Risk Stratification of Neonates using Machine Learning Techniques

Abstract

The process of classifying newly born babies into high-risk and low-risk is called risk stratification. Having a platform to stratify neonates according to severity of risk is the key to the success of any Neonatal Intensive Care Unit (NICU). The premature neonates are at a higher risk of developing the disabilities which could affect their future growth. However, the extent at which this can affect their entire life, strongly depends on how early they were born, the quality of care they received during and around birth and the days they follow in NICU. Establishing a decision support tool using machine learning algorithms will be useful for identifying neonates who are at high risk for proper diagnosis and efficient monitoring of neonates at NICU. The paper is focused on risk stratification of neonates using machine leaning algorithms such as Artificial Neural Network (ANN), K Nearest Neighbors (KNN) and Support Vector Machine (SVM). Furthermore, various evaluation parameters were used for comparing the results of the algorithms on the 66 cases of neonates admitted at Kasturba Medical College, Manipal. Based on Area Under Curve (AUC), ANN (0.91) performed better than KNN (0.83) and SVM (0.84). The result indicates the significant contribution of ANN with improved performance in identifying neonates who are at high risk better than other algorithms