Acute kidney injury (AKI) in hospitalized patients accelerates the progression to end-stage kidney disease (ESKD), but the greater challenge lies in outpatient AKI (AKIOPT)—a largely overlooked condition due to fragmented data and lack of consensus definitions. AKI can progress to chronic kidney disease and ESKD if the kidney function is not carefully cared for. Taiwan, known as the “dialysis island”, has the world's highest ESKD incidence and prevalence, reflected in an annual health insurance expenditure of NT$58.7 billion. This underscores the urgent need for an innovative way, data-driven intervention.
To address this, Taiwan established the National Health Insurance MediCloud to unsure a cloud-based health information exchange with cross-institution interoperability. Leveraging this infrastructure, CMUH launched the AKI Detection System (AKIDS) in 2017, integrating national and local electronic health records (EHRs) to enable real-time screening of AKIOPT and nephrotoxic agents use. AKIDS functions as an in-flow clinical service, streamlining kidney care through automatic function summaries, dialysis risk alerts, and a nephrotoxic agent dashboard. It can significantly reduce the time physicians spend gathering data and enables proactive referral and scheduling for high-risk patients.
By embedding algorithms into routine care, AKIDS sets a new benchmark for digital kidney care. Its design supports cross-institutional scalability, and its effectiveness has been validated in real-world clinical trials—demonstrating the transformative potential of smart, data-driven interventions.
To solidify the seamless integration of machine-human collaboration and assess clinical effectiveness, we devised a four-stage development plan for AKIDS:
Since the sandbox stage, AKIDS has screened for over 600 thousand outpatients in the CMUH, with an average monthly AKIOPT incidence of 11%. Of patients with AKIOPT, 23% of them required nephrology care, but only 3% actually received such care before the AKIDS implementation. In addition, patients with AKIOPT have significantly 16-fold risk of kidney failure and 3-fold risk of death within one year, indicating a huge gap in kidney care (Commun Med. 2025. https://doi.org/10.1038/s43856-025-00836-4). Furthermore, our pragmatic randomized clinical trial has provided solid evidence that for high-risk patients with AKIOPT, AKIDS implementation exhibited almost 40% reduction in the risk of kidney failure in 1 year (Figure 3). For every 6 patients covered by the AKIDS, 1 patient can be prevented to develop kidney failure. We estimate that the timely diagnosis of AKIOPT could potentially save at least 31 million USD spent on dialysis every year in Taiwan.
We believe AKIDS exemplifies the future of AI-driven clinical decision support in smart hospitals—proactively identifying silent threats like outpatient AKI, closing the gap in nephrology care, and transforming real-world clinical outcomes. With strong digital infrastructure, validated clinical impact, peer-reviewed publications, and both national and international recognition, AKIDS stands as a scalable and impactful solution to improve kidney health and reduce preventable dialysis burden on a population level.
2023
2017