BACKGROUND: Buprenorphine is the most frequently prescribed medication for treating substance use disorders in the United States, but few studies have evaluated the structure of treatment delivered in real-world settings. The purpose of this study is to investigate adherence to current buprenorphine treatment guidelines using administrative data for Massachusetts Medicaid. METHODS: We identified buprenorphine treatment episodes beginning in 2009 through pharmacy claims. We then used service claims to identify treatment-related physician, behavioral, and laboratory services received in the induction, stabilization, and maintenance phases of these treatment episodes. Rates of service utilization were compared with those recommended in treatment guidelines. RESULTS: A total of 3674 treatment episodes met inclusion criteria, representing 3005 unique Medicaid beneficiaries. Liver enzymes were tested in 47.3% of episodes, but testing for hepatitis C (23.2%), hepatitis B (19.6%), and human immunodeficiency virus (HIV; 13.7%) was less frequent. Adherence to recommended physician visit frequency was 37.6% during induction, 39.7% during stabilization, and 51.2% during maintenance. For behavioral care, adherence rates were 40.0% during induction, 41.2% during stabilization, and 41.0% during maintenance. Rates of toxicology testing met or exceeded recommendations in just over 60% of episodes in the induction (61.1%), stabilization (62.1%), and maintenance (61.4%) phases. Although rates varied by treatment phase, substantial proportions of episodes showed no evidence of physician visits (27.2-42.8%), behavioral care (44.3-60.0%), and toxicology screening (25.3-39.0%). CONCLUSIONS: Our data suggest that there is significant variability in the structure of buprenorphine treatment provided to Massachusetts Medicaid beneficiaries, and that half or less of episodes include physician and behavioral visits at recommended frequencies. The use of administrative data for this type of analysis is limited by the potential for missing or inaccurate data. More research is needed to establish the levels of services most closely associated with positive outcomes to help guide providers in offering the highest-quality care.