Abstract:Background: Acute kidney injury (AKI) is a common adverse outcome following nephrectomy. The progression from AKI to acute kidney disease (AKD) and subsequently to chronic kidney disease (CKD) remains a concern; yet, the predictive mechanisms for these transitions are not fully understood. Interpretable machine learning (ML) models offer insights into how clinical features influence long-term renal function outcomes after nephrectomy, providing a more precise framework for identifying patients at risk and supporting improved clinical decision-making processes. Objective: This study aimed to (1) evaluate postnephrectomy rates of AKI, AKD, and CKD, analyzing long-term renal outcomes along different trajectories; (2) interpret AKD and CKD models using Shapley Additive Explanations values and Local Interpretable Model-Agnostic Explanations algorithm; and (3) develop a web-based tool for estimating AKD or CKD risk after nephrectomy. Methods: We conducted a retrospective cohort study involving patients who underwent nephrectomy between July 2012 and June 2019. Patient data were randomly split into training, validation, and test sets, maintaining a ratio of 76.5:8.5:15. Eight ML algorithms were used to construct predictive models for postoperative AKD and CKD. The performance of the best-performing models was assessed using various metrics. We used various Shapley Additive Explanations plots and Local Interpretable Model-Agnostic Explanations bar plots to interpret the model and generated directed acyclic graphs to explore the potential causal relationships between features. Additionally, we developed a web-based prediction tool using the top 10 features for AKD prediction and the top 5 features for CKD prediction. Results: The study cohort comprised 1559 patients. Incidence rates for AKI, AKD, and CKD were 21.7% (n=330), 15.3% (n=238), and 10.6% (n=165), respectively. Among the evaluated ML models, the Light Gradient-Boosting Machine (LightGBM) model demonstrated superior performance, with an area under the receiver operating characteristic curve of 0.97 for AKD prediction and 0.96 for CKD prediction. Performance metrics and plots highlighted the model's competence in discrimination, calibration, and clinical applicability. Operative duration, hemoglobin, blood loss, urine protein, and hematocrit were identified as the top 5 features associated with predicted AKD. Baseline estimated glomerular filtration rate, pathology, trajectories of renal function, age, and total bilirubin were the top 5 features associated with predicted CKD. Additionally, we developed a web application using the LightGBM model to estimate AKD and CKD risks. Conclusions: An interpretable ML model effectively elucidated its decision-making process in identifying patients at risk of AKD and CKD following nephrectomy by enumerating critical features. The web-based calculator, found on the LightGBM model, can assist in formulating more personalized and evidence-based clinical strategies.

Optimal Predicting Method of Peritoneal Fluid Absorption Rate Using Genetic Algorithm Embedded in Neural Network

[An optimal predicting method based on improved genetic algorithm embedded in neural network and its application to peritoneal dialysis]

Intelligent Predicting Approach of Peritoneal Fluid Absorption Rate Based-on Neural Network

Prediction Method Of Peritoneal Fluid Absorption In Renal Failure Therapy Tbased On Least-Square Support Vector Machine

Prediction Model for Patients with Acute Respiratory Distress Syndrome: Use of a Genetic Algorithm to Develop a Neural Network Model

Random forest can accurately predict the technique failure of peritoneal dialysis associated peritonitis patients

Chronic Kidney Disease Survival Prediction With Artificial Neural Networks

Determining factors that predict technique survival on peritoneal dialysis: application of regression and artificial neural network methods

Machine learning algorithms for the prediction of adverse prognosis in patients undergoing peritoneal dialysis

Nursing Scheme Based on Back Propagation Neural Network and Probabilistic Neural Network in Chronic Kidney Disease

[Application of recurrent neural network in prognosis of peritoneal dialysis]

Construction and evaluation of an integrated predictive model for chronic kidney disease based on the random forest and artificial neural network approaches

Artificial neural networks analysis predicts long-term fistula function in hemodialysis patients following percutaneous transluminal angioplasty

A Deep Learning Approach for the Estimation of Glomerular Filtration Rate

Prediction of mortality in hemodialysis patients based on autoencoders

#766 A Prognostic Model (BLEND35) with Internal Validation for Treatment Failure in Patients with Peritoneal Dialysis-Associated Peritonitis

Achieving High Accuracy in Predicting Blood Flow Volume at the Arteriovenous Fistulas of Hemodialysis Patients by Intelligent Quality Assessment on PPGs

Development and validation of prediction model for technique failure in peritoneal dialysis patients: An observational study

Personalized Prediction of Long-Term Renal Function Prognosis Following Nephrectomy Using Interpretable Machine Learning Algorithms: Case-Control Study

Establishment and Evaluation of Artificial Intelligence-Based Prediction Models for Chronic Kidney Disease under the Background of Big Data

Automated prognosis of renal function decline in ADPKD patients using deep learning