Unlike intensive care units (ICUs), general hospital wards do not feature continuous, around-the-clock patient monitoring. When an inpatient’s condition suddenly deteriorates, early detection can be critically delayed—especially at secondary hospitals that lack dedicated emergency response teams. A new study suggests that an artificial intelligence (AI)-based medical device designed to predict cardiac arrest risk can effectively step in to fill that care gap.
Medical AI pioneer VUNO announced on June 18 that the results of a comprehensive multicenter study analyzing the clinical efficacy of its proprietary device, VUNO Med–DeepCARS, have been published in the peer-reviewed international journal Diagnostics.
DeepCARS utilizes advanced machine learning to analyze an inpatient's core vital signs—including blood pressure, pulse, respiratory rate, and body temperature—to alert clinicians to individuals at high risk of experiencing a cardiac arrest within a 24-hour window.
The study evaluated clinical outcomes before and after the deployment of DeepCARS among approximately 160,000 adult patients (aged 19 and older) admitted to general wards across three secondary hospitals that do not operate a traditional Rapid Response System (RRS): Gangdong Sacred Heart Hospital, Siwha Hospital, and Incheon Naeun Hospital. An RRS serves as a vital patient-safety infrastructure to catch deteriorating general-ward patients early, but secondary medical institutions frequently struggle to implement these programs due to high operational costs and chronic staffing constraints.
Substantial Clinical Impact Across Secondary Hospitals
The survival rate for a patient experiencing an in-hospital cardiac arrest is historically low, hovering around 13.4% at the one-year mark. Following the integration of DeepCARS, however, the participating hospitals saw a 21% reduction in the rate of in-hospital cardiac arrests and a 15% drop in overall in-hospital mortality.
Furthermore, the efficiency of hospital resource utilization improved: the average total length of stay decreased by 0.51 days, and the average ICU length of stay fell by 1.32 days. The clinical benefit was even more pronounced among patients suffering from sepsis—a notorious precursor to cardiovascular collapse—where cardiac arrests plummeted by 29% and mortality dropped by 22%.
"This study marks the second major clinical validation demonstrating that DeepCARS reduces in-hospital cardiac arrests on general wards, and it stands as our first published multicenter trial," said Joo Sung-hoon, CTO of VUNO. "The findings confirm that DeepCARS offers a highly scalable, cost-effective patient-safety solution for secondary hospitals facing prohibitive financial and staffing barriers to establishing a traditional RRS."
The multicenter findings reinforce a previous single-center study conducted at Inha University Hospital in January, which reported a 46% reduction in cardiac arrests and a 35% decline in mortality. While the earlier trial focused on a single academic center, this latest paper confirms that the technology's utility translates seamlessly across diverse secondary hospital environments.
To date, DeepCARS has been deployed at approximately 170 medical institutions nationwide, including active pilot sites, encompassing roughly 65,000 hospital beds. In South Korea, the predictive AI technology is currently expanding its market footprint primarily within the non-reimbursed self-pay sector.
