Machine Learning for Predicting Renal Replacement Therapy Onset in Chronic Kidney Disease
Date:
I presented our work at the Applications of Medical AI (AMAI) Workshop at MICCAI 2022, where our paper received the Best Paper Award. This work introduces a dynamic prediction model capable of identifying chronic kidney disease patients at high risk of requiring renal replacement therapy up to one year in advance.
The presentation summarized a large-scale analysis of Medicare claims data and a machine learning framework designed to support early clinical intervention. The paper demonstrates that a time-bucketed linear model can achieve high sensitivity and specificity across multiple prediction horizons (30–365 days), using only routinely available administrative data. This work highlights the potential for AI-driven early risk stratification to improve dialysis access planning, reduce emergency starts, and ultimately improve patient outcomes.
