维持性血液透析患者衰弱风险预测模型的构建研究

Construction of a predictive model for risk of frailty in patients with maintenance hemodialysis

目的分析维持性血液透析患者衰弱的危险因素,构建衰弱风险预测模型,为预防和降低透析患者衰弱提供参考依据。方法采用方便抽样的方法,选取航天中心医院血液净化门诊的145例透析患者,采用Logistic回归构建风险预测模型,Hosmer-Lemeshow卡方检验评价拟合程度,ROC曲线下面积验证模型的预测效果。结果最终纳入性别(OR=7.385,95%CI:4.965~56.529,P=0.045)、居住方式(OR=4.823,95%CI:1.138~20.446,P=0.033)、营养评分(OR=0.453,95%CI:0.255~0.807,P=0.007)、血红蛋白(OR=0.146,95%CI:0.015~1.392,P=0.030)、Charlson共病指数(OR=5.918,95%CI:0.465~75.240,P=0.012)、自理能力评分(OR=0.589,95%CI:0.551~1.142,P=0.032)6个因素构建风险预测模型。模型有较好的拟合程度(χ^(2)=6.889,P=0.549),ROC曲线下AUC为0.940(P<0.001,95%CI:0.886~0.973),灵敏度为86.4%,特异度为86.0%。结论该模型具有较好的拟合程度,且预测效果较好,可指导医护人员早期预测维持性透析患者衰弱情况,为临床医护人员制定具体干预措施提供参考。

Objective To analyze the risk factors for frailty in maintenance hemodialysis(MHD)patients,and to construct a prediction model for the risk of frailty so as to provide a reference for the prevention and alleviation of frailty in dialysis patients.Methods Convenient sampling was used to select a total of 145 MHD patients treated in the Blood Purification Clinic of our hospital.Logistic regression was used to construct a risk prediction model.The Hosmer-Lemeshow chi-square test was used to evaluate performance of the model.For the degree of fit,the area under the ROC curve was used to verify the predictive effect of the model.Results This study finally included 6 factors,namely gender(OR=7.385,95%CI:4.965~56.529,P=0.045),living style(OR=4.823,95%CI:1.138~20.446,P=0.033),nutritional score(OR=0.453,95%CI:0.255~0.807,P=0.007),hemoglobin(OR=0.146,95%CI:0.015~1.392,P=0.030),Charlson comorbidity index(OR=5.918,95%CI:0.465~75.240,P=0.012),and self-care ability score(OR=0.589,95%CI:0.551~1.142,P=0.032),to construct a risk prediction model.The results of Hosmer-Lemeshow chi-square test showed a better degree of fit(χ^(2)=6.889,P=0.549)of the prediction model;the AUC under the ROC curve of the prediction model was 0.940(P<0.001,95%CI:0.886~0.973),with the sensitivity of 86.4%and the specificity of 86.0%.Conclusion This prediction model has a better degree of fit and a better prediction effect.It is useful for medical staff to predict frailty in MHD patients earlier and to provide references in planning specific intervention measures.