Related Experiment Video
Updated: Jun 23, 2026

A Rat Orthotopic Renal Transplantation Model for Renal Allograft Rejection
Published on: February 2, 2022
HRRT: hierarchical reinforcement learning for renal replacement therapy decision support.
Qianyi Xu1, Feng Wu1, Zi Yi Christopher Thong2
1Saw Swee Hock School of Public Health, National University of Singapore, Singapore, Singapore.
An AI system called Hierarchical Reinforcement Learning for Renal Replacement Therapy (HRRT) was developed to aid clinical decisions for acute kidney injury (AKI) patients. This AI approach reduced intensive care unit mortality by personalizing treatment adjustments.
Area of Science:
- Nephrology
- Artificial Intelligence
- Clinical Decision Support Systems
Background:
- Acute kidney injury (AKI) necessitates renal replacement therapy (RRT), but clinical decisions on timing, modality, and management are complex and vary widely.
- Current evidence from randomized controlled trials (RCTs) does not adequately support adaptive, patient-specific RRT strategies.
- Dynamic adjustment of RRT based on individual patient progression remains a significant clinical challenge.
Purpose of the Study:
- To develop a holistic clinical decision support system (CDSS) using Hierarchical Reinforcement Learning for Renal Replacement Therapy (HRRT).
- To cover the entire decision-making process for RRT, from initiation to weaning.
- To reduce practice variation and improve outcomes in AKI patients undergoing RRT.
Main Methods:
- A retrospective multi-center study was conducted.
- The HRRT system was trained and internally tested using data from 2467 Intensive Care Unit (ICU) stays of 1439 patients in a US hospital.
- External validation was performed on datasets from the Netherlands (1085 ICU stays) and China (1230 ICU stays).
Main Results:
- The HRRT system demonstrated a reduction in the estimated mortality rate by 6.1 percentage points (from 47.7% to 41.6%).
- Performance was evaluated on external validation sets from the Netherlands and China.
- The AI-driven approach showed potential in optimizing RRT management compared to clinician-led outcomes.
Conclusions:
- An AI-driven, holistic approach to RRT can effectively reduce inappropriate practice variation.
- Personalized treatment adjustments recommended by the HRRT system can lead to lower ICU mortality.
- The HRRT CDSS offers a promising strategy for improving RRT management in AKI patients.
Related Concept Videos
Continuous Renal Replacement Therapy
Extracorporeal Removal of Drugs: Continuous Renal Replacement Therapy
Dialysis
Acute kidney injury develops suddenly and can be caused by pre-renal causes (e.g., hypovolemia, shock), intrinsic renal causes (e.g., acute tubular necrosis), or post-renal causes (e.g., urinary obstruction). In contrast, chronic renal failure progresses gradually over time and is often...
Acute Kidney Injury V: Interprofessional Care
Heart Failure VI: Adjunct Therapies
Chronic Kidney Disease III: Interprofessional Care