Author ORCID Identifier


Defense Date


Document Type


Degree Name

Doctor of Philosophy


Healthcare Policy & Research

First Advisor

Peter Cunningham, PhD

Second Advisor

Andrew Barnes, PhD

Third Advisor

David Harless, PhD

Fourth Advisor

Caitlin Martin, MD, MPH


This dissertation is unified by both policy opportunity and estimation method, examining the impact of Medicaid behavioral health policy in Virginia using a difference-in-differences framework. Medicaid-covered substance use disorder (SUD) treatment benefits in Virginia were significantly enhanced through a Section 1115 SUD waiver, which are required to be budget neutral. Therefore, it is critical to understand whether the objectives of the waiver are met using reliable, unbiased estimation methods. This dissertation includes two empirical research projects evaluating the impact of Medicaid policy on removing barriers in access to care and one methodological project comparing the performance of alternate approaches to difference-in-difference estimation with health policies implemented at different points in time.

Chapter 1 uses data from the National Survey of Substance Abuse Treatment Services (N-SSATS) to assess whether the enhancement in SUD treatment benefits increased Medicaid participation among treatment facilities in Virginia compared to states without similar changes in SUD benefits. Results indicate Virginia’s Section 1115 SUD waiver significantly increased facility acceptance of Medicaid as payment for SUD treatment services relative to comparison states. As this study has a single treated unit, the robustness of the difference-in-difference results was tested using two approaches appropriate for small samples—the synthetic control method and a form of outlier analysis with state-specific linear detrending.

Chapter 2 builds on methodologic advances on staggered difference-in-differences in the econometrics literature, using Monte Carlo simulations and an empirical application to compare the performance of two-way fixed effects (TWFE) with three alternatives to estimating overall treatment effects using extended two-way fixed effects (ETWFE). Across variation in sample size, treatment timing, treatment effects, and comparison groups, ETWFE outperforms TWFE, and weighting ETWFE estimates by the scaled share of units and time treated outperforms weighting by either units or time alone. The simulations and empirical application illustrate the bias of the TWFE-based effect of Medicaid expansion on insurance coverage will likely increase as more states implement Medicaid expansion, necessitating adoption of unbiased estimation approaches like ETWFE.

Chapter 3 uses ETWFE to examine whether the removal of the prior authorization (PA) requirement increases the prescribing rate for buprenorphine—one of the three medications approved by the Food and Drug Administration (FDA) for treatment of opioid use disorder (OUD). Compared to providers subject to PA requirements for the duration of the study, providers significantly increased their buprenorphine prescribing rates after the removal of the PA requirement. The gains in prescribing rates were observed in three of four cohorts of providers and increased over the post-treatment period. These findings suggest removing the PA requirement for buprenorphine is an effective policy solution for Medicaid programs to increase provider prescribing capacity and improve buprenorphine access.


© The Author

Is Part Of

VCU University Archives

Is Part Of

VCU Theses and Dissertations

Date of Submission