[A study of factors associated with neonatal necrotizing enterocolitis]

Area of Science:

• Neonatal Medicine
• Data Science
• Biostatistics

Context:

• Neonatal necrotizing enterocolitis (NEC) is a significant cause of morbidity and mortality in premature infants.
• Accurate identification of NEC risk factors is crucial for timely intervention and improved outcomes.
• Previous studies have explored various risk factors, but advanced predictive modeling can enhance understanding.

Purpose:

• To identify and compare risk factors associated with neonatal necrotizing enterocolitis (NEC) using nine distinct regression models.
• To determine the optimal predictive model for NEC risk assessment through rigorous comparison and validation.
• To elucidate the specific risk and protective factors for NEC using explainable AI techniques.

Summary:

• A Gradient Boosting Decision Tree (GBDT) model demonstrated superior performance in predicting NEC risk compared to eight other regression models.
• The GBDT model achieved an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.958, with 92.5% accuracy.
• Key identified risk factors for NEC include anemia, non-invasive ventilation, procalcitonin use, premature birth, and low birth weight. Breastfeeding and probiotics were found to be protective factors.

Impact:

• The study provides a robust, data-driven approach to identifying infants at high risk for NEC.
• Findings offer valuable insights for developing targeted prevention and treatment strategies for NEC.
• The use of explainable AI (Shap analysis) enhances the clinical interpretability of the predictive model, facilitating evidence-based decision-making.