目的 探讨慢性肾衰竭维持性血液透析（maintenance hemodialysis，MHD）患者泌尿系统感染的影响因素，构建并验证风险预测模型。方法 回顾性选取2021年1月至2023年1月于某区院行MHD的360例慢性肾衰竭患者为研究对象，按7∶3比例随机分为建模集（n=252）和验证集（n=108），根据建模集患者是否发生过泌尿系统感染分为感染组（n=37）和非感染组（n=215），对两组患者的资料进行对比分析，构建风险预测模型。绘制ROC曲线、决策曲线以及Bootstrap法内部验证校准曲线评估模型的预测效能，同时分析验证集的病例数据，对模型进行外部验证。结果 单因素及多因素Logistic回归分析结果显示，年龄、透析时间、泌尿道插管、糖尿病肾病、高血压肾病和血清白蛋白与慢性肾衰竭MHD患者泌尿系统感染有关（均P＜0.05）；预测模型接受者操作特性曲线（receiver operating characteristic curve，ROC）下面积为0.880，95%CI（0.820~0.940），Youden指数为0.640，灵敏度和特异度分别为91.9%和72.1%；决策曲线结果显示，该模型的潜在临床获益可观，可用性较高；Bootstrap法内部验证校准曲线的平均绝对误差为0.018；验证集ROC曲线下面积为0.894，95%CI（0.817~0.970），Youden指数为0.623，灵敏度和特异度分别为93.8%和68.5%，校准曲线MAE为0.042，提示模型的外部预测性能良好。结论 慢性肾衰竭MHD患者泌尿系统感染受年龄、透析时间、泌尿道插管等因素影响，基于各独立危险因素构建的风险预测模型的预测性能良好，可为临床预防慢性肾衰竭MHD患者发生泌尿系统感染提供护理参考。

Objective To investigate the contributing variables of urinary tract infection in individuals suffering from chronic kidney failure and undergoing maintenance hemodialysis,and to develop and verify a risk prediction model.Methods Retrospectively,360 chronic renal failure patients who received MHD in our hospital between January 2021 and 2023 were selected.3 out of 10 subjects were put into the validation set (n=108) and 7 into the modelling set (n=252).The latter were further divided into the infected group (n=37) and non-infected group (n=215) based on whether they had had urinary system infection.Data from the two groups were compared and examined to make a risk prediction model.The ROC curve,decision curve,and internal validation calibration curve in Bootstrap method’s were drawn to assess the prediction effectiveness of the model,and the case data of the validation set were also analyzed concurrently so as to externally validate the model.Results In MHD patients with chronic renal failure,urinary system infection was related to age,dialysis time,urinary tract intubation,diabetic nephropathy,hypertensive nephropathy,and serum albumin (P＜0.05),according to univariate and multivariate logistic regression analysis.The area under the ROC was 0.880,95%CI (0.820~0.940),with the values for the Youden index 0.640,scale of sensitivity and specificity 91.9% and 72.1%,respectively.The decision curve showed the substantial potential clinical benefit and availability of the model.The mean absolute deviation of the internal validation calibration curve in Bootstrap method was 0.894,95%CI(0.817~0.970),with the values for the Youden index as 0.623,scale of sensitivity and specificity 93.8% and 68.5%, respectively. The calibration curve was 0.042,which suggested that the model performed well in terms of external prediction.Conclusions Age,length of dialysis,urinary tract intubation,and other variables all influence the risk of urinary system infection in MHD patients with chronic renal failure.The risk prediction model based on each independent risk factor performs well in terms of prediction,and can offer nursing guidance for the prevention of urinary tract infection in these patients.

R459.5

湖州市科学技术局公益性应用研究项目（2022GYB55）