A machine learning prediction model for waiting time to kidney transplant

Area of Science:

• Nephrology
• Transplant Surgery
• Biostatistics

Background:

• Predicting deceased donor kidney transplant waiting times is crucial for patient management and resource allocation.
• Machine learning offers a novel approach to estimate time on the kidney transplant waiting list.

Purpose of the Study:

• To develop and validate a machine learning model for predicting kidney transplant waiting times.
• To provide a tool for better patient and clinician decision-making regarding kidney transplantation.

Main Methods:

• Retrospective cohort study of patients on the São Paulo State Organ Allocation System waiting list (2000-2017).
• Data split into training (75%) and testing (25%) sets.
• Cox regression model used to predict deceased donor transplant outcomes; Cox hazard ratios informed risk-prediction equations.

Main Results:

• Analysis included 48,153 patient registries; approximately one-third received a deceased donor transplant.
• Key predictors of transplant likelihood included age, subregion, calculated panel reactive antibodies (cPRA), and HLA-A, -B, and -DR frequencies.
• The developed model demonstrated good predictive performance with a c-index of 0.70 in internal validation.

Conclusions:

• A kidney transplant waiting time calculator was developed with good predictive accuracy.
• The tool can aid candidates and providers in managing expectations and care.
• The calculator has the potential to optimize resource utilization and pre-transplant patient management.