Modified Claims Algorithm Improves Spontaneous Abortion Identification in Large Commercial Database

A modified hierarchical algorithm applied to United States commercial health insurance claims identified spontaneous abortion in 17.5 percent of pregnancy episodes, improving classification sensitivity compared to the 14.9 percent incidence reported by earlier analytical models. In a retrospective cohort study conducted by researchers Megan Fuerst, Kaitlin Schrote, and Maria I. Rodriguez, the updated framework also successfully reduced unclassified or ambiguous pregnancy outcomes from 17.4 percent down to 13.4 percent across the evaluated records.

Distinguishing early pregnancy events within administrative healthcare databases presents persistent analytical difficulties for health researchers. According to the study authors, accurately identifying spontaneous abortion and induced abortion compared to other pregnancy outcomes remains challenging because these events frequently share billing codes. In addition, the research team highlighted that identifying these clinical episodes is further complicated by a widespread lack of reimbursement by many insurance providers.

To address these classification limitations, the investigators analyzed a cohort of 2,418,986 pregnancy episodes documented within the Merative MarketScan Commercial Database between 2018 and 2024. The authors adapted a previously published hierarchical algorithm specifically designed to reduce unknown or ambiguous pregnancy outcomes. They then systematically tested the assumptions of the hierarchy by alternating the ordering of diagnosis codes, procedural codes, and diagnosis-related group codes for spontaneous abortion and induced abortion within the algorithm structure.

Altering the algorithm hierarchy to prioritize induced abortion over spontaneous abortion identified an additional 8,623 induced abortion cases across the commercial dataset. However, even with this targeted adjustment, overall identification of induced abortion remained low, ultimately accounting for less than 5 percent of all evaluated pregnancy episodes.

The authors emphasized that transparent research methodology and precise identification of early pregnancy claims have become especially critical for researchers following the Dobbs decision. Although their modified algorithm improved sensitivity for spontaneous abortion and reduced unknown outcomes, the team concluded that claims-based analyses may still systematically undercount induced abortion, presenting significant implications for researchers attempting to inform evidence-based reproductive health policy.

Subject of Research: Hierarchical algorithm sensitivity for classifying early pregnancy outcomes in commercial health insurance claims

Article Title: Sensitivity of spontaneous abortion classification to hierarchical algorithm assumptions in US commercial claims data, 2018–2024

Article References: Fuerst, M., Schrote, K., & Rodriguez, M. I. (2026). Sensitivity of spontaneous abortion classification to hierarchical algorithm assumptions in US commercial claims data, 2018–2024. PLOS One, 21(10), e0361330. https://doi.org/10.1371/journal.pone.0361330

Image Credits: AI Generated

DOI: 10.1371/journal.pone.0361330

Keywords: spontaneous abortion, induced abortion, commercial claims data, Merative MarketScan, hierarchical algorithms, pregnancy outcomes, retrospective cohort study, billing codes, diagnosis-related groups, procedural codes, administrative healthcare data, Dobbs decision, reproductive health policy, algorithm sensitivity, unclassified pregnancy outcomes, early pregnancy claims