Leukocyte Subtype Variations in Preterm Births: Insights from Robson Group 10 Using Machine Learning and Conventional Analysis

Objective: This study aims to investigate the differences in leukocyte subtypes between preterm births classified as Robson Group 10 and term deliveries, focusing on how these hematological changes are evaluated using both conventional statistical methods and machine learning models like SHAP analysis. Methods: A retrospective case-control study was conducted at Ankara Etlik City Hospital between May and December 2023. Data from 2662 patients, including preterm deliveries (Robson Group 10) and term pregnancies (control group), were analyzed. Hematological parameters, such as white blood cells (WBCs), neutrophils, monocytes, and eosinophils, were compared between groups. Conventional statistical tests, including the Mann-Whitney U and logistic regression analyses, were employed. Additionally, machine learning models, such as XGBoost and SHAP analysis, were used to identify individual variations in leukocyte subtypes. Results: The Robson Group 10 showed significantly higher levels of neutrophils and total WBCs compared to the control group, while monocyte and eosinophil levels were significantly lower. Machine learning analysis revealed that higher immature granulocyte counts were predictive of Robson Group 10, while higher monocyte counts were predictive of the control group. Conclusion: Preterm births in Robson Group 10 exhibit distinct hematological profiles, with notable inflammatory markers, such as increased neutrophils and immature granulocytes. The combined use of conventional statistics and machine learning offers valuable insights into individual variations, suggesting a more personalized approach in evaluating preterm labor risks. Further studies with larger cohorts are needed to refine these findings.