Regional collaboration on the development of AI tools in healthcare

A national initiative exploring the use of AI to analyze incident reports, aiming to proactively increase patient safety and create a more learning-oriented and information-driven healthcare system.

AI-generated image (Gemini): Nurse reporting deviations.

Challenges

Today, vast amounts of valuable data in incident reports are not being fully utilized. Lessons learned and knowledge often remain siloed within the unit where an incident occurred, instead of being disseminated. Furthermore, inconsistent classification of these incidents hinders the generation of reliable statistics and a comprehensive national overview, which in turn impedes systematic, data-driven learning.

What is a patient incident?

A patient incident is an undesired event or circumstance that has occurred—or could have occurred—in connection with healthcare services, and which has resulted in, or could have resulted in, harm to a patient.

Incident management involves the systematic process of identifying, documenting, and reporting these events and risks.

All healthcare staff are obligated to report both actual harm and near-miss incidents. The purpose is to learn from these events in order to prevent similar situations from occurring in the future.

Project purpose

This is a national collaborative project between Region Skåne, Region Västmanland, Västra Götalandsregionen (VGR), Innovation Skåne, AI Sweden, and The Swedish Association of Local Authorities and Regions (SKR) that explores how artificial intelligence (AI) and large language models can be used to strengthen patient safety.

By analyzing data from healthcare incident reports, the goal is to identify risk patterns and create a more proactive approach.

The project does not aim to build a finished technical solution, but rather to develop a 'blueprint' with recommendations and optimal processes for data management that can guide Sweden's regions.

Expected outcomes

The project will result in an increased competence and capacity for collaboration between healthcare, academia, and industry.

Concrete outcomes include 1-3 proof-of-concept models for the analysis of incidents, as well as concrete guidelines for future national collaboration.

The results will form an important basis for SKR's ongoing work on a national strategy for regional data collaboration.