

Privacy-First Research with OpenSAFELY • Eli Holderness & Hannes Lowette
Jul 29, 2025
Eli Holderness, a Research Software Advocate known for his expertise in data science, joins Hannes Lowette, a Principal Consultant and whiskey lover, to discuss the innovative OpenSAFELY initiative. They delve into how this system allows researchers to safely access sensitive patient data through code execution in a secure environment, ensuring privacy. The conversation also covers the creation of EHRQL, a specialized query language for healthcare research, and highlights the ethical challenges of maintaining patient trust while advancing medical research.
AI Snips
Chapters
Books
Transcript
Episode notes
Privacy-First Research Approach
- OpenSAFELY prioritizes privacy by running researchers' code on sensitive data without exposing the data itself.
- Only aggregated, privacy-checked results are returned to researchers, avoiding direct data access.
Human Output Checking
- Output checking is a crucial step to ensure results do not reveal sensitive individual data.
- Trained researchers review outputs to prevent disclosive information from being released.
Enforcing Hypothesis-Driven Research
- Researchers must submit explicit hypotheses upfront to avoid p-hacking.
- OpenSAFELY enforces transparency by requiring publication of all analysis code with research papers.