Machine learning for predicting normalisation of physiology in hospital

Hospital doctors currently make a decision on when patients are ready to discharge from hospital based on clinical experience, the information that they can read in the patient’s clinical record, and the availability of appropriate support in the community. This project aims to create a tool that uses information from a patient’s clinical record to estimate accurately when that particular patient will have their health return to a “stable” condition. “Stability” is often partly defined by doctors based on scoring systems such as the “National Early Warning Score” (NEWS), which take the patient’s physiological measurements (such as heart rate and breathing rate) and give a score which increases as the patient’s health becomes more “unstable”. Using this definition, our tool will use a wide range of data that are usually recorded for patients in hospital to predict if and when the patient will return to a low “NEWS” score. The results could help doctors, nurses and hospital teams to discharge patients from hospital sooner.