NEW YORK, NY (April 2, 2025)--
April 2, 2025—An AI tool that analyzes nurses' data and notes detected when patients in the hospital were deteriorating nearly two days earlier than traditional methods and reduced the risk of death by over 35%, found a year-long clinical trial of more than 60,000 patients led by researchers at Columbia University.
The new AI tool, CONCERN Early Warning System , uses machine learning to analyze nursing documentation patterns to predict when a hospitalized patient is deteriorating before the change is reflected in vital signs, allowing for timely, life-saving interventions.
In the study, CONCERN shortened the average hospital stay by more than half a day and led to a 7.5% decrease in risk of sepsis. Patients monitored by CONCERN were roughly 25% more likely to be transferred to an intensive care unit compared to those who had usual care.
"Nurses are particularly skilled and experienced in detecting when something is wrong with patients under their care," said Sarah Rossetti, lead author of the study and an associate professor of biomedical informatics and nursing at Columbia University. "When we can combine that expertise with AI, we can produce real-time, actionable insights that save lives."
The findings were published today in Nature Medicine.
CONCERN Reflects Nurses' Concerns
Nurses often recognize subtle signs that a patient is deteriorating, such as pallor change or small changes in mental status. But their concerns, noted in a patient's electronic health record, may not cause immediate intervention, such as transfer to an intensive care unit.
CONCERN analyzes when nurses identify and respond to these small, but meaningful changes, by looking at nurses increased surveillance of patients, including frequency and time of assessments,, in a model that generates hourly, easy-to-read risk scores to support clinical decision-making.
"The CONCERN Early Warning System would not work without the decisions and expert opinions of nurses' data inputs," said Rossetti. "By making nurses' expert instincts visible to the entire care team, this technology ensures faster interventions, better outcomes, and ultimately, more lives saved."
Additional Information
The paper " Real-time surveillance system for patient deterioration: a pragmatic cluster-randomized 2 controlled trial of the CONCERN Early Warning System ," was published in Nature Medicine on April 2.
All authors (from Columbia unless noted): Sarah C. Rossetti, Patricia C. Dykes (Brigham and Women's Hospital), Chris Knaplund, Sandy Cho (Newton-Wellesley Hospital), Jennifer Withall, Graham Lowenthal (Brigham and Women's Hospital), David Albers, Rachel Y. Lee, Haomiao Jia, Suzanne Bakken, Min-Jeoung Kang (Brigham and Women's Hospital), Frank Y. Chang (Brigham and Women's Hospital), Li Zhou (Brigham and Women's Hospital), David W. Bates (Brigham and Women's Hospital), Temiloluwa Daramola, Fang Liu (University of Pennsylvania), Jessica Schwartz-Dillard, Mai Tran, Syed Mohtashim Abbas Bokhari, Jennifer Thate (Siena College), and Kenrick D. Cato (University of Pennsylvania).
The study was funded by grants from the National Institutes of Health (NINR 1R01NR016941 and T32NR007969).
Disclosures are noted in the paper.
