Artificial intelligence programs are likely to require hospitalization for ER patients, who could help doctors and nurses make predictions as quickly as hours, a new study says.
According to a survey released on August 11th, published in Mayo Clinic Proceedings: Digital Health Journal, AI programs trained with nearly 2 million patient visits have become slightly more accurate than ER nurses.
If this approach is successful, it could help reduce overcrowding in hospital emergency departments, researchers say.
“We are pleased to announce that we are committed to providing support for our clients with a wide range of services,” said Prime Minister Jonathan Norver, vice president of nursing and emergency services at Mount Sinai Health System in New York City.
“Industry such as airlines and hotels use bookings to forecast demand and plans. With ED, there are no bookings,” he continued in a news release. “Can you imagine an airline or hotel without reserves only predictions and plans from any historical trend? Welcome to Healthcare.”
A recent study by Journal Health Affairs found that up to 35% of ER patients who require hospitalization spend more than four hours in practice known as “boarding,” a practice that waits for a spare room or bed.
Worse, previous studies have found that nearly 5% of patients wait a whole day for bed during busy winter months.
“Our goal was to see if AI could combine with opinions from nurses to speed up admission plans, some sort of reservation,” Nover said. “We have developed tools to predict admission needs before orders are placed, and we have fundamentally improved how hospitals manage patient flow, providing insights that will lead to better outcomes.”
For this project, researchers trained AI on over 1.8 million ER visits that occurred between 2019 and 2023.
“By training the algorithm with visits of over 1 million patients, we aimed to capture meaningful patterns that would help predict admissions faster than traditional methods,” Dr. Eyal Klang, chief of Generation AI at Icahn School of Medicine in Mount Sinai, said in a news release.
The team then launched AI to over 500 ER nurse executives in assessing nearly 47,000 patient visits that occurred in six emergency departments at Mount Sinai Health System in September and October 2024.
The nurse was asked to determine if the patient needed to be hospitalized after a quick triage. The researchers also fed the results of the triage to AI to see what it predicts.
The nurses proved approximately 81% accuracy in predicting which patients would require hospitalization compared to the 85% accuracy of AI.
“We've been encouraged to make sure that AI can stand on its own in making complex predictions,” said Robert Freeman, a co-researcher of Robert Freeman, head of digital transformation at Mount Sinai Health Systems. “However, equally important, this study highlights that the key role of nurses (more than 500 people in person) demonstrates how human expertise and machine learning work closely together to rethink care delivery.”
Researchers then plan to implement AI in real-time workflows and monitor how the program affects boarding times and patient flow through the ER.
“This tool is not a replacement for clinicians, it's about supporting them. Early predictions of admission allow the care team to have time to plan, coordinate and ultimately provide better, more compassionate care,” Freeman said. “It's inspiring to see AI appear not as a futuristic idea, but as a practical, real-world solution shaped by people who provide care every day.”
More information
The American College of Emergency Physicians has more to do with ER boarding and crowding.
Copyright©2025 Healthday. Unauthorized reproduction is prohibited.
