AI&Optimization

Intelligent planning of ICU resources

The challenge

The management of intensive care units has a significant influence on the healing process of patients. However, the capacities of intensive care units are very limited due to the high care and staffing requirements. Unforeseeable emergencies that have to be treated alongside patients with planned operations and an uncertain length of stay complicate the management of intensive care capacities. The coordinators responsible for an intensive care unit must establish a dynamic strategy for the admission and transfer of patients under constantly changing conditions in order to provide intensive medical treatment to as many critically ill patients as possible.

How we will help

In the KISIK project, XITASO is working with partners from Augsburg University Hospital and the University of Augsburg on how artificial intelligence and, in particular, predictions about the length of stay and the probability of readmission of patients can facilitate the decisions of intensive care coordinators. We embed the AI predictions in a decision support system, which we then examine for applicability, acceptance and effectiveness in hospital operations. In doing so, we pay particular attention to the explainability of the forecasts (XAI), their visualization and the user-friendliness of the overall system. In this context, XITASO contributes its proven expertise in innovative software engineering with agile development methods, particularly in the healthcare sector.

Technologien

AI-based Assistance System Machine Learning Optimization Data Analytics Usability optimization Decision Support Systems

The research project in detail

The aim of the KISIK project is to support ICU coordinators in their complex decisions in everyday clinical practice and thus ensure optimal patient care. This is intended to optimize the utilization of the ICU and reduce the number of operations postponed at short notice and premature transfers.

To this end, XITASO and the project partners are developing AI-based algorithms to support decision-making in the management of intensive care capacities, for example to predict the need for intensive care, the length of stay and the probability of readmission.

XITASO is particularly interested in the explainability of the system: doctors need transparency in how the system arrived at its predictions in order to create trust on the one hand and to be able to contribute their own medical expertise to the decision-making process on the other. We are therefore working in particular on metrics that we can use to explain the predictions of the AI models (Explainable AI, XAI) and on the visualization of these explanations. The latter pose a particular challenge, as they have to be understood quickly and intuitively by users in the stressful day-to-day hospital environment. Explainability is made more difficult by the fact that the prognoses are often based on complex correlations between the patient’s vital data over a certain period of time. Accordingly, we are researching how the explainability of prognoses can be ensured under these circumstances and are thus making an important contribution to establishing AI-based decision support in everyday clinical practice.

The KISIK demonstrator

Doctors in the intensive care unit must be able to see relevant information at a glance. This includes the diagnosis, age and length of the previous stay, as well as the predictions developed in the research project regarding the remaining length of stay and the probability of readmission.

Our first draft of the KISIK demonstrator shows the possibilities of the user interface on a tablet.

Based on ideation and usability workshops on the requirements of ICU employees, we have designed a demonstrator with a wide range of functions and options for individual customization.

Project participants and funding

Project partners

University Augsburg
Research and teaching at the University of Augsburg (Prof. Brunner and Prof. Schiffels) focuses on the planning and analysis of strategic and operational service processes in the healthcare sector, particularly in hospitals. The focus is on modelling, analyzing and optimizing practical problems using quantitative methods. Central questions deal with process, resource, quality and information management. The University of Augsburg’s role in the project is to contribute its own expertise in the areas of machine learning and optimization.

Augsburg University Hospital
Augsburg University Hospital is the only maximum care hospital in Bavarian Swabia and one of the largest hospitals in Germany with around 1750 beds. Approximately 2500 employees work there in the field of nursing care, which has to be planned and managed. The UKA is a practice partner in the project and thus establishes the link to the clinical setting. The task of the UKA in the project is in particular to communicate the framework conditions and the requirements regarding nursing staff planning at the UKA.

Project sponsor

Federal Ministry of Education and Research

Funding program

AI-based assistance systems for in-process healthcare applications

Project management

VDI/VDE Innovation + Technik GmbH

Interested in helping with our research projects?

With research and innovation projects, we explore the potential of tomorrow’s technologies. For this purpose, we are always looking for committed colleagues who would like to continue and shape the XITASO path with us.

Show all job offers