Tackling algorithmic bias and promoting transparency in health datasets: the STANDING Together consensus recommendations

This paper presents recommendations for the documentation and use of health datasets. The recommendations were developed through a Delphi approach, supplemented by public consultation and international interviews. Over 350 representatives from 58 countries contributed, with 194 Delphi participants from 25 countries voting on 32 draft items over three electronic survey rounds and one in-person meeting. They aim to promote transparency regarding the composition and limitations of health datasets, and aid the identification and mitigation of AI algorithmic biases. From an editorial perspective, the recommendations would be useful for editors when handling manuscripts where the authors have used a health dataset (and particularly if they’re presenting a new dataset) or else an AI study using health datasets; they could be used to encourage authors to assess for themselves how closely the datasets underlying their study adhere to the recommendations, and to identify anything that’s missing or hasn’t been considered.

View the article

Lancet Digit Health. 2025 Jan;7(1):e64-e88  doi: 10.1016/S2589-7500(24)00224-3

Recommended on behalf of EASE by Silvia Maina, Italy