Integrated data models outperform single data source models, with highest AUROC for long short-term memory
By Elana Gotkine HealthDay Reporter
MONDAY, Sept. 15, 2025 (HealthDay News) — Machine learning and deep learning models applied to integrated clinical and claims data can improve prediction of chronic kidney disease (CKD) progression to end-stage renal disease (ESRD), according to a study published online Aug. 6 in the Journal of the American Medical Informatics Association.
Yubo Li and Rema Padman, Ph.D., from Carnegie Mellon University in Pittsburgh, used data from 10,326 CKD patients combining clinical and claims information from 2009 to 2018 to improve prediction of CKD progression to ESRD using machine learning and deep learning models. Five distinct observation windows were used, and key predictors were explored using feature importance and SHapley Additive exPlanations analysis.
The researchers found that integrated data models outperformed single data source models, with the highest area under the receiver operating characteristic curve and F1 score of 0.93 and 0.65, respectively, achieved with long short-term memory. Early detection and prediction accuracy was optimally balanced using a 24-month observation window. Prediction accuracy was improved with the 2021 estimated glomerular filtration rate equation, which also reduced racial bias, especially for African American patients.
“Our work bridges a critical gap by developing a framework that uses integrated clinical and claims data rather than isolated data sources,” Li said in a statement. “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.”
Copyright © 2025 HealthDay. All rights reserved.
