Sensitivity of spontaneous abortion classification to hierarchical algorithm assumptions in US commercial claims data, 2018–2024
- 1Division of Complex Family Planning, Oregon Health & Science University, Portland, Oregon, United States of America.
- 2Center for Reproductive Health Equity, Department of Obstetrics and Gynecology, Oregon Health & Science University, Portland, Oregon, United States of America.
Abstract
Background: Identifying spontaneous abortion (SAB) and induced abortion (IAB) in administrative data compared to other pregnancy outcomes is challenging due to shared billing codes and lack of reimbursement by many insurers. Methods: We performed a retrospective cohort study using the Merative MarketScan Commercial Database from 2018–2024. We adapted a previously published hierarchical algorithm to reduce the number of unknown or ambiguous pregnancy outcomes. We then tested the assumptions of the hierarchy by alternating the order of the diagnosis, procedural, and diagnosis-related group (DRG) codes within the algorithm for SAB and IAB. Results: Our dataset included 2,418,986 pregnancy episodes. Our modified algorithm improved the sensitivity for SAB, identifying a 17.5% incidence compared to 14.9% for previously published algorithms and reduced unclassified outcomes from 17.4% to 13.4%. Prioritizing IAB in the algorithm hierarchy over SAB identified an additional 8,623 IAB cases. However, IAB identification remained low at less than 5% of pregnancy episodes. Conclusions: Following the Dobbs decision, it is critical that researchers accurately identify early pregnancy claims and remain transparent in their methodology. Our algorithm improved sensitivity for SAB and reduced unknown outcomes. However, claims-based analyses may systematically undercount IAB, with implications for research-informed reproductive health policy.
Related
- A General Approach for Assessing Sensitivity to Violations of Causal Assumptions
- Mathematical Classification of Switching Types and Hierarchical Structures: Three Switching Types, Six Hierarchical Structures, Coefficient of Variation Lower Bound, and Discretization Plateau Effect in the k-Regime Linear Switching Model
- Existence and Absence of Spin-Glass Phases on Migdal--Kadanoff Hierarchical Lattices: A Complete Classification
- Hierarchical Majority Decomposition for Imbalanced Classification: A Recursive Binary Majority-Class Tree with Persistent Minority Comparison
- Maximum principle for quasilinear SPDE’s on a bounded domain without regularity assumptions
BibTeX
@article{fuerst2026sensitivity,
title = {{Sensitivity of spontaneous abortion classification to hierarchical algorithm assumptions in US commercial claims data, 2018–2024}},
author = {Megan Fuerst and Kaitlin Schrote and Maria I Rodriguez},
journal = {PLOS One},
year = {2026},
month = oct,
doi = {10.1371/journal.pone.0361330},
url = {https://doi.org/10.1371/journal.pone.0361330},
note = {CC BY},
}
RIS
TY - JOUR
TI - Sensitivity of spontaneous abortion classification to hierarchical algorithm assumptions in US commercial claims data, 2018–2024
AU - Megan Fuerst
AU - Kaitlin Schrote
AU - Maria I Rodriguez
PY - 2026
DA - 2026/10/28
JO - PLOS One
DO - 10.1371/journal.pone.0361330
UR - https://doi.org/10.1371/journal.pone.0361330
AB - Background: Identifying spontaneous abortion (SAB) and induced abortion (IAB) in administrative data compared to other pregnancy outcomes is challenging due to shared billing codes and lack of reimbursement by many insurers. Methods: We performed a retrospective cohort study using the Merative MarketScan Commercial Database from 2018–2024. We adapted a previously published hierarchical algorithm to reduce the number of unknown or ambiguous pregnancy outcomes. We then tested the assumptions of the hierarchy by alternating the order of the diagnosis, procedural, and diagnosis-related group (DRG) codes within the algorithm for SAB and IAB. Results: Our dataset included 2,418,986 pregnancy episodes. Our modified algorithm improved the sensitivity for SAB, identifying a 17.5% incidence compared to 14.9% for previously published algorithms and reduced unclassified outcomes from 17.4% to 13.4%. Prioritizing IAB in the algorithm hierarchy over SAB identified an additional 8,623 IAB cases. However, IAB identification remained low at less than 5% of pregnancy episodes. Conclusions: Following the Dobbs decision, it is critical that researchers accurately identify early pregnancy claims and remain transparent in their methodology. Our algorithm improved sensitivity for SAB and reduced unknown outcomes. However, claims-based analyses may systematically undercount IAB, with implications for research-informed reproductive health policy.
SN - 1932-6203
ER -