中心静脉导管相关血流感染风险预测模型的系统评价

Systematic evaluation of a risk prediction model for catheter-related bloodstream infections in patients with central venous catheterization

目的系统评价中心静脉导管相关血流感染风险预测模型,为临床实践提供参考依据。方法检索中国知网、万方数据库、中国生物医学文献数据库、维普数据库、PubMed、Web of Science、Cochrane Library等数据库,检索时限为建库至2023年6月2日,由2名研究者独立进行文献的筛选和资料的提取,并根据预测模型偏倚风险评估工具对文献进行质量评价。结果共纳入9篇文献,包括9个中心静脉导管相关血流感染风险预测模型,总样本量为80~11901例,结局事件数为19~403个,纳入模型的C指数为0.81~0.93,受试者操作特征曲线下面积为0.73~0.90,模型出现频率较高的预测因子主要为糖尿病史、白蛋白水平、导管留置时间、置管部位等。预测模型偏倚风险评估工具评估结果显示,纳入的风险预测模型整体适用性较好,但偏倚风险较高,与研究样本数据来源欠妥、连续变量方法处理不合适、未进行缺失数据的处理、模型性能评估不足及评价指标不规范等有关。结论中心静脉导管相关血流感染风险预测模型尚存在一些不足,未来应进一步提高相关模型研究的质量,尤其是在预测因子选择、模型评价等方面应更加规范。

Objective To systematically evaluate the risk prediction model for central venous catheter-related bloodstream infections and provide references for clinical practice.Methods Databases such as CNKI,Wanfang,CBM,VIP,PubMed,Web of Science,Cochrane Library,etc.were retrieved.The search period is from database establishment to June 2,2023.There are 2 researchers who independently screened and extracted the literature,and evaluated the quality of the literature using bias assessment tools of predictive model risk.Results A total of 9 articles were included,including 9 risk prediction models for catheter-related bloodstream infections.The total sample size was 80~11901 cases;the number of outcome events was 19~403 cases;the C index of the included model was 0.81~0.93.The area under the curve of the subjects ranged from 0.73~0.90.The predictors that appear more frequently in the model mainly included the history of diabetes,albumin value,the number of days of catheterization,the location of catheterization.The evaluation results of the bias assessment tool of predictive model risk for research show that the overall applicability of the included risk prediction models is good,but the bias risk is high.The reasons are related to the improper source of research sample data,inappropriate processing of continuous variable methods,failure to process missing data,insufficient model performance evaluation,and non-standard evaluation indicators.Conclusion There are still some shortcomings in the risk prediction model for central venous catheter-related bloodstream infections.In the future,the quality of related model research should be further improved,especially in terms of predictive factor analysis,model evaluation indicators,etc.,which should be further standardized.