Development and external validation of a machine learning model for cardiac valve calcification early screening in dialysis patients: a multicenter study
Xiaoxu Wang,
Yinfang Li,
Zixin Cao,
Yunuo Li,
Jingyuan Cao,
Yao Wang,
Min Li,
Jing Zheng,
Siqi Peng,
Wen Shi,
Qianqian Wu,
Junlan Yang,
Yaping Fang,
Aiqing Zhang,
Xiaoliang Zhang,
Bin Wang
Affiliations
Xiaoxu Wang
Department of Nephrology, Zhongda Hospital, Southeast University School of Medicine, Nanjing, P.R. China
Yinfang Li
Department of Pediatric, The Second Affiliated Hospital of Nanjing Medical University, School of Pediatric, Nanjing Medical University, Nanjing, P.R. China
Zixin Cao
Department of Nephrology, Zhongda Hospital, Southeast University School of Medicine, Nanjing, P.R. China
Yunuo Li
Department of Nephrology, Zhongda Hospital, Southeast University School of Medicine, Nanjing, P.R. China
Jingyuan Cao
Department of Nephrology, The Affiliated Taizhou People’s Hospital of Nanjing Medical University, Taizhou School of Clinical Medicine, Nanjing Medical University, Taizhou, P.R. China
Yao Wang
Department of Nephrology, The Affiliated Hospital of Yangzhou University, Yangzhou University, Yangzhou, P.R. China
Min Li
Department of Nephrology, The Third Affiliated Hospital of Soochow University, Soochow University, Changzhou, P.R. China
Jing Zheng
Department of Nephrology, Zhongda Hospital, Southeast University School of Medicine, Nanjing, P.R. China
Siqi Peng
Department of Nephrology, Zhongda Hospital, Southeast University School of Medicine, Nanjing, P.R. China
Wen Shi
Department of Nephrology, Zhongda Hospital, Southeast University School of Medicine, Nanjing, P.R. China
Qianqian Wu
Department of Nephrology, Zhongda Hospital, Southeast University School of Medicine, Nanjing, P.R. China
Junlan Yang
Department of Nephrology, Zhongda Hospital, Southeast University School of Medicine, Nanjing, P.R. China
Yaping Fang
Department of Nephrology, Zhongda Hospital, Southeast University School of Medicine, Nanjing, P.R. China
Aiqing Zhang
Department of Pediatric, The Fourth Affiliated Hospital of Nanjing Medical University, Nanjing Medical University, Nanjing, P.R. China
Xiaoliang Zhang
Department of Nephrology, Zhongda Hospital, Southeast University School of Medicine, Nanjing, P.R. China
Bin Wang
Department of Nephrology, Zhongda Hospital, Southeast University School of Medicine, Nanjing, P.R. China
Background Cardiac valve calcification (CVC) is common in dialysis patients and associated with increased cardiovascular risk. However, early screening has been limited by cost concerns. This study aimed to develop and validate a machine learning model to enhance early detection of CVC.Methods Data were collected at four centers between 2020 and 2023, including 852 dialysis patients in the development dataset and 661 in the external validation dataset. Predictive factors were selected using LASSO regression combined with univariate and multivariate analyses. Machine learning models including CatBoost, XGBoost, decision tree, support vector machine, random forest, and logistic regression were used to develop the CVC risk model. Model performance was evaluated in both validation sets. Risk thresholds were defined using the Youden index and validated in the external dataset.Results In the development dataset, 32.9% of patients were diagnosed with CVC. Age, dialysis duration, alkaline phosphatase, apolipoprotein A1, and intact parathyroid hormone were selected to construct the CVC risk prediction model. CatBoost exhibited the best performance in the training dataset. The logistic regression model demonstrated the best predictive performance in both internal and external validation sets, with AUROCs of 0.806 (95% CI 0.750–0.863) and 0.757 (95% CI 0.720–0.793), respectively. Calibration curves and decision curves confirmed its predictive accuracy and clinical applicability. The logistic regression model was selected as the optimal model and achieved excellent risk stratification in CVC risk prediction.Conclusion The predictive model effectively identifies CVC risk in dialysis patients and offers a robust tool for early detection and improved management.
