摘要目的:基于Cox比例风险回归模型构建Stanford A型主动脉夹层（AAD）死亡风险预测模型。方法:纳入2019年1月1日至2020年4月30日华中科技大学同济医学院附属同济医院经急诊科就诊转入心脏大血管外科确诊并手术治疗的AAD患者，收集患者的一般情况、临床表现、院前资料、实验室检查及影像学检查结果等，观察期到患者死亡或2021年4月30日。按照7∶3的比例将患者分为建模组和验证组。用Lasso法从建模组数据中筛选出预后预测变量，并纳入多因素Cox回归分析，构建AAD死亡风险预测模型，并采用列线图展示。绘制受试者工作特征曲线（ROC曲线）评价模型的区分度，校准曲线评价模型的准确度，临床决策曲线（DCA）评价模型的有效性。结果:最终纳入454例AAD患者，病死率为19.4%（88/454）。用Lasso回归分析从建模组317例患者数据中最终筛选出10个变量，并构建死亡风险预测模型：0.511×腹痛+1.061×晕厥+0.428×下肢疼痛/麻木-0.365×急诊入院-1.933×直接入科-1.493×转诊前确诊+0.662×术前收缩压（SBP）<100 mmHg（1 mmHg＝0.133 kPa）+0.632×超敏心肌肌钙蛋白I（hs-cTnI）>34.2 ng/L+1.402×De BakeyⅠ型+0.641×肺部感染+1.472×术后谵妄。AAD死亡风险预测模型的ROC曲线下面积（AUC）和95%可信区间（95% CI）为0.873（0.817~0.928），验证组AUC和95% CI为0.828（0.740~0.916）。DCA显示模型的净获益值较高；校准曲线显示，实际观察结果与模型预测结果有很好的相关性。 结论:基于腹痛、晕厥、下肢疼痛/麻木、入院方式、转诊前确诊、术前SBP<100 mmHg、hs-cTnI>34.2 ng/L、De BakeyⅠ型、肺部感染及术后谵妄构建的AAD死亡风险预测模型，能有效帮助临床医师识别AAD高风险患者，评估患者术后生存情况，并及时调整治疗策略。

abstractsObjective:To construct the prediction model of death risk of Stanford type A aortic dissection (AAD) based on Cox proportional risk regression model.Methods:AAD patients who were diagnosed and received surgical treatment admitted to the department of cardiothoracic surgery of Tongji Hospital, Tongji Medical College of Huazhong University of Science and Technology from January 1st, 2019 to April 30th, 2020 were enrolled. The general situation, clinical manifestations, pre-hospital data, laboratory examination and imaging examination results of the patients were collected. The observation period was up to the death of the patients or ended on April 30th, 2021. They were divided into the model group and the verification group according to the ratio of 7∶3. Lasso method was used to screen prognostic variables from the data of the modeling group, and multivariate Cox regression analysis was included to construct the AAD death risk prediction model, which was displayed by nomogram. The receiver operator characteristic curve (ROC curve) was used to evaluate the discrimination of the model, the calibration curve to evaluate the accuracy of the model, and the clinical decision curve (DCA) to evaluate the effectiveness of the model.Results:A totel of 454 patients with AAD were finally included, and the mortality was 19.4% (88/454). Lasso regression analysis was used to screen out 10 variables from the data of 317 patients in the model group, and the prediction model of death risk was constructed: 0.511×abdominal pain+1.061×syncope+0.428×lower limb pain/numbness-0.365×emergency admission-1.933×direct admission-1.493×diagnosis before referral+0.662×preoperative systolic blood pressure (SBP) < 100 mmHg (1 mmHg = 0.133 kPa)+0.632×hypersensitivity cardiac troponin I (hs-cTnI) > 34.2 ng/L+1.402×De Bakey type+0.641× pulmonary infection+1.472×postoperative delirium. The area under the ROC curve (AUC) and 95% confidence interval (95% CI) of the AAD death risk prediction model were 0.873 (0.817-0.928), and that of the verification group was 0.828 (0.740-0.916). DCA showed that the net benefit value of the model was higher. The calibration curve showed that there was a good correlation between the actual observation results and the model prediction results. Conclusion:The AAD death risk prediction model based on abdominal pain, syncope, lower limb pain/numbness, mode of admission, diagnosis before referral, preoperative SBP < 100 mmHg, hs-cTnI > 34.2 ng/L, De Bakey type , pulmonary infection, and postoperative delirium can effectively help clinicians identify patients at high risk for AAD, evaluate their postoperative survival and timely adjust treatment strategies.