FORTRESS-TeHR: Federated, Open and Reliable TREs for Synthetic Textual Healthcare Records

Free-text clinical records contain important contextual information but are rarely made available for research because of privacy and re-identification risks. Unlike structured datasets, free text can embed personal details in subtle ways that are difficult to remove or assess consistently. This creates uncertainty for data custodians and limits research using narrative clinical data.

FORTRESS-TeHR will explore how synthetic text generated using privacy-enhancing techniques, including differential privacy, can offer a safe alternative. Using cardiology as a test case, the project will assess whether synthetic free text preserves sufficient analytical value while reducing disclosure risk. It will also examine how such data can be used in federated settings across multiple Trusted Research Environments (TREs).

The project will produce validation frameworks and governance insights to help TRE operators, regulators, and researchers understand when synthetic text is appropriate and where its limitations lie. These outputs will support more informed and proportionate decisions about future access to free-text data.

Public and patient involvement is embedded throughout the project. Working with established public and patient engagement partners, the team will establish a Public Advisory Panel and run co-design activities at key decision points. These contributions will shape expectations around privacy, acceptable use, and public benefit.

By the end of the project, FORTRESS-TeHR will:

• Prototype federated workflows for generating synthetic clinical free text
• Develop validation approaches for assessing utility and privacy risk
• Produce guidance on the appropriate use of synthetic text in TREs
• Support regulatory and public confidence in synthetic data methods

Project information

Lead organisation: University of Manchester
Principal investigator: Professor Goran Nenadić
Project duration: 12 months
Project partners: Imperial College London, MHRA, NICE, Umbizo
Funding provided: £319,099
Primary contact email: gnenadic@manchester.ac.uk