Estimating Realized Access to Obstetric Care in Georgia: A Discrete Choice Modeling Analysis
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
The ongoing closure of obstetric units in the United States poses challenges to maternal care access, especially in rural communities. These closures leave patients with limited options for obstetric services and may force them to travel longer distances or rely on low-volume facilities, both associated with adverse maternal outcomes. While policymakers have proposed interventions to support obstetric care in underserved areas, evaluating their potential impact requires understanding how birthing patients select among available facilities and how those selections affect delivery volume and access patterns. This study develops a revealed preference discrete choice modeling framework to estimate realized access to obstetric facilities—the extent to which patients actually use available obstetric services—based on patient characteristics and facility attributes. Using 2016-2019 birth records in the state of Georgia, we developed three discrete choice models: a distance-only model, a multinomial logit model based on facility attributes, and a latent class model that considers heterogeneous weights patients place on facility attributes. We evaluated model performance in predicting low-volume facilities and regional access. The latent class model outperformed others and identified two patient classes: “less distance-sensitive” patients (34.5%) and “more distance-sensitive” patients (65.5%) with differing sensitivity to distance and levels of care. Our findings reveal regional patterns where patients disproportionately seek care far away or at low-volume facilities, and how these patterns may change as obstetric units close. We highlight the practical use of our model for forecasting underutilized facilities and informing policies to better distribute patient volume across the region.
Highlights
-
We analyzed how pregnant people in Georgia, USA sought obstetric care using statewide birth records from 2016-2019
-
We developed and compared three discrete choice models (distance-only model that assumes patients select nearest obstetric facilities, multinomial logit model, and latent class model) that considered patient characteristics and facility attributes to predict patients’ utilization of birth facilities and validated these models using out-of-sample datapoints.
-
Travel distance, facility level of care, rurality of facility location, and health system affiliation significantly influenced patients’ selection of delivery facilities. We also identified two classes of patients who varied in their social determinants of health and placed heterogenous weights on facility attributes that affected their selection of obstetric facilities.
-
Targeted interventions to improve maternal care access in underserved areas should consider patients’ underlying social determinants of health, as geographic proximity alone is not predictive of patients’ selection of delivery facility.
-
We demonstrate how our model can be applied to forecast low-volume obstetric facilities and access to care and to inform policies that avoid underutilized facilities and better distribute patient volume across the region.
