Ece AKÇA SALIK, Zafer BÜTÜN, Yeliz KAYA, Özer ÇELİK, Tuğba TAHTA, Arzu ALTUN YAVUZ
Zeynep Kamil Medical Journal - 2026;57(4):199-203
Objective: Preterm labor is a major obstetric challenge that contributes significantly to neonatal morbidity and mortality. This pilot study investigated the feasibility of machine learning techniques for predicting the risk of threatened preterm labor, rather than confirmed preterm birth, among first-time mothers using readily available sociodemographic and obstetric variables. Material and Methods: This retrospective, single-center study included 102 nulliparous women with singleton pregnancies. Five machine learning models were developed using basic clinical parameters. Model performance was evaluated using an 80/20 training-test split and assessed based on accuracy, ROC-AUC, sensitivity, and specificity. Results: The XGBoost classifier demonstrated the best performance, with an accuracy of 70% and a ROC-AUC of 0.64. Body mass index and a history of abortus imminens were the most influential predictors. Conclusion: Machine learning models based on simple clinical data can modestly predict threatened preterm labor. Although these models are not intended for immediate clinical application, the findings demonstrate feasibility and support further large-scale, externally validated studies incorporating maternal and neonatal outcomes.
