维持性血液透析患者衰弱风险预测模型的构建及应用研究

Construction and application of frailty risk prediction model in maintenance hemodialysis patients

目的 构建并应用维持性血液透析患者衰弱风险预测模型。方法 采用前瞻性研究设计,选取2020年3月-2022年4月在浙江省某三级甲等医院接受维持性血液透析治疗的876例患者作为研究对象,其中2020年3月-2021年7月为建模组(n=491),2021年8月-2022年4月为验证组(n=385)。采用单因素和多因素Logistic回归分析维持性血液透析患者衰弱的危险因素,建立风险预测模型并绘制列线图。采用Hosmer-Lemeshow拟合优度检验、受试者操作特征曲线下面积(area under curve,AUC)评价模型临床预测效果,采用Bootstrap抽样法对模型进行内部验证。结果 共226例患者(25.80%)发生衰弱,其中建模组123例(25.05%),验证组103例(26.75%)。年龄(OR=3.553)、日常生活活动能力(OR=37.804)、脑卒中史(OR=16.434)、血清白蛋白(OR=4.197)、C反应蛋白(OR=2.633)及血肌酐(OR=2.201)6个影响因素构建预测模型。建模组Hosmer-Lemeshow拟合优度检验χ^(2)=5.667,P=0.772,AUC为0.955[95%CI(0.936,0.974)],灵敏度89.7%,特异度87.7%,约登指数为0.774,内部验证C-statistic统计量值为0.950。验证组Hosmer-Lemeshow拟合优度检验χ^(2)=44.085,P=1.362,AUC为0.914[95%CI(0.882,0.946)],灵敏度85.5%,特异度85.4%,约登指数为0.709,准确度为73.5%。结论 维持性血液透析患者衰弱风险预测模型能较好地可视化预测患者衰弱的发生风险,为医护人员早期识别和干预提供支持。

Objective To develop predictive model of frailty risk prediction in maintenance hemodialysis patients.Methods Prospective study design was adopted,a total of 876 patients who received hemodialysis in a tertiary class A hospital in Zhejiang Province from March 2020 to April 2022 were recruited,including March 2020 to July 2021 as the modeling group(n=491)and August 2021 to April 2022 as the validation group(n=385).Univariate and multivariate logistic regression were used to analyze the risk factors of frailty in maintenance hemodialysis patients,and we established a risk prediction model and to draw a nomogram.Hosmer-Lemeshow test and area under receiver operating characteristic were used to evaluate the clinical prediction effect of the model.The bootstrap method sampling method was used to internally validate the model.Results 226 patients(25.80%)had frailty,123(25.05%)in the modeling group and 103(26.75%)in the validation group.Six influencing factors,including age(OR=3.553),activities of daily living(OR=37.804),history of stroke(OR=16.434),serum albumin(OR=4.197),C-reactive protein(OR=2.633),and serum creatinine(OR=2.201),were used to construct the prediction model.In the modeling group,Hosmer-Lemeshow goodness-of-fit test showedχ^(2)=5.667,P=0.772,the AUC was 0.955[95%CI(0.936,0.974)],the sensitivity was 89.7%,the specificity was 87.7%,the Youden index was 0.774,and internal validation C-statistic value was 0.950.In the validation group,Hosmer-Lemeshow goodness-of-fit test showed χ^(2)=44.085,P=1.362,the AUC was 0.914[95%CI(0.882,0.946)],the sensitivity was 85.5%,the specificity was 85.4%,the Youden index was 0.709,and the accuracy was 73.5%.Conclusion The risk prediction model of frailty in maintenance hemodialysis patients can better visually predict the occurrence risk of frailty in patients,providing support for the early identification and intervention of medical staff.