AI-based assistance systems for optimized control of the central emergency room
The Emergency Room (ZNA) is the point of contact in hospitals for the acute treatment of emergencies. Many different internal (e.g., staff absence due to illness) and external (e.g., increase in patient numbers due to extreme weather conditions) factors influence the capacities within an emergency department. For the continuous intactness of the ZNA, appropriate resources must be planned and made available in the right place on time. Currently, this planning is done primarily by humans based on experience. Due to the large number and the high rate of change of the existing influencing factors, it is usually impossible to include them in precise planning.
To prevent overload within the ZNA, the use of an intelligent assistance system to control processes in the ZNA is to be developed. The AI-enabled system will enable staff to recognize impending overloads at an early stage and initiate targeted measures semi-automatedly - for example, changes in duty scheduling, short-term requests for specialist staff, adjustments in bed management, timely transfer of patients, or keeping diagnostic capacities free.
To enable AI-supported planning, possible data sources will be analyzed and used (e.g., public health service data, weather data, data on upcoming major events, and internal historical and current data from the ZNA). At the same time, potential bottlenecks and restrictions within the ZNA will be included in the design of such a system. Furthermore, the challenges of using such a system will be investigated in terms of ethical, legal and social implications (ELSI). Within the scope of the project, a demonstrator will be developed, taking into account the collected findings. This will be evaluated in a field study directly in the running operation of a ZNA.
ZNAflow tests the use of an AI-based assistance system to optimize the processes in the ZNA in the hospital. In this context, fortiss supports continuous requirements elicitation and creates a domain-specific artifact model for the requirements elicitation of software-based solutions in hospitals. Furthermore, our scientists develop a prototype of a prediction model. For this purpose, various data sources and types are examined for their suitability, and a machine learning-based prediction model is developed in iterative steps. To evaluate the use of an AI-based system, ethical, legal, and social implications (ELSI) will be investigated. These findings will be processed and published as a guideline for integrating ethics and standards in AI engineering in the health sector.
01.11.2018 – 31.08.2022