AI-Driven Approach to Predict High-Risk Newborns to Reduce NICU Admission Overcrowding

This study presents an AI-driven approach designed to identify high-risk newborns and reduce unnecessary admissions to the Neonatal Intensive Care Unit (NICU). The initiative addresses critical challenges in many low- and middle-income countries (LMICs), this has resulted in transferring relatively stable newborns with risk factors who require observation and minimal support to NICU that are already overcrowded and understaffed. To outcome these issues, the paper proposes the use of machine learning algorithms including Support Vector Machine (SVM), Random Forest (RF), K-Nearest Neighbor (KNN), and Gradient Boosting Machine (GBM), These models are trained on parameters such as Fetal Heart Rate (FHR), Base Deficit in the Extracellular Fluid (BDecf), pH, and pCO2. Using these parameters, the model aims to predict the feasibility of admitting newborns to the NICU, enabling health care providers to prioritize and optimize resource allocation. The methodology involves training AI models on public (open access) historical neonatal data to classify newborns into different risk cases. The accuracy and efficiency of the model will look at comparing conventional treatment strategies. The aim is to improve neonatal outcomes in resource constrained settings.