AI-based assistance systems for in-process health applications
Thesis (BA/MA/IDP)
Advisor(s): Cornelius Born, Dr. Andreas Hein
Context
Despite many efforts, including legal ones, the structural problems of emergency care in German hospitals have been only imperceptibly reduced in recent years. The existing problems of the lack of networking between the players involved in outpatient emergency care, the high proportion of minor cases, and the continuing increase in the number of patients in the central emergency rooms (CER) have not been solved. At the same time, providing the necessary resources and ensuring functioning processes in this critical area, where patients with life-threatening conditions arrive just as unplanned as patients who only require outpatient clarification or treatment, is of utmost relevance. The greatest risks in the emergency department are not recognizing patients with urgent treatment needs in time or not being able to provide the necessary resources (staff, medical technology, treatment places) in time (Erkamp et al., 2021). In order to be able to counteract these risks preventively and at an early stage, measures according to the CER process chain are suitable. This includes both the prediction of crowding situations and an optimization of the targeted and timely management of patients, resources and staff. A tense situation with regard to personnel resources in the healthcare sector only permits a solution with AI in the long term.
One solution is the use of an interactive, AI-based assistance system. Such a system could predict crowding situations at an early stage and, based on this, initiate semi-automated measures to improve or avoid the situation. For this purpose, internal data sources (e.g., actual utilization, duty rosters, bed occupancy, availability of rooms and equipment), external data sources (e.g., weather, events, ambulance service/emergency physician, etc.), as well as interfaces to downstream IT systems can be used. In the research project ZNAflow we therefore aim to develop and evaluate an interactive, AI-based assistance system for data-based (partial) control of patient flows in the emergency department in a user-centric manner. The project therefore allows innovations in the field of interactive AI-based assistance systems to improve clinical, organizational or administrative processes in hospitals or treatment centers.
In CERs, the first AI-based systems for initial assessment of patients are already assisting medical staff in triaging (Moulik et al., 2020). Retrospective studies show the potential of machine learning (ML) based on historical data to predict patient flows prior to admission (Ehrlich et al., 2021; Hilton et al., 2020, Yao et al., 2020). Appropriate models have been able to identify associations between patient volumes in CERs with weather and events (Erkamp et al. 2021). However, these approaches are limited in terms of parameter selection and available data sets (Hilton et al., 2020) and use only historical data. Initial projects on the use of AI systems in the CER show the basic feasibility, for example, in the Hospital for Sick Children, Toronto, Canada (Varner, 2020). The concrete implementation and software modules of this project are not available and the transferability unclear, which is why an exchange with the project is being sought.
From a social and ethical perspective, socio-technical systems, as planned in ZNAflow, are subject to high complexity, as humans make decisions based on machine recommendations. Discrimination of the system and liability issues are key challenges. In particular, the quality of decisions made by machines is directly dependent on representativeness, completeness, and correctness of training data. Code of Conducts and similar guidelines are not sufficient to incorporate ethically relevant issues (Gogoll et al., 2021). Therefore, a dedicated and case-specific approach is also necessary in ZNAflow.
Potential research topics in this context are:
- Requirements analysis for applications, processes, and data used in CERs
- MVP development of the CER application and/or MVP development of a AI prescription model
- Integrating ethical, social, and legal requirements for AI applications in healthcare
- Analysis of potential business models and transfer strategies from research to practive for the ZNAflow application
Task(s) - depending on specific topic
- Review relevant literature in the respective field and on similar phenomena in related research fields
- Conduct interviews, workshops, or surveys with relevant stakeholders, with suitable data analysis, e.g., grounded theory
- Software (e.g., UX, AI/ML models) design, development, and validation
- Derivation of practical recommendations for the implementation of AI applications in healthcare
Requirements
- Interest in research on digital innovation, digital health and artificial intelligence
- A high degree of autonomy and individual responsibility
- Good communication skills e.g., to conduct interviews
- Strong analytical skills
- Structured, reliable, and self-motivated work style
- Experience in design science is benefitial
Further Information
The thesis should be written in English. If you have any further questions, do not hesitate to contact me directly.
Please send your application including our application form, "Notenauszug" from TUMonline, and CV to cornelius.born@tum.de . Please note that we can only consider applications with complete documents.
References
Ehrlich, H., McKenney, M., & Elkbuli, A. (2020). The niche of artificial intelligence in trauma and emergency medicine. The American Journal of Emergency Medicine, pp. 669-670. https://doi.org/https://doi.org/10.1016/j.ajem.2020.10.050
Erkamp, N. S., van Dalen, D. H., & de Vries, E. (2021, 2021/06/12). Predicting emergency department visits in a large teaching hospital. International Journal of Emergency Medicine, 14(1), 34. https://doi.org/10.1186/s12245-021-00357-6
Geissler, A., Wilm, Q., Busse, R. (2017) Umgestaltung der Notfallversorgung: Internationale Erfahrungen und Potenziale für Deutschland. Krankenhaus-Report 2017, Schattauer 2017. Stuttgart
Gogoll, J., Zuber, N., Kacianka, S., Greger, T., Pretschner, A., & Nida-Rümelin, J. (2021). Ethics in the Software Development Process: from Codes of Conduct to Ethical Deliberation. Philosophy & Technology. https://doi.org/10.1007/s13347-021-00451-w
Haas, P., & Schneider, U. K. (2021). Rahmenbedingungen Cloud-basierter Krankenhausinformationssysteme. https://kh-digitalisierung.de/files/downloads/Haas_Schneider_Cloud-KIS-Gutachten.pdf
Hevner, A. R., March, S. T., Park, J., & Ram, S. (2004). Design Science in Information Systems Research. MIS Quarterly, 28(1), 75-105. https://doi.org/10.2307/25148625
Hilton, C. B., Milinovich, A., Felix, C., Vakharia, N., Crone, T., Donovan, C., Proctor, A., & Nazha, A. (2020). Personalized predictions of patient outcomes during and after hospitalization using artificial intelligence. npj Digital Medicine, 3(1), 51. https://doi.org/10.1038/s41746-020-0249-z
Moulik, S. K., Kotter, N., & Fishman, E. K. (2020). Applications of artificial intelligence in the emergency department. Emergency Radiology, 27(4), 355-358. https://doi.org/10.1007/s10140-020-01794-1
Peffers, K., Tuunanen, T., Rothenberger, M. A., & Chatterjee, S. (2007). A Design Science Research Methodology for Information Systems Research. Journal of Management Information Systems, 24(3), 45-77. https://doi.org/10.2753/MIS0742-1222240302
Pin, M., Dodt, C., Somasundaram, R., Gräff, I., Dormann, H., Dietz-Wittstock, M., Wrede, C.E. (2018) Deutsche Gesellschaft Interdisziplinäre Notfall- und Akutmedizin (DGINA): Notfall + Rettungsmedizin. (21) 492–495
Varner, C. (2020, March 12). Can artificial intelligence predict emergency department crowding? Healthy Debate. Retrieved September 28, 2020 from https://healthydebate.ca/2020/03/topic/predicting-emergency-department-crowding-artificial-intelligence-mar2020/
Yao, L. H., Leung, K. C., Hong, J. H., Tsai, C. L., & Fu, L. C. (2020, 11-14 Oct. 2020). A System for Predicting Hospital Admission at Emergency Department Based on Electronic Health Record Using Convolution Neural Network. 2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC)
Zdun, U., Bener, A., & Olalia-Carin, E. L. (2012). Guest Editors' Introduction: Software Engineering for Compliance. IEEE Software, 29(3), 24-27.