Methodology
This project uses real-world health data and clinical research to better understand and improve care for people with difficult-to-treat rheumatoid arthritis (D2T RA) across Europe. A key feature of our approach is the use of federated learning, a secure and privacy-preserving way of analysing health data across multiple hospitals and countries.
.
Using real-world patient data with federated learning
We work with medical data from approximately 70,000 people with rheumatoid arthritis treated in hospitals and patient registries across several European countries. These data include information routinely collected during care, such as treatments, disease activity, healthcare visits, and patient-reported outcomes over time.
Because healthcare data are sensitive and collected differently in each country and hospital, we first carefully organise and harmonise the data locally at each centre. Importantly, the data never leave the hospital or registry where they were collected. Instead of sharing patient-level data, we use federated learning, which allows analytical models to be sent to the data, rather than moving the data to a central location. Only aggregated results or model updates are shared, ensuring patient privacy and full compliance with European data protection regulations.
.
Advanced analysis and prediction
Using the federated learning infrastructure, we apply advanced data analysis and artificial intelligence methods to address three main goals:
- Identify patients whose disease is difficult to treat
- Group these patients into clinically meaningful subtypes
- Predict early which patients are at higher risk of developing difficult-to-treat disease
These analyses consider many aspects of a patient’s journey, including treatment history, response to therapy, disease activity, other health conditions, and patterns of healthcare use. The use of federated learning ensures that insights are drawn from large, diverse populations without compromising data security.
.
Developing a clinical decision-support tool
Based on these analyses, we will develop a digital decision-support tool for healthcare professionals. This tool will use the federated learning models to:
- Alert clinicians when a patient may have difficult-to-treat rheumatoid arthritis
- Support personalised treatment decisions based on patient subtypes and clinical characteristics
The tool is designed to support clinical decision-making, not replace it, and builds on existing European clinical recommendations.
.
Testing the approach in clinical practice
The decision-support tool will be evaluated in a pilot clinical study. Patients will first be observed under usual care and then followed during a phase where treatment decisions are supported by the tool.
We will assess:
- How feasible and user-friendly the tool is in daily clinical practice
- Whether it supports more tailored treatment decisions
- Its impact on patients’ quality of life and disease outcomes
.
Biological samples and biomarkers
For patients in the clinical pilot study, blood samples will be collected and stored in local secure biobanks. These samples will be analysed centrally to identify biological markers linked to difficult-to-treat disease and its different subtypes. This work may help explain why some patients do not respond well to treatment and support future improvements in prediction and care.
.
Patient involvement, ethics, and trust
Patients are actively involved throughout the project, from study design to interpretation and communication of results. All activities follow strict ethical standards, and comply with European guidance on data protection, responsible artificial intelligence, and we use trustworthy federated learning technologies. The personalized treatment strategy enabled by the decision aid will include only approved treatments, but use is tailored to the individual patients and D2T RA phenotype.
.
.
Funded by the European Union (grant agreement no. 101080243). Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or the European Health and Digital Executive Agency. Neither the European Union nor the granting authority can be held responsible for them.
The project has also received funding from the Swiss State Secretariat for Education, Research and Innovation (SERI) and from Hungary’s National Research, Development and Innovation (NRDI) Fund.