the opensafely collaborative

Researchers often analyse large volumes of patients' medical records. This data is powerful, but also challenging. It contains sensitive confidential information on everyone, and is very difficult to process. Analysts write programs to prepare and analyse the data, but traditionally they work in a closed way: nobody outside the individual research team can see the computer programs that reshaped or analysed the data. Other teams can't quality-check that work, re-use it, or learn from it. The OpenSAFELY platform was created during COVID-19. It is mostly known for inventing new methods that protect patients' privacy. This meant the OpenSAFELY team could, for the first time, give lots of researchers access to analyse the GP records of all 58 million citizens in England. The team also created new working methods that make modern, open science the automatic default. Researchers can only run analyses in OpenSAFELY by first posting their programs on a code-sharing website. All code run in the platform is logged, in public, so any misuse or "fishing expeditions" would be visible to the entire research community. Lastly, all code for data preparation is standardised: it's easy for other users to read, understand, evaluate, amend, or re-use. OpenSAFELY is now used by hundreds of researchers from 22 institutions. It has delivered critical research in major journals like Nature, Lancet, and BMJ. But by making it easy - and obligatory - to do the right thing, OpenSAFELY has also successfully transformed the culture of openness among researchers using GP patient records.