Sponsored by the NIH National Center for Advancing Translational Sciences, CRITICAL is the first cross-CTSA attempt to create a multi-site, multi-modal, de-identified dataset that has both deep-data depth and broad-data width, the combination of which is still a major unmet need.

Our dataset will also include larger quantities of longitudinal in-patient and out-patient data, both pre- and post- ICU admissions, from over 400,000 distinct critical-care patients, which makes it the largest publicly shared, disease-independent benchmarking clinical dataset yet created. The diversified racial, ethnic, and geographic profiles of the data are expected to answer urgent and long-standing clinical problems and to support fair and generalizable AI translation for advanced patient monitoring and decision support.

CRITICAL will also lend itself not only to AI / machine-learning (ML) research, but also to outcomes-related research, opening the clinical translation to broader research communities.