Table 3.

Assessment of risk of bias for non-randomized studies.

#	Author, date	Was there a comparison?	How were the individuals/groups assigned?	Was the baseline allocation to intervention prospective?	Was outcome assessment prospective?	Was hypothesis generation prospective?	Study quality rating 1-9★
Psychosocial Interventions
9	Bloom-Foster and Mehl-Madrona (2020)	No	All patients initiating outpatient OUD treatment with buprenorphine	Yes	Yes	Probably Yes	n/a
10	Brown et al. (2014)	No	Patients recruited from the community	Yes	No Information	No Information	n/a
11	Montoya et al. (2005)	No	Secondary analysis RCT	No	Probably No	Probably No	n/a
12	Moore et al. (2012)	Yes	By therapist availability	Yes	Yes	Yes	7 stars
13	Moore et al. (2016)	Yes	Secondary analysis, RCT	No	Probably Yes	Probably No	5 stars
14	Weiss et al. (2014)	Yes	Secondary analysis, RCT	No	Probably Yes	Probably No	6 stars
Technological Interventions
17	Schuman-Olivier et al. (2018)	No	“Unstable young adults struggling in OBOT” were referred by buprenorphine prescribers.	Yes	Yes	Yes	n/a
Complementary & Integrative Treatments
19	Lander et al. (2018)	Yes	Volunteers stable on buprenorphine for 90+days, control group matched on key attributes (age, sex, days abstinent)	Yes	Yes	Probably Yes	5 stars
Structural Barriers to Care
20	Cochran et al. (2018)	No	Treatment-seeking volunteers	Yes	Probably Yes	Probably Yes	n/a

Risk of bias (ROB) for studies with a comparison was evaluated using the Newcastle-Ottawa ROB assessment tool. Nine stars represents the lowest risk of bias, 1 star the highest risk of bias. Studies without a comparison arm were not scored using the tool and are listed as n/a in the final column.