利奈唑胺致重症患者低钠血症预测模型的建立与评价  

Establishment and evaluation of a predictive model for linezolid-associated hyponatremia in critical patients

作　　者：杨万枝[1] 胡章海 居靖[2] YANG Wan-zhi;HU Zhang-hai;JU Jing(Department of Pharmacy,First Affiliated Municipal People's Hospital,Anhui Medical University,Anhui Anqing 246003,China;Department of Equipment,Affiliated Anqing Hospital,Anhui Medical University,Anhui Anqing 246003,China)

机构地区：[1]安徽医科大学附属安庆第一人民医院药学部,安徽安庆246003 [2]安徽医科大学附属安庆医院设备科,安徽安庆246003

出　　处：《中国医院药学杂志》2021年第18期1866-1869,1894,共5页Chinese Journal of Hospital Pharmacy

基　　金：安徽省高校自然科学研究项目(编号:KJ2020ZD21);安庆市第一人民医院2021年“三新”立项项目(编号:Y2021030)。

摘　　要：目的:建立利奈唑胺致重症患者低钠血症的预测模型,并验证模型的准确率。方法:选取2019年1月至2020年12月于安庆市第一人民医院使用利奈唑胺的89例重症患者作为训练集,回顾性分析患者的临床资料,包括性别、年龄、用药天数、用药前实验室检查指标、肝肾功能、合并疾病与联合用药等。采用Logistic回归分析筛选低钠血症发生的独立危险因素,建立预测模型,并采用受试者工作特征曲线(receiver operating characteristic curve, ROC)分析评价预测模型,另外选取2021年1-3月的30例符合纳入标准的患者的临床资料作为验证集,交叉验证预测模型的准确率。结果:89例患者低钠血症发生率为21.33%,单因素分析提示,与未发生组相比,低钠血症发生组用药天数、肾小球滤过率(eGFR)、C-反应蛋白、高血压和丙戊酸钠方面的差异有统计学意义(P<0.05)。多因素Logistic回归分析显示用药天数和eGFR是低钠血症发生的独立危险因素,2种危险因素联合建立Logistic回归方程,即Z=-1.751+0.313X_(用药天数)-0.059X_(eGFR),计算89例患者的Z值,ROC曲线下面积为(AUC=0.894,95%CI:0.828~0.960,P<0.001),Youden指数最大时(0.711)的切点为ROC曲线上的最佳临界值(-2.09),30例患者验证该模型准确率为83.33%。结论:用药天数和eGFR是重症患者使用利奈唑胺致低钠血症的高危因素,两者拟合模型预测低钠血症发生有一定价值,其准确率为83.33%。OBJECTIVE To establish a predictive model for linezolid-associated hyponatremia in critical patients and verify its accuracy.METHODS From January 2019 to December 2020,a total of 89 critical patients were treated with linezolid.We selected and analyzed retrospectively their clinical data, including gender, age, duration of medication, pre-medication laboratory examination indexes, hepatic and renal function, concurrent diseases and co-medication, etc.Logistic regression analysis was utilized for screening independent risk factors for the occurrence of hyponatremia and a predictive model was established.ROC curve analysis was performed for evaluating the predictive model.Clinical data of another 30 patients were selected for verifying the accuracy of the predictive model.RESULTS The incidence of hyponatremia was 21.33%.Univariate analysis showed that as compared with non-occurrence group, statistically significant differences existed in duration of medication, estimated glomerular filtration rate(eGFR),C-reactive protein, hypertension and sodium valproate in occurrence group(P<0.05).Multivariate Logistic regression analysis indicated that duration of medication and eGFR were independent risk factors for the occurrence of hyponatremia.Logistic regression equation was established by combining two risk factors.Z=-1.751+0.313X_(T)-0.059X_(eGFR),Z values were calculated, the area under the ROC curve was(AUC= 0.894,95%CI:0.828-0.960,P<0.001)with a maximal Youden index of 0.711.The cut-off point was the optimal value(-2.09)on the ROC curve and the accuracy of the model was 83.33%.CONCLUSION Duration of medication and eGFR are the high risk factors of hyponatremia caused by lineczolid in critical patients.Based upon the combination of two factors, the predictive model is valuable in predicting hyponatremia with an accuracy of 83.33%.

关 键 词：利奈唑胺 重症患者 低钠血症 预测模型 

分 类 号：R978[医药卫生—药品]