Chronic kidney disease (CKD) is a complex condition that tells you that renal function is gradually decreasing and can eventually progress to end-stage renal disease (ESRD). Globally, the prevalence of CKD ranges from 8% to 16%, with about 5% to 10% of people diagnosed ultimately reaching ESRD, which has become a major public health challenge.
In the new study, researchers evaluated integrated clinical and claims data with the aim of improving prediction of CKD progression to ESRD using machine learning and deep learning models, as well as explainable artificial intelligence (AI). The integrated model outperforms a single data source model that can enhance CKD management, support targeted interventions, and reduce healthcare disparities.
The study, by researchers at Carnegie Mellon University, is published in the Journal of the American Medical Informatics Association.
“Our research presents a robust framework for predicting ESRD outcomes and improves clinical decision-making through integrated multiple data and advanced analytics,” explains Rema Padman, professor of management science and healthcare informatics at Carnegie Mellonsheinz College, who led the study. “Future research will expand data integration and expand this framework to other chronic diseases.”
The progression of CKD is divided into five stages, with culmination in ESRD when renal function decreases from 10% to 15% of normal capacity and requires dialysis or implantation for patient survival. The economic impact of CKD is important, and the relatively small proportion of Medicare CKD patients in the US contributes a disproportionately high share of Medicare costs, particularly when going to ESRD. Additionally, over a third of ESRD patients are readmitted within 30 days of discharge, highlighting the important need for early detection and management of the disease to prevent progression to ESRD, improve patient health outcomes, and reduce health costs.
In this study, the researchers used data from over 10,000 CKD patients and used a combination of clinical and billing information from 2009 to 2018. Five different observation windows were used to evaluate multiple statistics, machine learning, and deep learning models. Their work was supported by explanatory AI to increase interpretability and reduce bias.
The integrated data model of the survey is superior to a single data source model. The 24-month observation window was optimally balanced, balancing early detection and prediction accuracy. The 2021 estimated glomerular filtration rate equation improved prediction accuracy and reduced racial bias, particularly in African American patients.
“Our work fills the critical gap by developing a framework that uses integrated clinical and claims data rather than isolated data sources,” said Yubo Li, a doctoral student at Carnegie Mellonsheinz College, who co-authored the study. “By minimizing the observation window required for accurate prediction, our approach balances patient-centered practicality with clinical relevance. This integration improves both predictive accuracy and clinical utility, allowing for more informational decisions to improve patient outcomes.”
In the study limitations, the authors state that reliance on data from one institution could limit the generalization of the model to other care settings. Furthermore, the use of data from electronic health records can introduce observational bias, incomplete records, and underestimation of specific patient groups.
/Public release. This material of the Organization of Origin/Author is a point-in-time nature and may be edited for clarity, style and length. Mirage.news does not take any institutional position or aspect, and all views, positions and conclusions expressed here are the views of the authors alone.
