How can critical changes in the condition of intensive care patients be detected at an early stage and assessed in a clinically meaningful and transparent way? This question is addressed in a recent scientific paper published in Communications Medicine, a peer-reviewed journal within the Nature portfolio. Markward Britsch, Data Scientist at HMS, contributed to the work within an interdisciplinary research team.
In the project, the team developed an AI-based system that provides daily updated predictions of clinical deterioration over a 48-hour time horizon. The data foundation is robust, comprising nearly 10,000 real-world datasets from electronic health records of intensive care units, including vital signs, laboratory values, and medication information.
At the core of the approach is the combination of dynamic modeling and interpretability. The model continuously adapts to the current condition of each patient while transparently indicating which factors influence the respective prognosis. This results in data-driven decision support for physicians, helping to substantiate medical assessments and support resource planning in intensive care units.
The paper was authored by an interdisciplinary team including Simone Britsch, Markward Britsch, Simon Lindner, Leonie Hahn, Verena Schneider-Lindner, Thomas Helbing, Manfred Thiel, Daniel Duerschmied, and Tobias Becher.
Within the team, Markward Britsch was primarily responsible for data preparation and modeling. His work ranged from the structured extraction of complex ICU data to the development of the machine learning algorithm. His long-standing experience as a Data Scientist at HMS was complemented by close technical exchange with HMS experts on specialized topics.
The publication in a Nature journal highlights the scientific rigor of the project. At the same time, it demonstrates how HMS’s expertise in data science and AI extends beyond classical project contexts and is responsibly transferred into highly regulated clinical research environments.
Paper title: An interpretable machine learning algorithm enables dynamic 48-hour mortality prediction during an ICU stay
The full paper is available here.
