Machine Learning for Decision Support Systems: Prediction of Clinical Deterioration

In-hospital clinical deterioration could lead to unfavorable adverse events, such as mortality, cardiac arrest, or unplanned admission to the intensive care unit. Early detection of deterioration allows clinical staff to respond in a timely manner in order to avoid such events. Advancements in data digitization and computational power enable the development of new algorithms for the prediction of clinical deterioration. Such algorithms, traditionally based on simple aggregate-based Early Warning Score (EWS) systems, seek to drive the inference engine of clinical decision support systems. They geared even more attention since the coronavirus disease 2019 pandemic to support the prognosis of in-hospital patients. The purpose of this chapter is to provide a brief overview of classical EWS systems as well as systems based on state-of-the-art machine learning and deep learning. We also compare their strengths and limitations, summarize current findings on the clinical impact of EWS systems in practice, and provide a future outlook on early warning clinical decision support systems based on current needs.