Reproduce “Enhancing End Stage Renal Disease Outcome Prediction: A Multi-Sourced Data-Driven Approach” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Yubo Li, Rema Padman

Carnegie Mellon University researchers combined clinical electronic health records and administrative claims data with deep learning models to improve prediction of End Stage Renal Disease progression in chronic kidney disease patients. The integrated approach yielded an AUROC of 0.93 and F1 score of 0.65 over a 24-month observation window, while also examining the influence of race-free eGFR equations on predictive accuracy and racial disparities.

Key insights

  • Integrating multi-sourced clinical and administrative claims data consistently yields superior predictive performance for ESRD progression compared to models based on single data sources.
  • Deep learning models, particularly LSTM, are effective in leveraging integrated datasets to capture complex temporal dependencies and achieve high prediction accuracy.
  • A 24-month observation window provides an optimal balance between early detection capabilities and high predictive accuracy for ESRD progression.
  • The 2021 race-free eGFR equation shows potential in mitigating racial bias in predictions, specifically by reducing misclassifications for African American patients.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.