Making More Reliable Decisions with AI Support: New REACT Research Consortium for Critical Care Medicine

Intensive care units constantly generate large amounts of treatment data: vital signs, lab results, medical notes, and medication plans. The new REACT research consortium aims to process this data in a standardized manner and translate it into actionable recommendations. The project is led by Charité – Universitätsmedizin Berlin. It is coordinated by Dr. Elias Grünewald, a postdoctoral researcher at the Institute for Medical Informatics at Charité – Universitätsmedizin Berlin and an associate member of the ECDF.

In the future, machine learning AI models will analyze and classify routine data. Clinical responsibility will remain with the physicians. However, they should be able to recognize critical situations earlier and make even more informed decisions.

Early Warning Instead of Rigid Thresholds

Until now, routine data from an intensive care unit has usually been evaluated only on a case-by-case basis using rigid thresholds. As a result, a deterioration in a patient’s condition is sometimes not detected until it is too late. Yet early detection of circulatory instability can be decisive for the further course of the illness. The goal is to develop a system that identifies relevant warning signs much earlier than before and, at the same time, transparently explains the basis for its assessment.

Much of the most valuable clinical information is not found in numbers, but in texts such as physician’s notes, progress reports, and test results. The project therefore aims to establish an infrastructure that harnesses both structured and unstructured data equally and converts it into a uniform, standardized format. The data will be processed according to the international interoperability standard FHIR (Fast Healthcare Interoperability Resources), a globally understood language for health data.

Medicine and Medical Informatics at the Same Table

In addition, the researchers aim to develop an AI-supported decision-making tool that provides treating physicians with clear, actionable recommendations in real time. Initially, two clinically applicable applications are planned:

• a dashboard that compares upcoming antibiotic treatments with existing guidelines, thereby supporting the responsible use of antibiotics,
• a second dashboard that provides early warning of impending circulatory instability.

The basis for development and testing is one of Europe’s largest harmonized intensive care data collections, containing treatment histories from approximately 200,000 ICU stays.

Open Source for Other Hospitals

“We are creating a common technical foundation on which clinical decision support can be implemented reliably, transparently, and in a way that is transferable across locations,” emphasizes Elias Grünewald. Following a clinical pilot phase, key components of the project—from the data infrastructure to both dashboards—will be released as open-source software. This will allow other hospitals in Germany to use the solutions as well.

About the Project

REACT stands for “Routine Data Enhanced Analytics for Clinical Trajectories.” Elias Grünewald leads the junior research consortium. Dr. Mirja Mittermaier and Dr. Nils Daum at Charité are responsible for the two clinical subprojects. Dr. René Hosch and his team at IKIM in Essen are responsible for the data infrastructure. In one of the subprojects, Elias Grünewald is responsible for data management and security, among other things.

The project is funded by the Federal Ministry of Research, Technology, and Space (BMFTR) under the “Future eHealth” funding priority with approximately 2.8 million euros over three years.

Contact
Dr. Elias Grünewald
REACT Consortium Coordinator
Institute for Medical Informatics, Charité – Universitätsmedizin Berlin
Phone: +49 30 450 581 055