Patients with early-stage kidney injuries typically receive very similar treatments, but a new artificial intelligence study shows that distinct groups within this population may have varying needs and could benefit from more personalized treatment methods. By examining the progress of more than 50,000 patients with stage 1 acute kidney injury, researchers identified three distinct groups that could alert physicians to their needs.
“I think it’s promising, because we saw a strong signal here. This is the first scientific paper that breaks down stage 1 acute kidney injury into higher resolution,” said first author Deyi Li, a doctoral student in the College of Medicine’s Department of Health Outcomes and Biomedical Informatics, mentored by Dr. Mei Liu. Li also hopes that future studies will confirm this signal and lead to guidance from medical associations.
Acute kidney injury (AKI), formerly known as acute renal failure, develops rapidly and can range in severity from temporary symptoms to life-threatening conditions that may require dialysis. It often occurs in intensive care units, with men and the elderly being particularly at risk. Fortunately, AKI can be detected by standard blood and urine tests.
The new study was published in July 2025 in Communications Medicine, a high-impact journal affiliated with the journal Nature.
Looking for Trends in Big Data
To study how AKI patients could be identified more precisely, researchers accessed massive data sets of patients who were hospitalized between 2009 and 2022. A total population of 107,130 people were divided between AKI and non-AKI patients, with the latter serving as a control group. The anonymous patient data came from university-based hospitals affiliated with the Patient-Centered Outcomes Research Institute.
All AKI patients had been tested during the week before and the week after the onset of AKI for the biomarker creatine, a waste product in urine that can also be detected in blood tests. By focusing on the week before the onset of AKI, researchers found statically significant patterns from lower to higher levels of creatine that became the three groupings of A, B and C.
Of the three AKI subtypes identified, type C was the most likely to die within one year. They also tended to be older and have more obesity and other diseases. Type B was the most likely group to develop long-term kidney disease after leaving the hospital. While Type A patients showed lower risks and lower creatine levels, they were still more likely to die within one year than non-AKI patients.
Future Benefits
While the results of this study are not yet ready for direct application to dividual patient care, they could inform future clinical guidelines. Physicians treating patients with acute kidney injury could use these new categories to quickly identify patients who may require more intense and personalized care.
Informed patients may want to know their categorization and how it may predict their health outcomes. However, it is important to note that these categories are based on large numbers of patients, and individual experiences may differ.
A great benefit of recognizing these trends during stage 1 of acute kidney injury is the potential to prevent progression to more advanced stages. AKI patients require close, lifelong monitoring of their kidney functions post-recovery.
Team Effort
This study’s team of 16 is led by the corresponding author Mei Liu, Ph.D., an AI specialist in machine learning and an associate professor in the Department of Health Outcomes and Biomedical Informatics. The team also included HOBI post-doctoral researcher Ho Yin Chan as well as data scientists and experts in nephrology from collaborating institutions.
“I was pleased to see how well this team came together, showcasing impressive collaboration and dedication to our research,” Liu said. “Each member brought unique expertise that significantly enriched our findings and fostered a productive environment. I am particularly proud of Deyi Li, who exemplified the commitment and innovation that drive our research forward.”
By finding patterns across a large population, the researchers showed that creatine biomarkers could reveal different groupings of patients experiencing stage 1 acute kidney injury. Because AKI develops rapidly, often within a matter of days or even hours, this new information could be crucial for optimizing timely treatment.
The researchers acknowledge that future studies could refine AKI patient classifications even further. Some data, particularly regarding drugs known to be toxic to the kidneys, were not used, and certain inconsistent data points were excluded.
“We encourage physicians to further study AKI-1, because we see some patients facing worse outcomes,” said lead author Li.
Grants from NIH’s National Institute of Diabetes and Digestive and Kidney Diseases, and the National Science Foundation, supported this research.
Read the full open access journal article in Communications Medicine, using the links in the citation below.
Citation
Li D, Chan HY, Yu ASL, Kellum JA, Fuhrman DY, Chrischilles EA, Cowell LG, Chandaka S, Kean J, McTigue KM, Mosa ASM, Taylor B, Waitman LR, Syed M, Hu Y, Liu, M. Clustering analysis of multi-site electronic health records reveals distinct subphenotypes in stage-1 acute kidney injury. Commun Med. 2025 July 3;5(274). https://doi.org/10.1038/s43856-025-00993-6