中心静脉置管患者导管相关性血栓风险预测模型的系统评价

目的 系统检索和评价中心静脉导管相关性血栓的风险预测模型，以期为构建更高质量的中心静脉导管相关性血栓的风险预测模型提供参考。方法 检索中国知网、万方数据库、中国生物医学文献数据库、PubMed、Embase、Web of Science、Cochrane Library、CINAHL中中心静脉导管相关性血栓的风险预测模型的相关研究，语种限定为中文和英文，检索时限为建库至2022年2月16日。由2名研究者独立筛选文献和提取数据，并应用预测模型研究的偏倚风险评估工具分析纳入文献的偏倚风险和适用性。结果 共纳入15项中心静脉导管相关性血栓的风险预测模型构建研究，共涉及16个模型。16个模型的受试者工作特征曲线下面积为0.641~0.850。12项研究的适用性较好，其余3项研究的适用性较差。15项研究的偏倚风险均较高，偏倚主要来自未选择合适的数据来源、研究对象的纳入和排除标准欠妥当、预测因子未完全从结局指标中被排除、样本量不足、连续性变量处理方式不合理、对缺失数据关注不足、变量筛选过程不合理、缺乏模型性能评估以及模型过度拟合等。结论 现有中心静脉导管相关性血栓的风险预测模型构建尚不完善，需要在后期的构建中关注对不同风险评估方法有效性的研究，以得到更好的高准确度的风险预测模型，为制订相关预防策略提供一定的参考和依据。

Systematic review of risk prediction models for catheter-related thrombosis in patients undergoing central venous catheterization

Objective To systematically analyze and evaluate the prediction models for central venous catheter-associated thrombosis，in order to provide references for constructing higher quality risk prediction models for central venous catheter-associated thrombosis. Methods Original articles in Chinese and English were systematically searched from Chinese databases（CNKI，Wanfang，Sinomed） and English databases（PubMed，Embase，Web of Science，Cochrane Library，CINAHL），and the search period was from the database establishment to February 16，2022.There were 2 investigators who independently screened the literature and extracted data，and applied PROBAST，a risk of bias assessment tool for predictive modeling studies，to analyze the risk of bias and applicability of the included literature. Results A total of 15 studies were included to construct risk prediction models for central venous catheter-associated thrombosis, involving a total of 16 models. 156 models had an area under the subject operating characteristic curve ranging from 0.641 to 0.85. 12 studies had good applicability，while the remaining 3 studies had poor applicability. 15 studies had a high risk of bias，mainly due to failure to select appropriate data sources，inappropriate inclusion and exclusion criteria for study subjects，insufficient sample size，inadequate treatment of missing data，unreasonable variable screening process， consideration of competing risk factors，lack of model performance.The risk of bias was high in all 15 studies，mainly due to inappropriate selection of data sources，inappropriate inclusion and exclusion criteria, incomplete exclusion of predictors from outcome indicators，insufficient sample size，unreasonable treatment of continuous variables，insufficient attention to missing data，unreasonable variable screening process，consideration of competing risk factors，lack of model performance evaluation and over-simulation of the model，etc. Conclusion The construction of existing risk prediction models for central venous catheter-associated thrombosis is still in its initial stage.It is necessary to focus on the study of the validity of different risk assessment methods in the later construction in order to obtain better and highly accurate risk prediction models and provide certain references and bases for the development of relevant prevention strategies.