摘要
目的:构建并比较Stanford A型和B型急性主动脉夹层（AAD）患者围术期导致急性肾损伤（AKI）的诺模图预测模型。方法:收集2019年1月至2021年12月在我院确诊为AAD的患者资料。采用LASSO回归和多因素logistic分别筛选出Stanford A型急性主动脉夹层AKI（TAAAD-AKI）、Stanford B型急性主动脉夹层AKI（TBAAD-AKI）的独立危险因素并构建诺模图预测模型。通过Bootstrapping内部验证，从准确性、校准度、临床效益3个角度综合评价模型优劣。结果:共收集464例AAD患者的资料，TAAAD-AKI的发生率为83.5%（147/176），TBAAD-AKI的发生率为41.0%（118/288）。用于构建TAAAD-AKI诺模图的独立危险因素有：入院首次血清肌酐（SCr）、入院D-二聚体值、体外循环（CPB）时间、机械通气（MV）时间、围术期使用升压药；构建TBAAD-AKI诺模图的独立危险因素有：入院首次SCr、入院肾脏灌注不良、ICU滞留时间、围术期使用袢利尿剂。区分度显示TAAAD-AKI模型受试者工作特征曲线（ROC）的曲线下面积（AUC）为0.899，提示准确性高；TBAAD-AKI模型AUC值为0.825，说明模型有中等准确性。绘制模型校准曲线，Hosmer-Lemeshow检验反映两种诺模图模型均有很好的校准度。而决策曲线同样发现模型均有不错的临床效益。结论:TAAAD-AKI和TBAAD-AKI的预测因子除了入院首次SCr外并不相同，TAAAD-AKI关键变量主要集中在术中术后，而TBAAD-AKI则多为术前。构建并验证可行的两种诺模图对于AKI的临床预警具有重要意义。
Abstract
Objective:To construct and compare the nomogram prediction model of perioperative acute kidney injury (AKI) in Stanford type A and type B acute aortic dissection (AAD) patients.Methods:The data of patients diagnosed with AAD in our hospital from January 2019 to December 2021 were collected. The independent risk factors of TAAAD-AKI and TBAAD-AKI were screened by LASSO regression and multi-factor logistic regression, respectively, and the nomogram prediction model was constructed. Through the internal verification of bootstrapping, the advantages and disadvantages of the model were evaluated from three aspects: accuracy, calibration, and clinical benefit.Results:Data from 464 patients with AAD were collected. The incidence of TAAAD-AKI was 83.5% (147/176), whereas TBAAD-AKI was 41.0% (118/288). First serum creatinine (SCr) on admission, D-dimer value on admission, cardiopulmonary bypass time, mechanical ventilation time, and perioperative use of pressor medications were the independent risk variables for the creation of the TAAAD-AKI nomogram. The variables screened by TBAAD-AKI were first SCr on admission, poor renal perfusion on admission, days of ICU retention, and perioperative use of loop diuretics. The area under the curve (AUC) of receiver operating characteristic curves (ROC) of TAAAD-AKI model was 0.899, which implied a high level of accuracy. The AUC value of the TBAAD-AKI model was 0.825, indicating moderate accuracy. The two nomogram models had good calibration, according to the model's calibration curve and the Hosmer-Lemeshow test. The decision-making curve also found that the model had good clinical benefits.Conclusion:The predictors of TAAAD-AKI and TBAAD-AKI are distinct except for the first SCr on admission. While the majority of the TBAAD-AKI are preoperative, the significant TAAAD-AKI variables are primarily focused during and after the procedure. Constructing and verifying two feasible nomograms is crucial for clinical early warning of AKI.
万方数据