Development and Validation of an Algorithm to Identify Prenatal Care in Administrative Data: Predictive Validity for Adverse Birth Outcomes
Health Services Research · 2026
DOI: 10.1111/1475-6773.70063 ↗Kurz gefasst
Validierungsstudie in der Zeitschrift Health Services Research, 2026. Verfasst von Songyuan Deng und 2 weiteren Autor:innen. Beteiligte Einrichtung: South Carolina Center for Rural and Primary Healthcare, University of South Carolina School of Medicine, Columbia, South Carolina, USA. Hauptthemen laut MEDLINE: Algorithms, Prenatal Care, Pregnancy Outcome. Die Arbeit nennt 28 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Objective
To develop and validate a hierarchical algorithm for assigning prenatal care (PNC) encounters using claims data while ensuring continuity of care. Study Setting and Design We conducted a retrospective cohort study among South Carolina Medicaid beneficiaries. Using a six‐step hierarchical algorithm—incorporating specialty designations, diagnostic/procedure codes, and adjustments for inpatient stays and supplemental visits—we assigned PNC encounters and identified predominant PNC providers. To assess predictive validity, we examined associations between predominant provider status and adverse birth outcomes (obtained from linked birth certificates and claims data) using logit‐binomial generalized estimating equations with robust standard errors, and we compared models' performance using both model fit statistics and 10‐fold cross‐validation. Data Sources and Analytic Sample We used South Carolina Medicaid data on live‐birth pregnancies from 2016 to 2021. We followed participants from conception until delivery. Principal Findings Initial screening identified 302 package/bundle payment claims, leading to the exclusion of 299 pregnancies (0.3%) from further analysis. The final analytic dataset contained 1,072,615 confirmed PNC encounters for 90,581 (97%) pregnancies. This study identified predominant providers for 87,573 pregnancies (98% of cases with at least two PNC encounters). The analysis of predictive validity revealed significant protective associations for two outcomes when comparing pregnancies with versus without predominant providers: preterm birth (adjusted RR: 0.68, 95% CI: 0.59–0.77) and low‐birth‐weight (adjusted RR: 0.68, 95% CI: 0.57–0.80).
Conclusions
This study developed and validated a claims‐based algorithm to identify PNC utilization in South Carolina Medicaid data. Predictive validity tests revealed that predominant provider status was associated with reduced adverse birth outcomes, suggesting care continuity may improve perinatal health. Future research could apply this algorithm to examine causal relationships between predominant provider status and specific outcomes (e.g., preterm birth, low birth weight), while accounting for institutional and socioeconomic confounders. These findings offer a foundation for optimizing PNC delivery through continuity‐focused interventions.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- South Carolina Center for Rural and Primary Healthcare, University of South Carolina School of Medicine, Columbia, South Carolina, USA.
- Clear Bell Solutions, Charleston, South Carolina, USA.
Zitieren
APA
Deng, S., Barabell, G., & Bennett, K. J. (2026). Development and Validation of an Algorithm to Identify Prenatal Care in Administrative Data: Predictive Validity for Adverse Birth Outcomes. Health Services Research. https://doi.org/10.1111/1475-6773.70063
Vancouver
Deng S, Barabell G, Bennett KJ. Development and Validation of an Algorithm to Identify Prenatal Care in Administrative Data: Predictive Validity for Adverse Birth Outcomes. Health Services Research. 2026. doi:10.1111/1475-6773.70063
BibTeX
@article{Deng2026,
author = {Deng, Songyuan and Barabell, Greg and Bennett, Kevin J},
title = {Development and Validation of an Algorithm to Identify Prenatal Care in Administrative Data: Predictive Validity for Adverse Birth Outcomes},
journal = {Health Services Research},
year = {2026},
doi = {10.1111/1475-6773.70063}
}
EUVIMED