The NHS Data Matryoshka: Progressive Synthetic OMOP Data Layers for Secure Health Research Collaboration Across UK TREs

Context
NHS hospitals hold extraordinary health data that could accelerate medical breakthroughs and improve patient care. Yet accessing this data remains painfully slow. Researchers wait months for approvals, then work within secure systems that constrain collaboration. More importantly, patients and the public have little visibility of the health information held within these enclaves.
Synthetic data offers a solution: artificial datasets that mimic real patient data without containing actual patient information. But Current tools fall short. They are either opaque "black boxes" that lack the transparency NHS information governance teams need to assess risks, or oversimplified systems that cannot handle the complexity of electronic health records.
NHS information governance teams need transparent systems. NHS analysts need tools designed for real-world healthcare data. Both need practical pathways that work within existing NHS structures.
Our Solution: Progressive Data Layers
Like a Russian nesting doll (Matryoshka), we propose progressive layers of health data using OMOP, the interoperable format being adopted across UK health systems.


Public Synthetic Data: Created transparently from summary statistics using differential privacy. Straightforward approval because generation methods are fully visible. Valuable for understanding data structures, testing methods, and education with no access controls needed.


Private Synthetic Data: Generated using complex algorithms that learn from individual patient records whilst creating entirely artificial data. Less transparent generation but maintains privacy budgets. Provides high-fidelity data for serious research with lighter access controls than secure systems.


Real Data: Actual patient information within existing secure controlled environments, reserved for final research stages after full approvals.


Aims and Objectives
In partnership with patients and public, NHS information governance and data teams,...