Unforeseen events and serious events (SE) occur time and again in hospitals. These include kidney failure, sepsis or the need for blood transfusions and many other events that cannot always be expected from the course of the illness, but can nevertheless result in a drastic deterioration in the patient’s state of health. The analysis and prevention of such SEs is part of quality assurance in hospitals. To date, such SEs have been investigated by qualified physicians through manual review of patient records and recommendations for action have been made for hospital processes and treatments.
In our research project, we are developing machine learning models to predict serious injury events from electronic patient data. A first step was the prediction of severe disease progression in COVID-19 patients, where we can identify high-risk patients with high accuracy as early as 24 hours after presentation to hospital. Another example is the early prediction of kidney failure, which is possible based on just a few laboratory values. In future, we will extend these models to other classes of severe damage in order to enable time-resolved, data-driven patient monitoring. This will increase patient safety and reduce the workload in hospitals.
Speaker: Maik Kschischo, Koblenz University of Applied Sciences, RheinAhrCampus, Department of Mathematics and Technology, Joseph-Rovan-Allee 2, 53424 Remagen, Germany
