Machine learning models for predicting the onset of chronic kidney disease after surgery in patients with renal cell carcinoma

Seol Whan Oh ^1,2, Seok-Soo Byun ³, Jung Kwon Kim ³

¹Department of Medical Informatics, College of Medicine, The Catholic University of Korea, 06591, Seoul, Korea.
²Department of Biomedicine & Health Sciences, The Catholic University of Korea, 06591, Seoul, Korea.
³Department of Urology, Seoul National University College of Medicine, Seoul National University Bundang Hospital, 13620, Seongnam, Korea.
⁴Department of Urology, Seoul National University College of Medicine, Seoul National University Hospital, 03080, Seoul, Korea.
⁵Department of Urology, Chonnam National University Medical School, 61469, Gwangju, Korea.
⁶Department of Urology, Korea University School of Medicine, 02841, Seoul, Korea.
⁷Department of Urology, National Cancer Center, 10408, Goyang, Korea.
⁸Department of Urology, Chungbuk National University College of Medicine, 28644, Cheongju, Korea.
⁹Department of Urology, College of Medicine, Chungbuk National University, 28644, Cheongju, Korea.
¹⁰Department of Urology, School of Medicine, Kyungpook National University Chilgok Hospital, Kyungpook National University, 41404, Daegu, Korea.
¹¹Department of Urology, Seoul St. Mary's Hospital, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea. toomey@catholic.ac.kr.

⁰Department of Medical Informatics, College of Medicine, The Catholic University of Korea, 06591, Seoul, Korea.

Bmc Medical Informatics and Decision Making +

|

March 23, 2024

View abstract on PubMed

Summary

A new model predicts chronic kidney disease (CKD) after kidney cancer surgery. This tool helps manage patient risk and improve outcomes using key factors like preoperative eGFR and tumor size.

Area Of Science

Nephrology
Oncology
Medical Informatics

Background

Patients with renal cell carcinoma (RCC) face increased chronic kidney disease (CKD) risk post-nephrectomy.
Postoperative renal function monitoring and intervention are crucial.
A predictive tool for CKD onset is vital for patient management.

Purpose Of The Study

To develop a machine learning model for predicting CKD after RCC surgery.
To identify key predictors of post-nephrectomy CKD.

Main Methods

Utilized data from 4389 RCC patients across eight Korean hospitals (KORCC database).
Trained and evaluated nine machine learning models to predict CKD occurrence.
Selected the best model based on AUROC and validated variable importance using SHAP and Kaplan-Meier analyses.

Main Results

The gradient boost algorithm achieved the highest performance with an AUROC of 0.826.
Preoperative eGFR, albumin levels, and tumor size were identified as significant predictors of post-surgical CKD.

Conclusions

A novel predictive model for post-surgical CKD in RCC patients was developed.
This quantitative tool aids in assessing CKD risk, enabling personalized postoperative care and improved patient prognosis.

Keywords:

Chronic kidney disease Gradient boost KOrean Renal Cell Carcinoma Machine learning Renal cell carcinoma