Hong Zhang,1,2 Haoxiang Zhang,3 Ronghua Li,2,4 Lin Zhuo,2,5 Ling Liu,2,5 Ling Tan,2,5 Rongrong Li,2,5 Sai Zhang2,6 

1Teaching and Research Section of Clinical Nursing, Xiangya Hospital, Central South University, Changsha, Hunan, People’s Republic of China; 2National Clinical Research Center for Geriatric Disorders, Xiangya Hospital, Central South University, Changsha, Hunan, People’s Republic of China; 3Xiangya School of Medicine, Central South University, Changsha, Hunan, People’s Republic of China; 4Nuclear Medicine Department, Xiangya Hospital, Central South University, Changsha, Hunan, People’s Republic of China; 5Organ Transplant Center, Xiangya Hospital, Central South University, Changsha, Hunan, People’s Republic of China; 6Institute of Medical Sciences, Xiangya Hospital, Central South University, Changsha, Hunan, People’s Republic of China

Correspondence: Sai Zhang, Email zhangsai2000@csu.edu.cn

Introduction: Renal transplant recipients (RTRs) are at high risk of renal dysfunction, and one contributing factor may be abnormal blood lipids. This study aimed to establish a risk prediction model using machine learning (ML).
Methods: This retrospective cohort study recruited 345 RTRs and followed up for one year. Patients’ demographic and clinical characteristics were retrieved from the electronic medical record system. The cohort was randomly split into training (n = 276) and validation (n = 69) groups at a 4:1 ratio. Predictors of renal dysfunction were determined using three ML models: RandomForest, XGBoost, and LightGBM.
Results: During the one-year follow-up, 193 (55.9%) patients developed renal dysfunction. Among 20 demographic and clinical variables screened, five were identified as significant predictors: age, gender, HDL-C, non-HDL-C, and LDL-C. A nomogram was developed as a visual predictive tool to present the interplay between these variables graphically. It demonstrated good diagnostic performance, with an area under the curve (AUC) of 0.87 (95% CI, 0.85– 0.89) in the training group and 0.81 (95% CI, 0.78– 0.83) in the validation group.
Conclusion: Our study developed a risk prediction model to identify RTRs at high risk of renal dysfunction based on preoperative lipid profiles, which is crucial for optimizing patient management and improving the prognosis.

Keywords: kidney transplantation, renal dysfunction, eGFR, risk prediction, blood lipid levels, machine learning, nomogram