Title: Operations Research for Improved and Equitable Maternal Health
Date: May 1st, 2025
Time: 10 AM – 12:00 PM EST
Meeting Link: Microsoft Teams
Meghan E. Meredith
Ph.D. Candidate in Operations Research
School of Industrial and Systems Engineering
Georgia Institute of Technology
Committee:
Dr. Lauren N. Steimle (Advisor)
School of Industrial and Systems Engineering, Georgia Institute of Technology
Dr. Gian-Gabriel Garcia
School of Industrial and Systems Engineering, Georgia Institute of Technology
Dr. Pinar Keskinocak
School of Industrial and Systems Engineering, Georgia Institute of Technology
Dr. David Goldsman
School of Industrial and Systems Engineering, Georgia Institute of Technology
Dr. Sheree Boulet
Department of Obstetrics & Gynecology, Boston University Chobanian & Avedisian School of Medicine, Boston Medical Center
Abstract:
The United States (U.S.) rate of pregnancy-related deaths is the highest among developed countries and is increasing, even as upwards of 80% of these deaths are preventable. Additionally, there are staggering racial/ethnic and urban/rural disparities in maternal outcomes. These poor health outcomes and persistent disparities point to systemic issues and the need for evidence-based systems-level solutions. The complexities of the U.S. maternal healthcare system are extremely difficult to capture and study using classical methods in medicine and epidemiology. In this thesis, we present operations research approaches that are capable of incorporating these complexities to inform policy that optimizes delivery of and access to maternal healthcare in the U.S.
In the first technical chapter, we assess the racial and ethnic disparities in pre-pregnancy conditions and analyze their contribution to disparities in adverse maternal outcomes using data from an observational cohort of first-time mothers. Using logistic regression and causal inference methods, we find that non-Hispanic Black race/ethnicity is significantly associated with adverse maternal outcomes, which is consistent with existing literature. However, we find that accounting for pre-pregnancy conditions, specifically cardiovascular conditions, explains some of the elevated risk of non-Hispanic Black patients experiencing severe preeclampsia. This result suggests that the prevention and management of pre-pregnancy conditions could be an important factor in decreasing racial and ethnic disparities in adverse maternal outcomes.
In the remaining technical chapters, we focus on informing systems-level solutions to maintain and improve access to obstetric care. In the second technical chapter, we evaluate existing measures of access to obstetric care in the U.S., including the county-based “maternity care deserts” measure. Using optimization models, we determine the implied facility location policy implications of these measures. In a state like Georgia, with many small counties, eliminating “maternity care deserts” would require a prohibitively large number of new obstetric hospitals, suggesting that additional tools are needed to estimate the optimal number and distribution of hospitals to meet obstetric care needs.
In the third technical chapter, we use mathematical modeling to better understand the implications of the widespread trend of obstetric hospital closures on travel distance to care and delivery volume of care. Most rural residents travel far distances for obstetric care and are more likely to deliver in hospitals with low delivery volume, which are associated with poor maternal outcomes. We propose a multi-criteria bilevel optimization model to characterize the trade-off between travel distance and delivery volume. We find that a single obstetric hospital closure would increase affected patients’ travel distance by an average of 4.6 miles but would decrease the proportion of deliveries that occur in low-volume rural hospitals from 8.1% to 6.9%, which may lead to maternal outcome gains due to consolidated delivery volume. This work emphasizes that travel distance and volume of care cannot be considered in isolation to maintain access to high-quality care.
In the fourth technical chapter, we extend our previous work to incorporate bypassing behavior, where pregnant women do not seek labor & delivery care at their closest obstetric hospital. We assume that pregnant women seek care according to a random utility maximization model, which we integrate into a facility location model to determine worst-case obstetric closures. This modeling framework, the Maximal Choice-Based Expectation Facility Location Problem, extends existing literature beyond the maximum capture objective to a generalized choice-based expectation. We explore the properties of this model and design decomposition methods to solve large-scale instances. We then use this framework to determine the obstetric closures that would maximize choice-based travel distance, to identify which obstetric hospitals are vital to maintaining access in a maternal healthcare system.
The overall objective of my thesis work is to better understand and improve the delivery of maternal healthcare at an individual- and systems-level to reduce adverse maternal health outcomes and disparities.