Integrating Clinical and Histopathological Data to Predict Delayed Graft Function in Kidney Transplant Recipients...

Area of Science:

• Nephrology
• Medical Informatics
• Biostatistics

Background:

• Delayed graft function (DGF) significantly impacts kidney transplant outcomes.
• Accurate prediction of DGF risk is crucial for deceased-donor kidney transplantation (DDKT).

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting DGF risk in DDKT.
• To compare the performance of ML models against traditional logistic regression.

Main Methods:

• Retrospective cohort study using clinical and histopathological data (2018-2022).
• Development and evaluation of neural network, random forest, and extreme gradient boosting (XGBoost) models.
• Performance assessment using Area Under the Receiver Operating Characteristic Curve (AUROC) and Brier score.

Main Results:

• 64 out of 354 (18.1%) DDKT recipients experienced DGF.
• Predictors included donor BMI > 23, donor diabetes, prolonged cold ischemia, male recipient, and specific biopsy scores.
• XGBoost model achieved an AUROC of 0.989, outperforming other models and logistic regression.

Conclusions:

• ML models, particularly XGBoost, show strong potential for DGF risk assessment in DDKT.
• These models can aid in donor acceptance decisions and potentially reduce the need for invasive procedures.
• XGBoost models may be particularly valuable in resource-limited settings for optimizing kidney transplantation outcomes.