慢性肾脏病高钾血症风险评估模型的建立

Development of a hyperkalemia risk assessment model for patients with chronic kidney disease

目的:探讨我国慢性肾脏病(CKD)患者发生高钾血症的影响因素,并建立风险评估模型。方法:回顾性收集2017年5月至2020年6月来自全国14家医院的CKD 3~5期患者的临床数据。通过随机均衡抽样分为模型训练集和模型验证集,在模型训练集中通过单因素及多因素logistic回归分析方法筛选CKD患者发生高钾血症的影响因素并赋分。在模型验证集中,绘制受试者工作特征(ROC)曲线并计算曲线下面积(AUC),验证模型的评估效果。结果:共有847例CKD患者的临床数据被纳入分析,年龄(57.2±15.6)岁,男555例,女292例。其中训练集675例,验证集172例。多因素logistic回归模型纳入了年龄、CKD分期、心力衰竭史、血钾≥5.0 mmol/L史、糖尿病、酸中毒及使用升高血钾的药物,并根据这些因素建立评估模型。在验证集中,评估模型的ROC曲线下面积为0.809,具有较好的准确性,当cut-off值为4分时,对于高血钾事件预测灵敏度为87.1%,特异度为57.0%。结论:本研究初步建立了CKD患者发生高钾血症的风险评估模型,可进一步优化临床医师对于CKD患者的血钾管理。

Objective To investigate risk factors for hyperkalemia among chronic kidney disease(CKD)patients and establish a risk assessment model for predicting hyperkalemia events.Methods Clinical data of CKD patients(stage 3 to 5)hospitalized between May 2017 and June 2020 from 14 hospitals were retrospectively collected and divided into training dataset and validation dataset through balanced random sampling.Multivariate logistic regression analysis was used to analyze risk factors for hyperkalemia in CKD patients and the factors were scored.Receiver operating characteristic(ROC)curve was plotted and the area under the curve(AUC)was calculated.Meanwhile,the cut-off value with the best sensitivity and specificity were used to verify the accuracy of the model in validation dataset.Results A total of 847 CKD patients were enrolled and further divided into training dataset(n=675)and validation dataset(n=172).There were 555 males and 292 females,with a mean age of(57.2±15.6)years.Multivariate logistic regression analysis showed that age,CKD stage,history of heart failure,history of serum potassium≥5.0 mmol/L,diabetes,metabolic acidosis,and use of medications that increase serum potassium levels were risk factors for causing hyperkalemia in patients with CKD.Risk assessment model was established based on these risk factors.The AUC of the ROC curve was 0.809.Using 4 as the cut-off value,the sensitivity and specificity for predicting hyperkalemia events reached 87.1%and 57.0%,respectively.Conclusion The model established in the current study can be used for predicting hyperkalemia events in clinical practices,which offers a new way to optimize serum potassium management in patients with CKD.