基于决策树法预测肿瘤患者PICC置管后上肢静脉血栓风险研究

Prediction Risk of Upper Limb Venous Thrombosis after PICC Catheterization in Cancer Patients Using Decision Tree Model

目的探讨决策树模型对肿瘤患者PICC置管后预测血栓风险的应用价值。方法回顾性分析178例肿瘤PICC置管患者临床资料,采用CHAID决策树分析构建PICC置管后上肢静脉血栓风险因素决策树模型,并验证模型效应。结果决策树生长4层,共9个生长节点,终末节点5个,筛选出4类血栓高危人群:未使用抗凝剂的肿瘤患者;未使用抗凝剂,合并甘油三酯高于正常的患者;未使用抗凝剂,甘油三酯正常或偏低,但NRS2002营养评分≥3分的患者;未使用抗凝剂,甘油三酯正常或偏低,NRS2002营养评分<3分,无手术史。决策树模型ROC曲线下面积为0.909,敏感度87.32%,特异度96.23%。结论决策树模型对预测肿瘤患者PICC置管后上肢静脉血栓形成风险简便易行,可用于上肢静脉血栓风险高危人群的筛选。

Objective To explore the application value of decision tree model in predicting the risk of thrombosis in cancer patients after PICC catheterization.Methods The clinical data of 178 cancer patients with PICC were retrospectively analyzed.CHAID decision tree analysis was used to construct a decision tree model of risk factors for upper limb venous thrombosis after PICC,and the model effect was verified.Results The decision tree consisted of 4 layers,9 growth nodes and 5 terminal nodes.Four types of high-risk groups of thrombosis were screened out:cancer patients without anticoagulant use;patients without anticoagulants but with higher triglyceride than normal;patients without anticoagulant use,with normal or low triglyceride but NRS2002 nutritional score≥3;patients without anticoagulant use,with normal or low triglyceride but NRS2002 nutritional score<3,and no history of surgery.The area under the ROC curve of the decision tree model was 0.909,the sensitivity was 87.32%,and the specificity was 96.23%.Conclusion The decision tree model is simple and feasible to predict the risk of upper extremity venous thrombosis in cancer patients after PICC catheterization,and can be used to screen high-risk groups of upper limb venous thrombosis.