As AI is rapidly being introduced into nursing practice, perceptions among frontline healthcare workers appear to vary significantly depending on the type of technology being used.
AI systems that reduce repetitive tasks -- such as logistics robots and real-time location-tracking systems (RTLS) -- received high satisfaction ratings from nurses. In contrast, clinical decision-support tools like fall-prediction AI and pressure-ulcer-detection AI struggled to gain trust in clinical settings due to errors and the added burden of verification.
The findings were highlighted in the International Labour Organization (ILO) report, “
Nursing and Artificial Intelligence in the Republic of Korea
,” featured in the National Assembly Library’s weekly report publication released on Thursday.
Published by the ILO last month, the report analyzed the real-world impact of AI on nursing labor through interviews and on-site observations at tertiary general hospitals in Seoul.
(Credit: KBR)
According to the report, the technologies that received the most positive feedback were the logistics robot “Magic Cart” and RTLS equipment-tracking systems. These technologies proved effective at reducing repetitive, time-consuming tasks, such as checking ward inventory and locating medical devices.
A nurse working in a urology ward said, “In the past, nurses had to manually check every evening whether supplies had arrived correctly and in the right quantities. Now the entire process has been systematized and handled by logistics companies, which has significantly reduced our workload.”
RTLS also reduced nurses’ workload by enabling real-time tracking of medical equipment locations. The report noted that the system effectively eliminated the long-standing practice of arriving one to two hours early to search for devices. Another nurse explained, “I used to come to work about an hour early, but thanks to this technology, that task has been greatly reduced, and now I can arrive on time.”
In contrast, evaluations of fall-prediction AI and pressure-ulcer-detection AI were generally negative. The clinical decision-support system (CDSS) for fall-risk prediction calculates risk based on factors such as patient age, medication use, mobility, and cognitive status. However, nurses noted that the system failed to adequately account for behavioral variables among patients and caregivers in actual clinical settings.
A rehabilitation ward nurse said, “We trust our own experience with patients and caregivers more than the numerical fall-risk score.” An intensive care unit nurse added, “At first, when I doubted the system’s results, I contacted the relevant department. But later, I just followed whatever the system instructed.” The report described this reaction as “closer to resignation than trust.”
Pressure-ulcer-detection AI was also found useful in some wards, but showed limited accuracy. The report explained that in wards caring for many bedridden patients, diagnostic accuracy varied depending on lighting conditions, creating additional work for nurses. Some nurses felt it was more efficient to rely on their own judgment rather than go through the process of taking photos, uploading them, and reviewing AI-generated results.
The report also highlighted structural issues within Korea’s healthcare system. While Korea has rapidly advanced hospital digitalization and AI adoption, nurses continue to face high patient loads, intense workloads, and high turnover rates. Only 52.8 percent of licensed nurses in Korea currently work in clinical settings, lower than the Organisation for Economic Co-operation and Development (OECD) average of 68.2 percent. About 12,000 nurses are estimated to leave hospitals each year.
The report emphasized that frontline participation is a critical factor for successful AI adoption. In the case of the Magic Cart system, hospital wards collaborated with IT, logistics, and purchasing departments for months through repeated consultations to reflect practical workplace needs. A digital transformation (DX) manager explained, “The longest part of the process involved continuous feedback through preliminary meetings, gathering requirements, and repeated first-, second-, third-, and fourth-round design revisions.”
The report concluded, “Designing technology based on problems identified in actual workplaces together with frontline workers is the key condition for successful automation.”
Kim Eun-young
댓글
0