Andrew Taylor, MD, MHS
A new artificial intelligence (AI) model developed to predict agitation in emergency departments could help clinicians intervene earlier and reduce the need for physical restraints or sedating medications. The study, published in JAMA Network Open, involved more than 3 million emergency visits across nine hospitals in New England and represents a major step forward in using data to improve patient safety.
Andrew Taylor, MD, MHS, vice chair for research and innovation in the Department of Emergency Medicine at the University of Virginia, served as senior author on the study. The work reflects his long-standing focus on using AI to address real-world challenges in high-acuity care settings.
“Agitation in the emergency department can escalate quickly and put both patients and healthcare workers at risk,” said Dr. Taylor. “By identifying early warning signs through AI, we can take proactive steps to support patients before a crisis occurs.”
The model uses over 50 routinely collected clinical variables including vital signs, prior ED visits, medical and psychiatric history, and reason for visit to flag patients at higher risk of agitation. With strong predictive performance the model shows promise in guiding earlier, targeted de-escalation strategies.
Agitation-related events are a growing concern in emergency care, often requiring rapid interventions that carry their own risks. By supporting more timely and informed clinical decisions, the model offers a path toward safer, more patient-centered care.
Dr. Taylor is now expanding on this work as part of UVA’s broader efforts in clinical AI and digital health innovation. “This is a great example of how we can bring rigorous, real-world AI models to the bedside not just to improve outcomes, but to do so equitably and responsibly,” stated Dr. Taylor.
Filed Under: Research