Information-driven care – federated learning and synthetic data generation

The project intends to build a scalable federated structure that enables information-driven care. This is done by using health data from three regions in project stage 1: Region Halland, Region Örebro län and Region Kronoberg. For stage 2, the project intends to scale up and include more regions, patient groups and data sources in a national system demonstrator to improve and optimise Swedish healthcare.


An aging population, the increase of chronic diseases and insufficient access to healthcare staff require the regions to improve the quality of care without increasing costs. Part of the solution is to make better use of health data to work more fact-based with information-driven care. Information-driven care is a concept where conclusions are drawn from health data to get an overall picture of healthcare – from an individual to a system perspective. The information can be used for early disease detection, to identify risk groups or risks in individuals and implement preventive measures with high precision. The management of the regions can save resources without compromising the quality of care for the patients, by directing resources to where the they do the most good.


The project applies and scales up:

  • AI modeling for increased precision and early action
  • Federated learning for co-use of sensitive data
  • Synthetic data generation for analyzes without risk of privacy preservation
  • Patient Encounter Costing for resource description of care chains
  • Climate footprint of care chains

The project will begin with stage 1 in the autumn of 2022. Stage 2 is expected to start in 2023 and involves further scaling nationally and internationally to include more stakeholders, data sources and analysis of additional patient groups as well as in-depth research and development of the technology.

The project's three overall objectives during stage 1:

  1. To create a scalable structure of federated learning by sharing health data from three regions: Region Halland, Region Örebro län and Region Kronoberg. The project will analyse data at population level from all three regions and as a first step develop AI-based prediction models for two specific patient groups – chronic kidney disease and heart failure.
  1. To begin the development of so-called synthetic generation of health data. Synthetic data is artificially generated information that carries no trace of personal information. Synthetic data that reflects real health data enables more companies, organisations and authorities to use data for the development of new products and solutions for health care.
  1. To initiate dialogues, both between the involved regions and with national authorities and organisations, to meet the challenges and changes that the new technology entails.

About the project

Project period, stage 1:

  • August 1, 2022, to December 31, 2022


  • Vinnova

Involved partners:

  • Halmstad University
  • Region Halland
  • Region Örebro län
  • Region Kronoberg
  • Örebro University
  • AI Sweden
  • Hallandia V
  • AstraZeneca


Health Data Centre (HDC) at Halmstad University

The project belongs to Technology Area Aware Intelligent Systems (AIS) at the Department of Intelligent Systems and Digital Design (ISDD) at the School of Information Technology (ITE).