Estimating the Impact of Medication Assisted Opioid Treatment

The number of opioid prescriptions tripled between 1990 and 2014, with coincident escalations in opioid addiction rates and opioid-related overdose deaths. These trends have led to extensive supply side oriented policies. However, the possibility of unintended consequences and restricting access for patients that have a clinical need suggests that demand-side interventions are an essential component of a comprehensive response to U.S. opioid epidemic. One such policy is expanding access to buprenorphine or buprenorphine/naloxone medication-assisted treatment for opioid addiction (hereafter referred to as “buprenorphine MAT”). Our study estimates a reduced-form effect of increasing buprenorphine MAT use on reducing prescription opioid use and opioid overdose, assuming that the reduced-form effect comes primarily from the structural effect through reducing opioid addiction.

Our primary data source is the Medicaid Drug Utilization (MDU) files from CMS. States are required to report to the CMS on prescription activities of all Medicaid-covered outpatient drugs in exchange for federal matching funds. Each drug product in the MDU files is identified by a National Drug Code (NDC) number, which allows us to identify buprenorphine medications for MAT, Schedule-II prescription opioids, and injectable naloxone medications for opioid overdose treatment. 

The first study outcome is the use of addictive prescription opioids, measured by the per-capita amounts of Medicaid prescriptions for, and spending on, Schedule-II prescription opioids. The second outcome is the incidence of opioid overdose, which we proxy for using the per-capita amounts of Medicaid prescriptions for and spending on injectable naloxone medications. The independent variable of interest is the use of buprenorphine MAT, measured by the per-capita amounts of Medicaid prescriptions for and spending on buprenorphine MAT.

To address the potential endogeneity of buprenorphine MAT use with respect to addictive prescription opioid use and opioid overdose incidents, we included county, quarter two-way fixed effects to isolate the within-state variations in buprenorphine MAT use over time. We further exploited the exogenous variations in buprenorphine MAT use induced by availability of the Drug Addiction Treatment Act (DATA)-waived physicians qualified for buprenorphine MAT prescription within 30- and 100-patient limits (hereafter referred to as “30-patient-waived physicians” and “100-patient-waived physicians”). The preferred models were estimated using two-stage least squares (TSLS) with two-way fixed-effect in both stages. All estimates were population-weighted and state-clustered to correct for heterogeneous policy effect and the within-state serial correlation in a difference-in-differences context.

In the first stage we find that the availability of 100-patient-waived physicians was strongly associated with increases in Medicaid prescriptions for and spending on buprenorphine MAT; the availability of 30-patient-waived physician was also associated with increased buprenorphine MAT prescriptions and spending, albeit significant only at the 0.10 level. The second-stage main estimates suggest that one more Medicaid prescriptions for buprenorphine MAT per 1,000 enrollees reduced Medicaid opioid prescriptions by 1.14 per 1,000 enrollees (state-clustered S.E.=0.37); $1 more Medicaid spending on buprenorphine MAT per 1,000 enrollees reduced the opioid spending by $0.69 per 1,000 enrollees (state-clustered S.E.=0.20). Estimated reductions in opioid prescriptions and spending were concentrated largely in Oxycodone, Hydrocodone, and Oxymorphone medications.