Machine Learning Decision Support for Macronutrients in Neonatal Parenteral Nutrition

Abstract

Parenteral nutrition plays a crucial role in the care of hospitalized patients in the neonatal intensive care unit. Determining the appropriate amount and composition of parenteral nutrition bag for each infant is a complex and individualized process. Using a dataset of 1210 neonatal intensive-care unit patients of the University Hospital of Perugia, collected over 17 years, this work aims to establish the basis for an evidence-based decision support system to reduce the time and effort required for manual calculations from doctors working in such a critical emergency environment. After the data was presented, we compared different machine learning techniques (i.e., random forest, gradient boosting machine, support vector machine, multilayer perceptron) that could predict nutritional requirements. We discuss the feasibility of the proposed approach, evaluating the methods in terms of their explainability, and using performance measures (i.e., RMSE, $R^{2}$ ). Preliminary results revealed promising predictive ability for macronutrients and volume of parenteral bags. These findings highlight the potential of machine learning as a valuable tool for nutritional outcome estimation in neonatal clinical practice.

Keywords

• Machine learning
• intelligent decision systems
• neonatal intensive care
• AI for decision in healthcare
• machine learning for parenteral nutrition
• infant nutrition