Abstract:

Objective To explore the combination patterns of high-risk characteristics for catheter-related thrombosis（CRT） formation in patients undergoing artificial liver support therapy via femoral vein catheterization，and to develop and validate a cumulative risk prediction model，so as to provide a reference for implementing precise and anticipatory nursing care. Methods Patients who underwent artificial liver support therapy via femoral vein catheterization in a tertiary hospital in Hangzhou from January to December 2024 were consecutively included as the study population，and their relevant clinical data were collected. The Apriori algorithm was applied to mine the core high-risk association rules for CRT formation. Clustering and visualization analyses of these core rules were performed using grouping matrix clustering and network topology graphs. Based on the analytical results，a cumulative risk prediction model was constructed，and logistic regression analysis was employed to validate the predictive performance of the model. Results A total of 206 patients were included，and the incidence of CRT was 53.39%. Using the Apriori algorithm，1，712 core high-risk association rules were extracted. Cluster analysis was performed on the top 15 core rules ranked by lift，and 5 clusters of core characteristic combinations were identified. Among these，the combination of “impaired consciousness and number of artificial liver support therapy sessions ≥3” showed the highest average lift. Network topology analysis revealed that “number of artificial liver support therapy sessions ≥3”，“Caprini score of 3-4”，and “pre-treatment fibrinogen concentration <1.5 g/L” were the 3 hub nodes with higher connectivity in the network. Based on the 5 clusters，a cumulative risk prediction model was constructed. Logistic regression analysis showed that the model score was positively associated with the risk of CRT（OR=1.689，P<0.001）. Conclusion The model can effectively quantify the cumulative effect of multiple risk factors for CRT formation in patients undergoing artificial liver support therapy via femoral vein catheterization，and demonstrates a relatively high accuracy in identifying high-risk patients.

Key words: Artificial Liver, Catheter-Related Thrombosis, Association Rules, Predictive Model, Nursing Care

摘要：

目的 探讨人工肝经股静脉置管患者导管相关性血栓（catheter-related thrombosis，CRT）形成的高风险特征组合规律，构建并验证累积风险预测模型，为实施精准的预见性护理提供参考。 方法 连续纳入2024年1—12月杭州市某三级甲等医院收治的人工肝经股静脉置管患者作为调查对象，采集相关资料。运用Apriori算法挖掘CRT形成的核心高风险关联规则，通过分组矩阵聚类及网络拓扑图对核心高风险关联规则进行聚类与可视化分析，并基于分析结果构建累积风险预测模型，采用Logistic回归分析验证模型的预测效能。 结果 共纳入206例患者，CRT发生率为53.39%。运用Apriori算法，提取出1 712条核心高风险关联规则。对提升度排序前15位的核心高风险关联规则进行聚类分析，提炼出5类核心特征组合，其中有意识障碍且人工肝治疗次数≥3次的平均提升度最高。网络拓扑图分析结果显示，人工肝治疗次数≥3次、Caprini评分为3~4分、治疗前纤维蛋白原浓度<1.5 g/L是网络中连接度较高的3个枢纽节点。依据5类核心特征组合构建累积风险预测模型，Logistic回归分析结果显示，模型评分与CRT的发生风险呈正相关（OR=1.689，P<0.001）。 结论 该研究构建的模型能有效量化人工肝经股静脉置管患者CRT形成中多风险因素的叠加效应，对高风险患者的识别准确率较高。

关键词: 人工肝, 导管相关性血栓, 关联规则, 预测模型, 护理