Gary Collins

RESEARCH

Gary's research interests are primarily focused on methodological aspects surrounding the development and validation multivariable prediction (prognostic) models (design and analysis) and he has published extensively in this area. He is particularly interested in the sample size considerations and the role of big data in developing and evaluating prediction models. 

Gary led an international initiative to develop guidance for studies using artificial intelligence and machine learning (TRIPOD+AI). He has also been involved in other guidance for reporting artificial intelligence/machine learning studies including CONSORT-AI/SPIRIT-AI (for reporting AI intervention studies), STARD-AI (for reporting AI based diagnostic test accuracy [DTA] studies), and DECIDE-AI (bridging the development-implementation gap). He is also involved in developed risk of bias tools for machine learning diagnostic test accuracy studies (QUADAS-AI) and prediction model studies (PROBAST+AI). Gary is also involved with colleagues from the University of Southern California developming guidance fir the responsible use of large language models such as ChatGPT for research (the CANGARU guidelines) and with colleagues from McMaster University for developing reporting guidance on studies evaluating chatbots for providing medical advice (CHART guideline).

Gary is also been involved in the development of reporting guidelines including the GATHER statement for reporting global health estimates, and published in the Lancet and PLoS Medicine, and the AGReMA statement for reporting mediation analyses, published in JAMA. More recently he is involved in updating the SPIRIT and CONSORT guidelines. Gary is also a steering group member of the international STRATOS Initiative, which aims to provide accessible and accurate guidance in the design and analysis of observational studies, and currently sits on the external advisory board for the Centre for Open Science Transparency and Openness Promotion Guidelines