Predicting Risk for Patent Ductus Arteriosus in the Neonate: A Machine Learning Analysis

Predicting Risk for Patent Ductus Arteriosus in the Neonate: A Machine Learning Analysis

<i>Background and Objectives</i>: Patent ductus arteriosus (PDA) is common in newborns, being associated with high morbidity and mortality. While maternal and neonatal conditions are known contributors, few studies use advanced machine learning (ML) as predictive factors. This study asse...

Full description

Saved in:
Bibliographic Details
Main Authors: Ana Maria Cristina Jura, Daniela Eugenia Popescu, Cosmin Cîtu, Marius Biriș, Corina Pienar, Corina Paul, Oana Maria Petrescu, Andreea Teodora Constantin, Alexandru Dinulescu, Ioana Roșca
Format: Article
Language:English
Published: MDPI AG 2025-03-01
Series:Medicina
Subjects:
patent ductus arteriosus
neonatal risk
machine learning
maternal pathology
neonatal outcomes
Online Access:https://www.mdpi.com/1648-9144/61/4/603
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:<i>Background and Objectives</i>: Patent ductus arteriosus (PDA) is common in newborns, being associated with high morbidity and mortality. While maternal and neonatal conditions are known contributors, few studies use advanced machine learning (ML) as predictive factors. This study assessed how maternal pathologies, medications, and neonatal factors affect the risk of PDA using traditional statistics and ML algorithms: Random Forest (RF) and XGBoost (XGB). <i>Materials and Methods</i>: A retrospective 3-year cohort study of 201 NICU neonates assessed maternal and neonatal factors. Logistic regression (LR) and chi-square analyses identified significant predictors, while ML models enhanced predictive accuracy and pinpointed key PDA factors. <i>Results</i>: LR identified prolonged rupture of membranes (>18 h) as the most significant predictor (OR: 13.03, <i>p</i> < 0.001). The ML models identified gestational age, maternal anemia, prenatal care level, birth weight, prolonged rupture of membranes, medication usage, diabetes, pregnancy-induced hypertension, SARS-CoV-2 infection, and cervical cerclage as key predictors. The RF model had 76.3% accuracy, moderate sensitivity (47.4%), and high specificity (90%). XGB performed better with 81.4% accuracy, an AUC of 0.872, sensitivity of 92.5%, and specificity of 57.9%. <i>Conclusions</i>: This study shows that maternal and neonatal factors significantly influence the risk of PDA. ML, particularly XGBoost, enhances predictive abilities, guiding targeted interventions and improving neonatal outcomes.
ISSN:1010-660X
1648-9144

Similar Items