The MI-CLAIM checklist: Minimum information about clinical artificial intelligence modeling

September 25, 2020

BIDS Faculty Affiliate Bin Yu is part of a team that developed the mI-CLAIm checklist, a tool intended to improve transparent reporting of AI algorithms in medicine. The team is led by Beau Norgeot and Atul Butte from the UCSF Bakar Computational Health Sciences Institute (BCHSI), and also includes Milena Gianfrancesco from the UCSF School of Medicine, Rima Arnout from the UCSF BCHSI, Ziad Obermeyer from UC Berkeley’s School of Public Health, and other collaborators from Scripps Research Translational Institute, Harvard Medical School, Johns Hopkins University, and Bayesian Health.

The checklist integrates newer machine-learning and artificial intelligence methods -- which have clear advantages in performance, adaptability, and scalability -- with older ‘supervised machine learning’ models that are more interpretable by expert medical professionals. 

More effective modelling becomes crucial in clinical testing, as well as the medical products and services that are already being approved by the US FDA, especially as researchers attempt to generalize and make recommendations across larger populations. 

By providing the first steps toward a minimum set of guidelines for better documentation about test-cohort selection, development methodology, and validation systems, the tool can be used to inform readers and users about the machine-learning models being used in a given study, as well as how they were developed and tested. 

The tool is targeted for a variety of audiences -- including medical-algorithm designers, repository managers, manuscript writers and readers, journal editors, and model users -- and the long-term goal of the project is to develop a robust documentation standard that can be used by clinical scientists, data scientists, and the clinicians of the future who will be relying on these tools to implement effective diagnoses and treatments. 

To that end, the research team invites the user community to provide their feedback and suggestions on how the mI-CLAIm checklist can be improved, and a public Github repository ( has been set up to allow the community to comment on existing sections and suggest additions.

Minimum information about clinical artificial intelligence modeling: the MI-CLAIM checklist
September 9, 2020  |   Nature Medicine
Beau Norgeot, Giorgio Quer, Brett K. Beaulieu-Jones, Ali Torkamani, Raquel Dias, Milena Gianfrancesco, Rima Arnaout, Isaac S. Kohane, Suchi Saria, Eric Topol, Ziad Obermeyer, Bin Yu and Atul J. Butte

Featured Fellows

Bin Yu

Statistics, UC Berkeley
Faculty Affiliate