AI Predicts Acute Kidney Injury, Supporting the Critical First 24 Hours in the ICU

(Central News Agency reporter Chen Chieh-ling, Taipei, 7th) Taiwan has long been one of the countries with a high prevalence of kidney disease worldwide. With the rise of artificial intelligence, AI-based acute kidney injury (AKI) prediction support software has become a growing trend. It aims to predict the probability of ICU patients developing AKI within the next 24 hours, enabling early intervention for high-risk patients to prevent the need for dialysis.

Care for chronic kidney disease and acute kidney injury (AKI) is a key policy focus of Taiwan’s National Health Insurance system. According to data from the National Health Insurance Administration under the Ministry of Health and Welfare, medical expenditures related to kidney diseases, including dialysis, have consistently ranked first among Taiwan’s annual healthcare spending categories, highlighting the importance of kidney care and its long-term burden on the healthcare system.

In recent years, the National Health Insurance Administration has continued to promote policies such as early-stage kidney disease care programs, AKI care management, and digital healthcare applications. Lee Hui-Ling, General Manager of Huede Healthtech, told CNA reporters today that AI-based acute kidney injury prediction support software has become one of the key representative projects selected in this year’s Taiwan 50 Outstanding SMART Apps competition.

Lee Hui-Ling explained that acute kidney injury is difficult to detect early. This AI-based AKI prediction support software is designed for use in intensive care units, providing clinical data analysis and decision support. It is applicable to ICU patients aged 20 and above, and uses machine learning algorithms to provide healthcare professionals with a predicted probability of a patient developing AKI within the next 24 hours.

Lee Hui-Ling noted that with the assistance of AI, physicians can make rapid clinical judgments. For patients identified as extremely high risk, nephrologists can provide personalized treatment to prevent progression to acute kidney injury or even the need for dialysis. The system has an area under the ROC curve (AUROC) of 0.86, demonstrating excellent predictive performance. It has obtained a Class II medical device license from the Taiwan Food and Drug Administration, and has already been adopted by several medical centers to enhance patient care quality.

SMART on FHIR has become an important international standard framework for next-generation smart healthcare. The Ministry of Health and Welfare organized the first “Taiwan 50 Outstanding SMART Apps” competition, selecting 50 representative innovative applications from 115 submissions. Among them, the AI-based AKI prediction support software was chosen as one of the key representatives advancing the standardization of smart healthcare and clinical AI applications in Taiwan.

Lee Chien-chang, Director of the Department of Information Management at the Ministry of Health and Welfare, previously stated in a press release that under the SMART on FHIR framework, future healthcare applications can be downloaded and used like smartphone apps across different healthcare information systems with “plug-and-play” functionality. This not only lowers adoption barriers but also enables high-quality applications to rapidly spread across medical institutions at all levels, reducing redundant testing and improving clinical efficiency.

Other awardees in the Taiwan 50 selection include the AI chest X-ray abnormality screening system from Tri-Service General Hospital, a post-cardiac catheterization bleeding risk assessment tool from Linkou Chang Gung Memorial Hospital, an AI imaging interpretation integration platform from National Taiwan University Hospital, and an intelligent emergency department admission risk prediction system for non-trauma patients from Chung Shan Medical University Hospital. (Editor: Kuan Chung-wei) 1150607