An integrated AI, telemedicine, and Internet of Things system for the ICU has brought down the overall mortality rate in critical care at China University Medical Hospital in Taiwan.
At HIMSS25 APAC, Dr Wen-Sheng Feng, director of the Digital Transformation Technology Office at CMUH, presented how they started building a scalable AI-powered system in the ICU and how it works.
Previously, their clinicians had to switch between multiple data platforms to piece together a complete view of the patient's condition.
"This fragmentation increases the risk of errors and delays, overwhelming clinical teams and making it harder to respond to clinical issues in time. It highlights the urgent need for a smarter solution," Dr Feng said.
CMUH's HiThings Tele-ICU system collects and analyses live data from thousands of medical devices and features AI modules that support clinical decision-making. It also supports seamless teleconsultation and remote monitoring, which can be done from various devices.
According to Dr Feng, the integrated ICU system has two main user interfaces: the overview panel, which presents up-to-date ICU data, and the digital twin panel, which creates a patient-specific model to offer a detailed, near real-time view of a patient's condition.
The overview panel can help clinical teams identify high-risk patients who require urgent attention. A key feature of this panel is the infection control map, which helps staff identify and quickly respond to potential outbreaks using colour-coded indicators. Another feature, the equipment tracking map, monitors the location and status of ICU devices in near real-time, helping reduce over 60% of nurses' inventory time during their handovers.
Meanwhile, the digital panel has 3D human modelling, which provides a detailed view of patients' images. It also features a life support device dashboard that can offer treatment recommendations.
The tele-ICU system runs six AI applications that each address specific critical conditions and "proactively and continuously analyse up-to-date data," noted Dr Feng.
It has AI for detecting sepsis, STEMI (severe heart attack), acute respiratory distress syndrome (ARDS), and stroke. There is also AI that provides nutrition recommendations and a generative AI-based speech-to-text tool that helps with medical and nursing documentation.
Dr Feng shared that the integrated system has helped reduce the overall ICU mortality rate by 4%. Besides helping lower the deaths caused by MRSA infections and heart attacks, the system has also decreased the instances of multi-drug resistance and delirium and increased the nutrition target achievement rate.
In the workflow side, the ICU system has led to a shorter diagnosis time for stroke based on brain CT scans and ARDS, as well as reduced the time spent on gathering data, documentation, inventory and nutrition evaluation.
HiThings Tele-ICU has been implemented in six of CMUH's ICUs, customised for each unit, which made it "practical and adaptable for [daily] use," said Dr Feng.
