急性白血病患者PICC置管相关性感染风险预测模型建立及临床指导价值

Risk prediction model for PICC-related infection in patients with acute leukemia and its clinical value

目的探讨急性白血病患者经外周静脉中心静脉置管(PICC)相关性感染的危险因素,建立风险预测模型。方法选取2016年2月-2021年5月新疆医科大学第五附属医院收治的急性白血病PICC置管患者450例,按照PICC置管相关性感染发生情况,分为感染组71例和未感染组379例,多因素Logistic回归分析感染危险因素,基于回归系数构建预测模型。结果71例感染患者,培养分离病原菌83株,其中革兰阳性菌49株占59.04%,革兰阴性菌32株占38.55%,真菌2株占2.41%;多因素Logisitic回归分析显示,年龄≥60岁、合并糖尿病、化疗次数≥6次、穿刺次数≥3次、导管留置时间≥3个月是急性白血病PICC置管患者发生置管相关性感染的独立危险因素;基于回归系数构建的预测模型,Hosmer-Lemeshow检验具有较高的拟合优度(P=0.280);对模型进行内部验证,其ROC曲线下面积(AUC)为0.855(95%CI:0.814-0.896),灵敏度为83.10%,特异度为82.10%。结论基于回归系数建立的风险评分模型具有较好的判别效度,可为针对性加强急性白血病患者PICC置管相关性感染预防控制措施提供重要的理论依据。

OBJECTIVE To explore the risk factors for peripherally inserted central catheter(PICC)-related infection in patients with acute leukemia and set up the risk prediction model.METHODS Totally 450 acute leukemia patients who were treated with PICC in the Fifth Affiliated Hospital of Xinjiang Medical University from Feb 2016 to May 2021 were enrolled in the study and divided into the infection group with 71 cases and the no infection group with 379 cases according to the status of PICC-related infection.Multivariate logistic regression analysis was performed for risk factors for the infection,and the risk prediction model was established based on regression coefficients.RESULTS A total of 83 strains of pathogens were isolated from the 71 patients with infection,49(59.04%)of which were gram-positive bacteria,32(38.55%)were gram-negative bacteria,and 2(2.41%)were fungi.Multivariate logistic regression analysis showed that the no less than 60 years of age,complication with diabetes mellitus,no less than 6 times of chemotherapy,no less than 3 times of puncture and catheter indwelling time no less than 3 months were independent risk factors for the PICC-related infection.Hosmer-Lemeshow test showed the risk prediction model established based on regression coefficients had high goodness of fit(P=0.280).The result of internal validation indicated that the area under ROC curve of the model was 0.855(95%CI:0.814~0.896),with the sensitivity 83.10%,the specificity 82.10%.CONCLUSION The risk scoring model established based on the regression coefficients has good discriminative validity and provides significant theoretical basis for putting forward targeted prevention measures for PICC-related infection in the patients with acute leukemia.