Utilization of Machine Learning in Patient Admission into Intensive Care Units

Abstract

Intensive care units (ICUs) around the world suffer from capacity-related problems, the biggest of which involve the over-utilization of beds, which can be attributed to the inaccurate decisions of admitting patients into the ICU and the volatile demand. This problem is of significance because it correlates with hospital resources and the well-being of patients. In an attempt to solve this problem several solutions are presented in this paper ranging between big data approaches, mathematical models, software, and simulation. The paper presents an approach where machine learning techniques to determine the best discharge location based on patient diagnostic and biographical data. The best machine learning method was selected based on defined criteria through Analytical Hierarchy Process (AHP). Based on a large data set, decision trees were found to be the best choice due to its accuracy and interpretability. The proposed decision tree can be used to reduce the utilization of ICU beds by assisting the nurses in decision making of dispatching patients.