Finally, this analysis limited its ex-
amination of costs to the payer's perspective. We were unable to esti-
mate costs to the individual and society as a result of treatment or
overdose.
5. Conclusion
The growing opioid epidemic necessitates comparative research on
available therapies so that individuals affected by opioid use, providers,
payers, policymakers, and other stakeholders can make informed de-
cisions about appropriate interventions. This study finds that there are
complex relationships between MAT adherence, risk of nonfatal over-
dose, and cost of care.
Declaration of Competing Interest
None.
p
0.0013
−2.135 (−213.168, 208.898) 0.984
50.089 (−312.215, 412.393) 0.786
−44.574 (−243.673, 154.525) 0.661
< 0.0001
< 0.0001
−46.236 (−391.469, 298.996) 0.793
59.416 (−377.788, 496.619) 0.790
1173.493 (252.840, 2094.146) 0.013
−697.552 (−1636.610, 241.509) 0.145
0.017
−192.475 (−675.854, 290.903) 0.435
1264.017 (256.801, 2271.233) 0.014
S. Kinsky, et al. Journal of Substance Abuse Treatment 104 (2019) 15–21

Methadone and Buprenorphine New Starts

Medicaid Deidenfied Datamart

Methadone Buprenorphine
Jan 2015—Dec 2016 Jan 2015—Dec 2016
N=955 N=2,274

Exclusions Exclusions
Overlap:
N=260

No overlap No overlap
N=695 N=2,014

<6 Months post and < 3 <6 Months post and < 3
Months prior months prior
membership: N=483 membership: N=1,042

Members for adherence Members for adherence

and OD analysis and OD analysis
N=212 N=972

<9 Months pre and < 9 <9 Months pre and < 9
Months post months post
membership: N=87 membership: N=405

Members for cost analysis Members for cost analysis

N=125 N=567

Fig. A.1. Consort chart of sample selection steps.

References
Barnett, P. G. (2009). Comparison of costs and utilization among buprenorphine and
methadone patients. Addiction, 104(6), 982–992. https://doi.org/10.1111/j.1360-
0443.2009.02539.
Baser, O., Chalk, M., Fiellin, D. A., & Gastfriend, D. R. (2011). Cost and utilization out-
comes of opioid-dependence treatments. American Journal of Managed Care, 17(Suppl.
8), S235–S248.
Campbell, C. I., Bahorik, A. L., VanVeldhuisen, P., Weisner, C., Rubinstein, A. L., & Ray, G.
T. (2018). Use of a prescription opioid registry to examine opioid misuse and over-
dose in an integrated health system. Prevention Medicine, 110, 31–37. https://doi.org/
10.1016/j.ypmed.2018.01.019.
Chou, R., Korthuis, P. T., Weimer, M., Bougatsos, C., Blazina, I., Zakher, B., ... McCarty, D.
(2016). Medication-assisted treatment models of care for opioid use disorder in primary
care settings (vol. AHRQ publication no. 16(17)-EHC039-EF). Rockville, MD: Agency
for Healthcare Research and Quality.
Clark, R. E., Samnaliev, M., Baxter, J. D., & Leung, G. Y. (2011). The evidence doesn't
justify steps by state Medicaid programs to restrict opioid addiction treatment with
buprenorphine. Health Affairs (Millwood), 30(8), 1425–1433. https://doi.org/10.
1377/hlthaff.2010.0532.
Clark, R. E., Baxter, J. D., Aweh, G., O'Connell, E., Fisher, W. H., & Barton, B. A. (2015).
Risk factors for relapse and higher costs among Medicaid members with opioid de-
pendence or abuse: Opioid agonists, comorbidities, and treatment history. Journal of
Substance Abuse Treatment, 57, 75–80. https://doi.org/10.1016/j.jsat.2015.05.001.
DEA (2017). Analysis of drug-related overdose deaths in Pennsylvania, 2016. Drug
Enforcement Administration (DEA) Philadelphia Division & University of Pittsburgh.
DEA (2018). The opioid threat in Pennsylvania. Drug Enforcement Administration (DEA)
Philadelphia Division & University of Pittsburgh.
Greenwald, M. K., Comer, S. D., & Fiellin, D. A. (2014). Buprenorphine maintenance and
mu-opioid receptor availability in the treatment of opioid use disorder: Implications
for clinical use and policy. Drug and Alcohol Dependence, 144, 1–11. https://doi.org/
10.1016/j.drugalcdep.2014.07.035.
Hedegaard, H., Minino, A. M., & Warner, M. (November 2018). Drug overdose deaths in the
United States, 1999–2017. NCHS data brief no. 329 NCHS Data Brief No. 329. National
Center for Health Statistics.
Hser, Y. I., Saxon, A. J., Huang, D., Hasson, A., Thomas, C., Hillhouse, M., ... Ling, W.
(2014). Treatment retention among patients randomized to buprenorphine/naloxone
compared to methadone in a multi-site trial. Addiction, 109(1), 79–87. https://doi.
org/10.1111/add.12333.
Indivior UK Ltd Prescribing information: Suboxone (R) (buprenorphine and naloxone)
sublingual film (CIII). Retrieved April 24, 2019, from https://www.suboxone.com/
pdfs/prescribing-information.pdf.
Kimber, J., Larney, S., Hickman, M., Randall, D., & Degenhardt, L. (2015). Mortality risk
of opioid substitution therapy with methadone versus buprenorphine: A retrospective
cohort study. The Lancet Psychiatry, 2(10), 901–908. https://doi.org/10.1016/S2215-
0366(15)00366-1.
Klabunde, C. N., Warren, J. L., & Legler, J. M. (2002). Assessing comorbidity using claims
data: An overview. Medical Care, IV26–IV35.
Lo Ciganic, W. H., Gellad, W. F., Gordon, A. J., Cochran, G., Zemaitis, M. A., Cathers, T.,
... Donohue, J. M. (2016). Association between trajectories of buprenorphine treat-
ment and emergency department and inpatient utilization. Addiction, 111(5),
892–902. https://doi.org/10.1111/add.13270.
Lynch, F. L., McCarty, D., Mertens, J., Perrin, N. A., Green, C. A., Parthasarathy, S., ...
Pating, D. (2014). Costs of care for persons with opioid dependence in commercial
integrated health systems. Addiction Science & Clinical Practice, 9, 16. https://doi.org/
10.1186/1940-0640-9-16.
Ma, J., Bao, Y. P., Wang, R. J., Su, M. F., Liu, M. X., Li, J. Q., ... Lu, L. (2018). Effects of
medication-assisted treatment on mortality among opioids users: A systematic review
and meta-analysis. Molecular Psychiatry. https://doi.org/10.1038/s41380-018-
0094-5.
Maas, J., Barton, G., Maskrey, V., Pinto, H., & Holland, R. (2013). Economic evaluation: A
comparison of methadone versus buprenorphine for opiate substitution treatment.
Drug and Alcohol Dependence, 133(2), 494–501. https://doi.org/10.1016/j.
drugalcdep.2013.07.018.
Manhapra, A., Petrakis, I., & Rosenheck, R. (2017). Three year retention in buprenorphine
treatment for opioid use disorder nationally in the Veterans Health Administration.
The American Journal on Addictions, 26(6), 572–580. https://doi.org/10.1111/ajad.
12553.
Mattick, R. P., Breen, C., Kimber, J., & Davoli, M. (2014). Buprenorphine maintenance
versus placebo or methadone maintenance for opioid dependence. Cochrane Database
Systematic Review, 2, CD002207. https://doi.org/10.1002/14651858.CD002207.
pub4.
McCarty, D., Perrin, N. A., Green, C. A., Polen, M. R., Leo, M. C., & Lynch, F. (2010).
Methadone maintenance and the cost and utilization of health care among in-
dividuals dependent on opioids in a commercial health plan. Drug and Alcohol
Dependence, 111(3), 235–240. https://doi.org/10.1016/j.drugalcdep.2010.04.018.
Rieckmann, T. R., Gideonse, N., Risser, A., DeVoe, J. E., & Abraham, A. J. (2017).
Treating opioid dependence with buprenorphine in the safety net: Critical learning
from clinical data. The Journal of Behavioral Health Services & Research, 44(3),
351–363. https://doi.org/10.1007/s11414-017-9553-z.
SAS Institute Inc (2014). SAS/STAT® 13.2 user's guide. Cary, NC: SAS Institute Inc.

20
S. Kinsky, et al.
Shrank, W. H., Patrick, A. R., & Brookhart, M. A. (2011). Healthy user and related biases
in observational studies of preventive interventions: A primer for physicians. Journal
of General Internal Medicine, 26(5), 546–550. https://doi.org/10.1007/s11606-010-
1609-1.
Singh, G. K. (2003). Area deprivation and widening inequalities in US mortality,
1969–1998. American Journal of Public Health, 93(7), 1137–1143.
Sordo, L., Barrio, G., Bravo, M. J., Indave, B. I., Degenhardt, L., Wiessing, L., ... Pastor-
Barriuso, R. (2017). Mortality risk during and after opioid substitution treatment:
Systematic review and meta-analysis of cohort studies. BMJ, 357, j1550. https://doi.
org/10.1136/bmj.j1550.
Substance Abuse and Mental Health Services Administration (2006). The determinations
report: A report on the physician waiver program established by the drug addiction treat-
ment act of 2000 (“DATA”). Rockville, MD: SAMHSA.
Tkacz, J., Volpicelli, J., Un, H., & Ruetsch, C. (2014). Relationship between buprenor-
phine adherence and health service utilization and costs among opioid dependent
21
Journal of Substance Abuse Treatment 104 (2019) 15–21
patients. Journal of Substance Abuse Treatment, 46(4), 456–462. https://doi.org/10.
1016/j.jsat.2013.10.014.
Veehof, L., Stewart, R., Haaijer-Ruskamp, F., & Jong, B. M.-d. (2000). The development of
polypharmacy. A longitudinal study. Family Practice, 17(3), 261–267.
Weinstein, Z. M., Kim, H. W., Cheng, D. M., Quinn, E., Hui, D., Labelle, C. T., ... Samet, J.
H. (2017). Long-term retention in office based opioid treatment with buprenorphine.
Journal of Substance Abuse Treatment, 74, 65–70. https://doi.org/10.1016/j.jsat.2016.
12.010.
Wickizer, T. M., Krupski, A., Stark, K. D., Mancuso, D., & Campbell, K. (2006). The effect
of substance abuse treatment on Medicaid expenditures among general assistance
welfare clients in Washington state. Milbank Quarterly, 84(3), 555–576. https://doi.
org/10.1111/j.1468-0009.2006.00458.
Zur, J., & Tolbert, J. (2018). The opioid epidemic and Medicaid's role in facilitating access to
treatment issue brief. Henry J. Kaiser Family Foundation.