The Vogel Foundation supports intensive care data analysis with €11,000
01/26/2026The Vogel Foundation Dr. Eckernkamp is providing €11,000 in funding for the project submitted by Professor Werner and PD Dr. med. Dirk Weismann, Medical Clinic and Polyclinic I of the University Medical Center Center for Internal Medicine (ZIM), on the topic of “Analysis of data from intensive care patients.” This project aims to apply more complex mathematical methods and models, e.g., from time series analysis or deep learning, in order to make data-based predictions about a patient's stability, for example.
Randomized clinical trials are the medical gold standard for independent and unbiased assessment of interventions. The limited increase in evidence provided by intensive care studies can be explained by patient heterogeneity, the generally overly optimistic estimation of effect size, the excessive focus on mortality as an endpoint, and the dichotomization of results into significant and non-significant. The heterogeneity observed in studies could also be an expression of an insufficient understanding of critical disease progression, which makes appropriate patient selection difficult or impossible.
At the same time, large amounts of data are collected from intensive care patients in internal emergency and intensive care medicine. These usually include measurements of heart rate, blood pressure, body temperature, medication administered, fluid balance, and various laboratory parameters. Thanks to digital measurement methods, some of these measurements are available in high temporal resolution, in some cases with measurements taken at intervals of 30 seconds. It seems obvious that this data contains information about the patient's constitution and state of health that goes beyond purely intensive care use. Thus, by analyzing this data appropriately, possible treatment recommendations for the future can be derived that have not yet been systematically investigated in this form (intensive care therapy is predominantly off-label and not evidence-based). Ideally, it would even be possible to make data-based predictions about a patient's stability (i.e., when they should be transferred back to the normal ward and when not).
From a mathematical perspective, the data mentioned above represent different time series for each intensive care patient, each corresponding to one of the measurement parameters. We base our hypothesis on the assumption that this data can be understood as autoregressive with a moving average, which raises the question of regularities that could also be used for randomization, for example.
In this project, the first approach is to analyze the existing data using classical methods of statistics and data science such as clustering, UMAP, and PCA. In a second step, i.e., once a basic understanding has been established, more complex mathematical methods and models, e.g., from time series analysis or deep learning, will also be applied.
