中心静脉导管相关性血流感染风险预测模型的系统评价

Risk prediction models for central venous Catheter-Related Bloodstream Infections:a systematic review

目的:系统检索并评价中心静脉导管相关性血流感染的风险预测模型,以期为临床选择及应用中心静脉导管相关性血流感染的风险预测模型提供参考。方法:检索中、英文数据库中关于中心静脉导管相关性血流感染的有关文献,检索时限为建库至2023年12月11日。2名研究员独立进行文献筛选和资料提取,使用预测模型研究的偏倚风险评价工具对纳入的研究进行偏倚风险和适用性评价。结果:共纳入16篇文献,涉及15个中心静脉导管相关性血流感染风险预测模型。15个模型的AUC或C指数为0.67~0.930;5个模型进行了内部验证,8个模型进行了外部验证。16篇文献质量评价结果均为高偏倚风险、低适用性风险。出现频次较高的预测因子包括:肠外营养、糖尿病、中心静脉导管留置时间、重症监护室患者或住院史、管腔数、使用抗生素、股静脉置入、恶性肿瘤。结论:现有的中心静脉导管相关性血流感染风险预测模型的预测性能较好,预测因子对临床具有一定的指导意义,但偏倚风险高,未来应对模型进行进一步验证,为制订有效的预防策略提供依据。

Objective:To systematically search and evaluate the risk prediction models for central venous Catheter-Related Bloodstream Infections,to provide reference for constructing a high-quality risk prediction model for central venous Catheter-Related Bloodstream Infections.Methods:We searched for relevant literature on central venous Catheter-Related Bloodstream Infections from Chinese and English databases from their inception to December 11,2023.Two researchers independently conducted literature screening and data extraction,using PROBAST tool to evaluate the risk of bias and applicability of included studies.Results:A total of 16 articles were included,involving 15 risk prediction models for central venous Catheter-Related Bloodstream Infections.The AUC or C index of 15 models ranges from 0.67 to 0.930;5 models received internal validation and 8 models external validation.Quality evaluation of the 16 articles showed high bias risk and low applicability risk.The predictive factors with higher frequency of occurrence included:parenteral nutrition(TPN),diabetes,duration of central venous catheterization,ICU patient or hospital history,number of lumens,use of antibiotics,femoral vein insertion,and malignant tumors.Conclusion:The existing risk prediction models for central venous Catheter-Related Bloodstream Infections have good predictive performance to determine the predictive factors which possess certain guiding significance for clinical practice,but the risk of bias is high.Further validation of future response models is needed to provide a basis for developing effective prevention strategies.