慢性肾脏病患者血钾异常风险预测模型的构建及验证

Construction and validation of a risk prediction model for abnormal serum potassium in patients with chronic kidney disease

目的探讨慢性肾脏病(CKD)患者发生血钾异常的危险因素,构建CKD患者血钾异常的风险预测模型。方法采用便利抽样的方法,选取2020年11月—2021年12月就诊于徐州医科大学第二附属医院肾脏内科的520例CKD患者为研究对象,将2020年11月—2021年9月的416例CKD患者作为建模组,将2021年10—12月的104例CKD患者作为验证组。统计并分析CKD患者发生血钾异常的危险因素,采用多重线性回归分析构建风险预测模型并利用受试者工作特征曲线下面积检验预测模型效果。结果416例患者中,血钾异常者129例,其中低血钾者101例,发生率为24.27%;高血钾者28例,发生率为6.73%。多重线性回归分析结果显示,是否有高血压、是否使用利尿剂及肾素-血管紧张素-醛固酮类药物、阴离子间隙、血尿素及患者肾脏病饮食依从态度量表得分对CKD患者血钾异常均有不同程度的影响(P<0.05)。最终构建的高血钾预测模型的受试者工作特征曲线下面积为0.844(95%CI:0.775~0.913),灵敏度为92.3%,特异度为52.6%,最大Youden指数为0.602;低血钾预测模型的受试者工作特征曲线下面积为0.805(95%CI:0.730~0.881),灵敏度为88.5%,特异度为52.6%,最大Youden指数为0.566。结论本模型预测效果良好,可为临床治疗和制订预防血钾异常的护理措施提供参考。

Objective To explore the risk factors of abnormal serum potassium in patients with chronic kidney disease(CKD),and construct a risk prediction model for abnormal serum potassium in CKD patients.Methods Totally 520 CKD patients who were treated in the Department of Nephrology,the Second Affiliated Hospital of Xuzhou Medical University from November 2020 to December 2021 were selected by convenience sampling,among whom 416 treated from November 2020 to September 2021 were included into the modeling group,and 104 treated from October to December 2021 were included into the verification group.The risk factors of abnormal serum potassium in CKD patients were statistically analyzed,the risk prediction model was constructed by multiple linear regression analysis,and the area under the receiver operating characteristic curve was used to test the effect of the prediction model.Results Among the 416 patients,129 cases had abnormal serum potassium,including 101 cases of hypokalemia,with an incidence rate of 24.27%;28 cases had hyperkalemia,with an incidence rate of 6.73%.Multiple linear regression analysis showed that whether there was hypertension,whether diuretics and RAAS drugs were used,AG,BUN,and patients'RAAQ scores all had varying degrees of influence on abnormal serum potassium in CKD patients(P<0.05).The area under the receiver operating characteristic curve of the finally constructed hyperkalemia prediction model was 0.844(95%CI:0.775-0.913),with a sensitivity of 92.3%,a specificity of 52.6%,and the maximum Youden index of 0.602;the area under the receiver operating characteristic curve of the hypokalemia prediction model was 0.805(95%CI:0.730-0.881),with a sensitivity of 88.5%,a specificity of 52.6%,and the maximum Youden index of 0.566.Conclusions The prediction effect of this model is good,which can provide a reference for clinical treatment and formulation of nursing measures to prevent abnormal serum potassium.