Abstract:

Objective To construct and validate a prediction model for bleeding complications at the puncture site in liver cancer patients after percutaneous femoral artery interventional therapy，providing a reference for healthcare professionals to early identify and prevent the occurrence of such complications. Methods Using a prospective study design，915 liver cancer patients undergoing percutaneous femoral artery interventional therapy were consecutively enrolled from a tertiary hospital in Shanghai between June and December 2024 as a modeling cohort. An external validation cohort comprising 582 liver cancer patients from 3 tertiary hospitals in Shanghai and Beijing was prospectively enrolled between March and May 2025. Characteristic predictors were screened using Lasso-Logistic regression to construct a Logistic regression model and visualized as a Nomogram. Model performance was evaluated using receiver operating characteristic curves，Hosmer-Lemeshow tests，and decision curve analysis to assess discrimination，calibration，and clinical utility. Internal validation was performed via the bootstrap method. Results Based on Lasso-Logistic regression，5 predictors were selected for model construction：thrombocytopenia，postoperative hypertension，total bilirubin level，number of vascular punctures，and compression method. The prediction model demonstrated an area under the receiver operating characteristic curve of 0.760（95%CI：0.719~0.800），with a sensitivity of 66.9%，a specificity of 74.6%，and an accuracy of 82.8%. The calibration curve，Hosmer-Lemeshow test（P=0.818），and decision curve analysis all indicated favorable goodness-of-fit and clinical utility of the model. Conclusion The constructed predictive model effectively identifies the risk of puncture site bleeding complications in liver cancer patients，demonstrating robust generalizability and reliability. It provides a clinical basis for healthcare providers to recognize high-risk patients and implement personalized，precision preemptive care strategies.

Key words: Hepatoma, Interventional Therapy, Bleeding, Prediction Model, Nomogram, Nursing Care

摘要：

目的 构建并验证肝癌患者经股动脉介入治疗术后穿刺处出血并发症预测模型，为医护人员早期识别及预防患者穿刺处出血并发症提供参考。 方法 采用前瞻性研究，便利选取2024年6—12月上海市某三级甲等专科医院收治的915例经股动脉行介入治疗的肝癌患者作为建模组，2025年3—5月上海市及北京市3所三级甲等医院收治的582例肝癌患者为验证组。采用Lasso-Logistic回归筛选特征性预测因子后构建Logistic预测模型，以列线图形式呈现。应用受试者操作特征曲线下面积、Hosmer-Lemeshow检验及临床决策曲线对模型进行区分度、校准度及临床实用性评价。采用Bootstrap法进行内部验证。 结果 基于Lasso-Logistic回归，筛选出血小板减少症、术后高血压、总胆红素、血管穿刺次数及压迫方式5个预测因子进行模型构建。预测模型的受试者操作特征曲线下面积为0.760（95%CI：0.719~0.800），灵敏度为66.9%，特异度为74.6%，准确率为82.8%，校准曲线、Hosmer-Lemeshow检验（P=0.818）及临床决策曲线均显示，该模型的拟合优度及临床实用性良好。 结论 构建的预测模型可有效预测肝癌患者穿刺处出血并发症的发生，具有良好的可靠性，可为医护人员识别高危患者并实施个性化、精准化预见性护理策略提供依据。

关键词: 肝肿瘤, 介入治疗, 出血, 预测模型, 列线图, 护理