Applying Advanced Data Privacy Methods to Real World Data (RWD)

Project Scope 

The goal of this project is to develop a comprehensive, cross-disciplinary resource to support individuals and organisations in applying advanced privacy-preserving techniques to real-world data (RWD). Given the evolving landscape of privacy regulations and the increasing complexity of data sources – including integrating machine learning (ML), artificial intelligence (AI) and large language models (LLMs) into healthcare pipelines – this project will serve as a critical foundation for advancing privacy-preserving data integration frameworks for RWD.

Integrating and sharing real-world data (RWD) introduces profound privacy risks due to its high dimensionality, subject uniqueness and sensitivity. While privacy-preserving technologies such as differential privacy, federated learning and secure harmonisation have emerged, there remains a lack of consolidated guidance that addresses practical implementation strategies across these techniques, especially in the context of ML, AI and LLM applications.

By addressing gaps in current guidance – including practical implementation of differential privacy, federated learning and synthetic data generation –this initiative will empower researchers, developers and regulators to adopt privacy-preserving AI in ways that are compliant, scalable and scientifically robust. This project will impact on the industry by delivering a foundational, actionable resource focused on applying privacy-preserving methods to real-world data (RWD), particularly through the lens of ML, AI and LLM. Open-source databases (e.g. MIMIC-IV, UK Biobank, ADNI) are ideal for prototyping and validating these methods due to their reproducibility and real-world relevance.

Abhishek Mishra, GSK

abhishek.2.mishra@gsk.com

Elena Valkanova, Independent

evalkanova75@yahoo.com

Nicola Newton, PHUSE Project Coordinator

nicky@phuse.global 

White Paper/Guideline 

TBC