CHUV is coordinating LLM4CIRS (Large Language Model for Critical Incident Reporting Systems), a national project funded with 480,000 francs by the Federal Quality Commission, half of its total budget. Several thousand adverse events are reported every year in a university hospital: medication errors, falls, violent behaviour. The project aims to speed up their analysis and to propose corrective measures whose effectiveness has been scientifically demonstrated.
The tool will take the form of a secure, multilingual chatbot interface for the professionals in charge of quality processes. Its development is entrusted to the Biomedical Data Science Center, a structure shared by CHUV and the University of Lausanne; the evaluation will be carried out by the Institut universitaire de formation et de recherche en soins in Lausanne. The solution will be connected to the national CIRRNET network and to the PatBox.ch platform, then deployed at CHUV, University Hospital Basel and the Ente Ospedaliero Cantonale. The project is due to end in 2028.