个体化预测维持性血液透析患者口渴风险的列线图模型建立

Establishment of a nomogram model for individualized prediction of thirst risk in maintenance hemodialysis patients

目的 构建个体化预测维持性血液透析(maintenance hemodialysis,MHD)患者口渴风险的列线图模型。方法 以便利抽样调查方法选取2021年5月~2022年5月淮安市第二人民医院血液净化中心收治的200例MHD患者,采用自制调查问卷、透析口渴量表(dialysis thirst inventory,DTI)进行调查,依据DTI量表得分情况将200例MHD患者分为口渴组(DTI≥16分,n=36)与无口渴组(DTI<16分,n=164),用Logistic回归分析筛选影响MHD患者口渴风险的危险因素;采用R软件构建预测MHD患者口渴风险的列线图模型,且用ROC曲线、校准曲线进行列线图模型验证。结果 Logistic回归分析结果提示有糖尿病史(OR=3.174,95%CI:1.033~9.750,P=0.044)、口腔干燥量表(xerostomia inventory,XI)评分(OR=1.331,95%CI:1.201~1.475,P<0.001)、透析间期体质量增长率(OR=4.417,95%CI:2.054~9.497,P<0.001)是MHD患者口渴风险的独立危险因素。利用以上3个风险预测指标构建预测MHD患者口渴风险的列线图模型,模型验证结果显示:ROC曲线下面积0.897(95%CI:0.840~0.953)。校准曲线斜率接近1,且H-L拟合优度检验χ^(2)=8.830,P=0.357。结论 基于糖尿病史、XI评分、透析间期体质量增长率等3项指标构建的列线图模型对MHD患者口渴风险有较好的预测效能。

Objective To construct an individualized nomogram model for predicting the risk of thirst in maintenance hemodialysis(MHD) patients. Methods A convenience sampling survey method was performed to select 200 MHD patients admitted to the Blood Purification Center of Huai’an Second People’s Hospital from May 2021 to May 2022. A self-made questionnaire and the dialysis thirst inventory(DTI) were used for the investigation. According to the score of DTI scale, the 200 MHD patients were grouped into thirst group(DTI≥16 points, n=36) and non-thirst group(DTI <16 points, n=164). Logistic regression was used to identify the risk factors for thirst in MHD patients;R software was used to construct a nomogram model for predicting the risk of thirst in MHD patients. The ROC curve and calibration curve were used to verify efficiency of the nomogram model. Results Logistic regression showed that diabetes history(OR=3.174, 95%CI: 1.033~9.750, P=0.044), xerostomia inventory(XI) score(OR=1.331, 95% CI : 1.201~1.475, P<0.001),and interdialysis weight gain rate(IDWG%)(OR=4.417, 95% CI: 2.054~9.497, P<0.001) were the independent risk factors for thirst in MHD patients. The above three risk prediction indexes were used to construct a column graph model for predicting thirst risk in MHD patients. The verification results showed that the area under the ROC curve was 0.897(95% CI: 0.840~0.953), the slope of the calibration curve was close to 1, and the H-L goodness-of-fit test χ^(2)=8.830 and P=0.357. Conclusion The nomogram model constructed based on the three indicators of diabetes history, XI score and IDWG% has better predictive performance on thirst risk in MHD patients.