Weekend urgent care centers aim to relieve overcrowding in emergency departments by diverting patients with mild conditions. AI-assisted precision triage is expected to play a key role in this effort. The Ministry of Health and Welfare (MOHW) is actively promoting AI to build a patient-centered medical network, with Taiwan’s unified health data standards described as a crucial first step.

The “2025 Smart Sustainable Cities International Forum,” hosted by the Central News Agency, kicked off yesterday. In the evening closing keynote titled “Building a Patient-Centered Medical Network, Sustainable AI-Powered Protection,” MOHW Director-General for Technology and Information, Li Chien-Chang, shared insights on Taiwan’s medical system. He noted that while Taiwan ranks first globally in healthcare accessibility and affordability, challenges like emergency department congestion persist, and AI intervention offers an opportunity to reverse the situation.

An emergency physician himself, Li Chien-Chang used ED congestion as an example. He explained that authorities hope to alleviate ED pressure by diverting patients with minor ailments to weekend urgent care centers. Triage in emergency rooms follows a standard scale from Level 1 to 5, with lower numbers indicating more severe conditions and higher treatment priority. In Taiwan, Level 4 and 5 patients account for 15% of ED visits, while Levels 1–3 account for 85%, with Level 3 patients being the largest group.

Li noted that analysis of emergency department big data, tracking patient mortality 30 days after an ED visit, revealed the highest mortality among Level 3 patients—higher even than Levels 1 and 2. This underscores the critical need for accurate triage, especially when clinicians cannot instantly access a patient’s extensive medical records, making it difficult to effectively assess life-threatening risks.

Li believes AI can address this issue in emergency triage. By combining cloud-based medical records with historical patient data, AI can improve triage accuracy. Analysis of over 100,000 samples in the U.S. shows that automated triage can reduce misclassification errors by up to 50%, accurately identifying mild versus severe cases and efficiently directing patients to appropriate medical facilities such as clinics.

Although Taiwan’s smart healthcare innovations have repeatedly won international startup awards, real-world AI adoption remains limited. Li pointed out that fragmented medical systems and data remain a major challenge. Even with electronic health records (EHRs), Taiwan’s hospitals and clinics use a wide variety of systems that cannot communicate with each other and lack uniform standards, preventing effective integration and utilization of data.

Li added that in 2025, two medical centers exchanging patient data still cannot do so electronically; CDs or printed records are required. “Data is the oil of the 21st century. Taiwan has abundant health data—from clinics and pharmacies to medical centers—but without interoperability, it is like shale oil that cannot be efficiently extracted or utilized on a large scale,” he said.

Li stressed the urgent need for a unified data platform, standardization, and improved system compatibility. The MOHW is working to establish a “data middle platform” to standardize Taiwan’s health data, integrating not only medical institutions but also official data from the Health Promotion Administration and Centers for Disease Control. This approach aims to solve data fragmentation, achieve effective integration, and enhance overall healthcare efficiency.

Li also addressed ethical concerns regarding AI in healthcare. He referenced a movie from three years ago, in which the protagonist invents an AI screening system that predicts patients likely to die during a 20-day hospital stay and labels them as receiving “futile care,” restricting their access to medical resources. Ultimately, the protagonist’s mother dies because she is deemed to receive futile care.

Li emphasized that while AI can improve efficiency and accuracy, it lacks “humanity.” In real medical settings, physicians sometimes perform interventions they know may be futile. For example, in emergency care, if efforts prove ineffective after two hours, treatment may be stopped—but when a child is involved, parents’ distress may lead the medical team to continue.

“Even after 4 to 5 hours of resuscitation, we continue for the sake of the parents,” Li said. In healthcare, this empathy is crucial—a factor AI cannot fully judge or understand. Medical decisions involve emotional, ethical, and human considerations that remain beyond the reach of current AI technology.

Resource: AI精準檢傷分級助急診分流 衛福部拚健康資料整合