Help Us Improve Code Sharing in Health Research

TL;DR: We’re building a code sharing resource for health and medical research. If you have expertise in health research, journal publishing, or in creating and sharing reproducible code, and are interested in potentially contributing to the development of the resource through online consensus feedback meetings later this year, you can express your interest here.

Sharing the code underlying our analyses is a core value of the Bennett Institute. We’ve written quite a bit about this in the past, including what code sharing entails, why we think you should do it, and some tips on doing it well. We’ve also always tried to lead by example, making all of our software and analysis code open source for our projects, including the entire OpenSAFELY platform and nearly every piece of analysis code that’s ever been run on the platform.

Despite what we’d like to see, code sharing in the health and medical sciences is basically non-existent. You don’t have to take our word for it – the most recent systematic review on the subject shows when researchers go looking for shared code, they find it less than 1% of the time.

We’ve been doing a lot of thinking over the last year about strategies to improve code sharing practice. Requiring code sharing, as The BMJ recently (and heroically) did, is certainly a major step in the right direction. But, ideally, we don’t just want to see code shared, but shared in ways that make it comprehensible, re-usable, and reproducible. Imposing a requirement on a population that doesn’t know how to meet that requirement can lead to issues. We want to see a minimum standard for shared code that allows for both interpretability and utility to the broader community, not PDFs of .do files in supplementary material.

We’ve begun to develop this idea internally at the Bennett Institute through a checklist to aid authors in taking steps to share their code aligned with some minimum best practice standards. Now, we are very excited to have been awarded an inaugural UKRI Metascience Research Grant to further develop, and pilot test, this resource.

We’re incredibly lucky to be partnering with The BMJ, as their recent code sharing requirement makes them an ideal test bed to pilot this as an intervention. Additionally, Professor Florian Naudet at the University of Rennes will bring a wealth of knowledge on data and code sharing in medical research to the project (this study is a particular favorite of ours).

Look out for more updates as the project advances but if all of this is of interest to you, and you have some expertise relevant to the project (eg, health research, journal editors, academic software developers, code sharing enthusiast), we will be hosting online expert feedback sessions to help develop the materials later this year. You can express your interest here.