目的 构建及验证脑出血患者静脉血栓栓塞症（venous thromboembolism，VTE）风险预测模型，并与Caprini评分表比较预测性能。 方法 建模组回顾性选取2008年—2019年美国重症监护医学信息数据库Ⅳ中的脑出血患者，利用单因素和多因素分析确定危险因素，构建并评价预测模型。采用Bootstrap抽样法对模型进行内部验证。验证组回顾性选取2016年—2024年中国某三级甲等医院的脑出血患者，进行外部验证，并与Caprini评分表比较。 结果 ICU入住时长（OR=1.064）、淋巴细胞绝对计数（OR=0.711）及BMI（OR=1.932）是脑出血患者发生VTE的独立危险因素。建模组受试者操作特征曲线下面积（area under curve，AUC）为0.767，灵敏度为0.718，特异度为0.684。Hosmer-Lemeshow检验结果显示，χ²=3.075，P=0.930。验证组模型的AUC为0.785，Caprini评分表的AUC为0.608。 结论 该预测模型具有良好的预测性能，可为临床医护人员早期识别脑出血患者中的VTE高危人群提供理论依据。

Objective To develop and validate a risk prediction model for venous thromboembolism（VTE） in patients with intracerebral hemorrhage（ICH） and compare its predictive ability with Caprini score. Methods The modeling group retrospectively selected ICH patients from the Medical Information Mart for Intensive Care-IV （MIMIC-IV） database（2008—2019）. Risk factors were identified using univariate and logistic regression analyses，and a risk prediction model was established. Internal validation was conducted using the bootstrap resampling method. The validation group retrospectively selected ICH patients in a tertiary hospital in China from 2016 to 2024 for external validation and compared it with the Caprini score. Results ICU length of stay（OR=1.064），absolute lymphocyte count（OR=0.711），and BMI（OR=1.932） were identified as the three independent risk factors for VTE in ICH patients. In the modeling group，the model demonstrated an AUC of 0.767，with a sensitivity of 0.718 and specificity of 0.684. The Hosmer-Lemeshow test indicated good calibration（χ²=3.075，P= 0.930）. In the external validation group，the model achieved an AUC of 0.785，AUC of the Caprini score was 0.608. Conclusion This prediction model has good predictive performance and can provide theoretical basis for clinical medical staff to early identify high-risk VTE populations in ICH patients.