Speech Recognition and AI-Based Documentation in Specialized Health Services (KI-DOK)
Background:
Healthcare professionals in specialized health services spend a significant share of their working time on clinical and administrative documentation, which reduces patient contact and contributes to increased workload. At the same time, developments in Norwegian speech-to-text technology and large language models have made it possible to go far beyond traditional dictation, with AI-based support for structuring, summarizing, and suggesting medical record entries. However, there is a lack of experience, methodology, and frameworks for how such solutions can be adopted in a safe, effective, and cost-efficient manner within specialized health services.
Objectives:
The project will map user needs and pilot both commercial and in-house developed AI-based solutions for speech recognition and documentation (KI-DOK) in selected clinical and administrative settings. Furthermore, the project will develop methodologies for evaluating time use, documentation quality, and clinical relevance, and carry out legal and technical assessments to ensure responsible use. The project also includes the development of two medical speech-to-text datasets: one open-source dataset for training Norwegian medical speech-to-text models, and one for internal validation of models that can be used in Helse Nord. The goal is to establish a solid knowledge base for later procurement, implementation, and large-scale deployment of AI-based documentation support in specialized health services.
Project partners:
The project is a collaboration between seven different clinical environments and SPKI, HN IKT, the Norwegian Centre for E-Health Research, as well as the legal and innovation teams at UNN. The project manager is Karl Øyvind Mikalsen, head of SPKI. The project is funded for 2025–2027 through innovation funding from Helse Nord.
