Machine Learning Models Improve Prediction of CKD Progression to End-Stage Renal Disease

New research published in Journal of American Medical Informatics Association demonstrates a framework to predict end-stage renal disease (ESRD) in patients with chronic kidney disease (CKD). The authors expressed their optimism that such information can improve clinical decision-making via integrated multisourced and advanced analytics, and future research may further expand upon the integration into other chronic disease spaces.¹

Image credit: Sansert | stock.adobe.com

CKD is a complex, multimorbid condition that is characterized by a gradual decline in kidney function, which can ultimately progress to ESRD. CKD’s global prevalence ranges from approximately 8% to 16%, with estimates suggesting that around 5% to 10% of individuals diagnosed with CKD eventually reach ESRD, representing a crucial public health challenge. This is especially significant in those with comorbidities, such as diabetes and hypertension.¹

To improve the prediction of CKD progression to ESRD, the investigators utilized machine learning and deep learning models, and when applied to clinical trials and claims data with various observation windows, artificial intelligence (AI) can help enhance interpretability and reduce bias. The investigators utilized data from 10,326 patients with CKD from 2009 to 2018.¹

After preprocessing, cohort identification, and feature engineering, the investigators assessed multiple statistical machine learning and deep learning models that used 5 different observation windows. Feature importance and SHapley Additive exPlanations analysis were employed to understand key predictors. Models were specifically tested for robustness, clinical relevance, misclassification patterns, and bias.¹

The authors observed that integrated data models outperformed single data source models, with long short-term memory achieving the highest area under the receiver operating characteristic curve (AUROC; 0.93) and F1 score (0.65). Additionally, a 24-month observation window optimally balanced early detection and prediction accuracy. The 2021 estimated glomerular filtration rate equation improved prediction accuracy and reduced racial bias, particularly for Black patients.¹

“Our study presents a robust framework for predicting ESRD outcomes, improving clinical decision-making through integrated multisourced data and advanced analytics,” lead study investigator Rema Padman, professor of management science and healthcare informatics at Carnegie Mellon’s Heinz College, said in a news release. “Future research will expand data integration and extend this framework to other chronic diseases.”²

When CKD progresses to ESRD, dialysis or transplantation is necessary for patient survival. The economic impact of CKD is also significant, with a relatively small proportion of patients with CKD on US Medicare contributing to a disproportionately high share of Medicare expenses, especially when they progress to ESRD. Notably, more than a third of patients with ESRD are readmitted to a hospital within 30 days of discharge, emphasizing the critical need for early detection and management of CKD to prevent its progression to ESRD, improve patient health outcomes, and reduce health care costs.^1,2

Among the study’s limitations, the investigators wrote that their reliance on data from 1 institution may limit the generalizability of their model to other care settings. Also of note, their use of data from electronic health records can introduce observational bias, incomplete records, and underrepresentation of certain patient groups, which can undermine both accuracy and fairness amid evaluation.^1,2

“Our work bridges a critical gap by developing a framework that uses integrated clinical and claims data rather than isolated data sources,” coauthor Yubo Li, a PhD student at Carnegie Mellon’s Heinz College, explained. “By minimizing the observation window needed for accurate predictions, our approach balances clinical relevance with patient-centered practicality; this integration enhances both predictive accuracy and clinical utility, enabling more informed decision-making to improve patient outcomes.”²

REFERENCES

1. Li Y, Padman R. Enhancing end-stage renal disease outcome prediction: a multisourced data-driven approach. J Am Med Inform Assoc. Published online August 6, 2025. doi:10.1093/jamia/ocaf118

2. Carnegie Mellon University. Study used AI models to improve prediction of chronic kidney disease progression to end stage renal disease. News release. September 8, 2025. Accessed September 15, 2025. https://www.eurekalert.org/news-releases/1097590