AI Model May Accurately Predict Kidney Failure Risk in Chronic Kidney Disease Patients: Study Shows - Video

A study by researchers at Carnegie Mellon University has shown that using integrated clinical and insurance claims data, analyzed through machine learning, deep learning, and explainable artificial intelligence (AI), can significantly improve the prediction of chronic kidney disease (CKD) progression to end-stage renal disease (ESRD). The findings, published in the Journal of the American Medical Informatics Association, could help clinicians intervene earlier and reduce healthcare disparities and costs.

Chronic kidney disease is a progressive condition in which kidney function gradually declines over time. It affects 8% to 16% of the global population, and around 5% to 10% of those diagnosed ultimately reach end-stage renal disease, a life-threatening stage requiring dialysis or kidney transplant. Early identification of high-risk patients is critical.

To address this challenge, the researchers analyzed data from more than 10,000 chronic kidney disease patients between 2009 and 2018, combining electronic health records with insurance claims data. They tested a variety of predictive models over five different observation periods and found that integrated models consistently outperformed those using a single data source. A 24-month observation window was found to be optimal, offering a balance between early detection and predictive accuracy.

“Our study presents a robust framework for predicting end-stage renal disease outcomes, improving clinical decision-making through integrated multisourced data and advanced analytics,” said Rema Padman, professor at Carnegie Mellon’s Heinz College, who led the research. “Future research will expand data integration and extend this framework to other chronic diseases.”

The team also found that using the 2021 estimated glomerular filtration rate (eGFR) equation significantly improved prediction accuracy and reduced racial bias—particularly for African American patients.

Despite promising results, the researchers acknowledged some limitations, including the use of data from a single institution and potential observational bias in electronic health records.

Reference: Yubo Li, Rema Padman, Enhancing end-stage renal disease outcome prediction: a multisourced data-driven approach, Journal of the American Medical Informatics Association, 2025;, ocaf118, https://doi.org/10.1093/jamia/ocaf118