JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Developing clinical prognostic models to predict graft survival after renal transplantation: comparison of statistical and machine learning models

  • 0Department of Statistics, Bahir Dar University, Bahir Dar, Ethiopia. gech.marr@gmail.com.
Bmc Medical Informatics and Decision Making +

|

February 3, 2025

|

February 3, 2025

View abstract on PubMed

Summary

This summary is machine-generated.

Predicting kidney transplant success is vital. Machine learning, particularly Stochastic Gradient Boosting, offers superior accuracy for graft survival prediction compared to traditional methods.

Area Of Science

  • Nephrology
  • Transplantation Medicine
  • Biostatistics

Background

  • Graft failure is a major challenge in renal transplantation for end-stage renal disease patients.
  • Accurate prediction of graft survival is essential for identifying high-risk individuals and improving patient outcomes.
  • This study aimed to develop and compare statistical and machine learning models for predicting renal graft survival.

Purpose Of The Study

  • To develop prognostic models for renal graft survival prediction.
  • To compare the performance of various statistical and machine learning models.
  • To identify key predictors of renal graft survival.

Main Methods

  • Utilized data from 278 renal transplant recipients.
  • Applied SMOTE resampling to address class imbalance.
  • Evaluated Standard/penalized Cox models, Random Survival Forest, and Stochastic Gradient Boosting (SGB).

Main Results

  • Median graft survival was 33 months; mean hazard of graft failure was 0.0755.
  • Stochastic Gradient Boosting (SGB) model showed superior discrimination (C-index=0.943) and calibration.
  • Key predictors included rejection episodes, urological complications, nonadherence, BUN, exercise, and marital status.

Conclusions

  • The SGB model achieved the highest predictive performance for renal graft survival.
  • The Ridge-Cox model provided comparable performance with enhanced interpretability.
  • Findings support integrating advanced models for improved risk stratification and personalized care in kidney transplant recipients.

Keywords: