Methodological Advances in Population Health: Policy Variation, Geospatial Access, and Target Trial Emulation with Applications to Abortion

Abstract

Rigorous evaluation of abortion access and effectiveness in the United States is challenged by structural policy variation, geographic heterogeneity, and the ethical and logistical constraints of randomized trials. Generating policy-relevant evidence in this setting requires methodological approaches that can leverage observational data to study both access to care and clinical outcomes under real-world conditions. In this dissertation, I apply and integrate geospatial methods and target trial emulation to evaluate abortion access and comparative effectiveness in populations disproportionately affected by structural inequities, particularly those impacted by the Hyde Amendment, including communities on Native lands and enrolled in Medicaid. In Chapter 1, I quantify changes in geographic access to abortion facilities for individuals residing on Native lands before and after the Supreme Court’s decision in Dobbs v. Jackson Women’s Health Organization. Using geospatial accessibility measures, I estimated changes in service coverage of Native lands by abortion clinics as well as travel to the nearest abortion clinic. I demonstrate that the post-Dobbs policy landscape substantially increased travel burdens for many Native communities, compounding pre-existing geographic inequities. In Chapter 2, I assess the geographic availability of crisis pregnancy centers (CPCs) near Native lands. CPCs deter access and divert funding from reproductive healthcare. Indigenous people have been found to be among those populations most proximate to CPCs while also being furthest from abortion care. Using proximity-based measures, I evaluate the presence of CPCs relative to abortion clinics around Native lands. This chapter extends spatial access metrics to examine the potential for structural interference in access to reproductive healthcare. I show that CPCs are more geographically proximate to Native lands than abortion clinics, effectively crowding out services, demonstrating that the role of CPCs may be important when considering Native populations, especially those on Native lands, and structural barriers to comprehensive, evidence-based reproductive healthcare. In Chapter 3, I estimate the comparative effectiveness of mifepristone–misoprostol medication abortion versus vacuum aspiration procedural abortion using a target trial emulation framework in Medicaid claims data. I explicitly specify the protocol of a hypothetical randomized trial, define eligibility criteria, treatment strategies, assignment procedures, and outcomes, and estimate both intention-to-treat and per-protocol effects on complete abortion. By applying causal inference methods to large-scale administrative data, this study generates real-world effectiveness estimates in a population often underrepresented in clinical trials and directly affected by coverage restrictions under the Hyde Amendment. By leveraging large-scale administrative data and rigorous causal methods, I generate real-world evidence showing high effectiveness of contemporary abortion practice among a large US Medicaid population. Collectively, this dissertation demonstrates how advancing methods—from geospatial accessibility measures to target trial emulation—can support the study of abortion access and effectiveness. By formalizing causal questions, leveraging policy variation, and quantifying geographic constraints, this work provides advances in methods for evaluating reproductive healthcare under structural policy inequities.