Construction and evaluation of hyperkalemia risk prediction model in patients with chronic kidney disease
10.3760/cma.j.cn441217-20240315-00319
- VernacularTitle:慢性肾脏病患者高钾血症风险预测模型的构建与评价
- Author:
Dong DONG
1
;
Qiuling ZHANG
Author Information
1. 杭州市临平区第一人民医院肾内科，杭州 311100
- Keywords:
Renal insufficiency, chronic;
Hyperkalemial;
Risk factors;
Forecasting mode
- From:
Chinese Journal of Nephrology
2024;40(11):894-900
- CountryChina
- Language:Chinese
-
Abstract:
This study retrospectively analyzed the risk factors of hyperkalemia in patients with chronic kidney disease (CKD), established a risk assessment model, and validated it. It was a retrospective study. Data from 505 CKD patients who visited the Department of Nephrology at Linping First People's Hospital from June 2022 to January 2024 were retrospectively collected. Patients were divided into a modeling set ( n=354) and a validation set ( n=151) in a 7∶3 ratio. The independent risk factors for hyperkalemia in CKD patients were screened using least absolute shrinkage and selection operator (Lasso) regression, and a risk prediction model was constructed using multivariate Logistic regression. Hosmer-Lemeshow test was used to judge the fit degree of the prediction model, and the nomogram, the receiver operating characteristic curve, the calibration curve and clinical decision curve were drawn. The results showed that 155 out of 505 CKD patients developed hyperkalemia, with an incidence rate of 30.69%. Lasso regression and multivariate logistic regression analysis showed that serum albumin, blood sodium, blood phosphorus, parathyroid hormone, and risk score X (the value was based on the CKD hyperkalemia risk prediction model established in China in 2020) were the independent factors associated with hyperkalemia in CKD patients (all P<0.05). The analysis of the receiver operating characteristic curve of the subjects showed that the area under the modeling set curve was 0.840 (95% CI 0.796-0.884), and the area under the validation set curve was 0.849 (95% CI 0.784-0.915). The calibration curve suggested good consistency between the predicted and actual results of the model, and the decision curve analysis suggested that the model could increase patient benefits. Finally, the authors drew a visual nomogram of the model. The authors believe that the column chart prediction model for hyperkalemia in CKD patients constructed based on the predictive variables selected by Lasso regression ha good predictive ability and is helpful for clinical medical personnel to detect and manage hyperkalemia early in CKD patients.
