China Medical University Hospital scales AI with clinical oversight

The hospital links AI deployment to lower mortality and stronger ICU governance standards.

ICU burnout is nearing 40% as hospitals struggle with rising patient complexity, staffing pressure, and fragmented clinical systems, accelerating AI adoption in critical care operations.

China Medical University Hospital said its AI-enabled ICU systems helped “reduce the risk of mortality from drug resistance by about 40% and reduce mortality from acute respiratory distress syndrome by about 90%,” as healthcare providers seek measurable returns from digital investments.

Dr. Wei-Cheng Chen, Chief Secretary and Director of the Respiratory Intensive Care Unit at China Medical University Hospital, said AI systems must fit directly into clinical workflows instead of adding administrative pressure on healthcare workers.

“The key point of our highly centralised ICU is that we use real-time data integration and also proactively alert high-risk patients to clinicians to reduce workload and improve patient care,” Chen said.

The hospital said clinicians and engineers work together to identify operational bottlenecks and automate time-consuming processes. Chen said “all the process creation is a collaboration between the clinical teams and engineering teams,” helping hospitals deploy digital tools without increasing clinician fatigue.

The discussion also highlighted demand for evidence-backed AI systems as hospitals move beyond pilot programmes.

“For the first one, trust from published data is very important,” Chen said. “Through this evidence data, the clinical teams will have more trust to use new tools.”

Dr. Vincent Feng, Director of the Digital Transformation Technology Office at China Medical University Hospital, said fragmented data remains a major barrier inside ICUs.

“In the ICU, the problem is usually not the lack of data. The problem is the data is scattered across many different systems,” Feng said, referring to bedside monitors, laboratory systems, imaging platforms, and electronic medical records.

Feng said Fast Healthcare Interoperability Resources (FHIR) standards help consolidate information into a unified patient view, improving bedside decision-making and reducing time spent navigating disconnected systems.

The hospital also outlined oversight measures for generative AI deployment. Feng said dedicated committees review clinical feasibility, privacy, security, and ethics before AI systems are deployed.

“This helps ensure AI is used in the right way and under proper oversight,” he said.