AID_MIND – AI-driven automation of cross-sectoral communication

When patients transition from hospital to community care, accurate and timely information exchange is essential for a seamless care process. However, current manual processes, where nurses prepare reports, are time-consuming and often interrupted. Nurses typically spend 1–2 hours per patient preparing care plans and discharge reports, which diverts time from patient-centered care. A solution is therefore needed that can streamline this process, ensure guideline adherence, and reduce workload.

Aim

The aim of the AID_MIND project is to develop and test an AI-based Agentic Retrieval-Augmented Generation (RAG) system that can automate the preparation of cross-sectoral communication documents, such as care plans and discharge letters, based on data from electronic patient records.

The system will:

• use an advanced hybrid search pipeline to retrieve relevant information from the patient record (combining semantic search, keyword search and reranking) 
• follow any specified documentation guideline for a given document type, making it adaptable to future use cases such as discharge summaries 
• include a validation agent that checks factual consistency, source adherence and formatting of the generated document 
• generate document drafts with explicit references to the underlying patient record, requiring only proofreading and final approval by nurses.

The system will be tested at the Department of Geriatric Medicine at Odense University Hospital. By integrating this technology, the project aims to free up nurse time for patient-centred care, improve documentation workflows, and create a scalable solution that can be extended to other departments and document types.

Partners

• CAI-X
• Department of Geriatric Medicine, Odense University Hospital
• Research Unit of Geriatrics, University of Southern Denmark

Funding

The project is funded by the innovation fund and co-financed by the project’s partners.