论文题名(中文): | 静脉-动脉体外膜氧合支持患者急性肾损伤预测模型的构建与验证 |
姓名: | |
论文语种: | chi |
学位: | 硕士 |
学位类型: | 专业学位 |
学位授予单位: | 北京协和医学院 |
学校: | 北京协和医学院 |
院系: | |
专业: | |
指导教师姓名: | |
论文完成日期: | 2025-03-28 |
论文题名(外文): | The Construction and Validation of a Prediction Model for Acute Kidney Injury in Patients Supported by Venous-Arterial Extracorporeal Membrane Oxygenation |
关键词(中文): | |
关键词(外文): | Venous-arterial extracorporeal membrane oxygenation (V-A ECMO) Acute kidney injury Risk factors Prediction model |
论文文摘(中文): |
研究背景 静脉-动脉体外膜氧合(venous-arterial extracorporeal membrane oxygenation, V-A ECMO)是严重心肺功能衰竭患者的重要支持手段,但也常伴一些并发症的 发生,其中急性肾损伤(acute kidney injury, AKI)较为常见,尤其是 3 期 AKI 与不良预后密切相关。影响 3 期 AKI 发生的因素繁多复杂,且目前针对 V-A ECMO 患者 3 期 AKI 的预测模型研究有限,有必要对其危险因素进行分析研究, 识别高风险患者,以实现早期干预和精准治疗。 研究目的 分析 V-A ECMO 支持患者的临床资料,识别与 V-A ECMO 支持的成年患者 3 期 AKI 相关的独立危险因素,进一步建立 3 期 AKI 的列线图预测模型并进行 验证,用于早期识别高危患者和改善预后。 研究方法 本研究为一项回顾性研究,收集了 2016 年 1 月到 2024 年 8 月期间在中国医 学科学院阜外医院接受 V-A ECMO 支持的 192 例成年患者的临床资料。按照患 者术后是否发生 3 期 AKI 分为 3 期 AKI 组与非 3 期 AKI 组,采用最小绝对收缩 选择算子(least absolute shrinkage and selection operator, LASSO)回归分析选择预 测因子,使用多因素 Logistic 回归分析方法筛选出 3 期 AKI 的独立影响因素,并 通过列线图使预测模型可视化。模型的内部验证采用 Bootstrap 重抽样法,受试 者工作特征曲线(receiver operating characteristic Curve, ROC)及其对应的曲线下 面积(area under the curve, AUC)用于量化模型对 3 期 AKI 的区分能力。同时,校 准曲线通过对比预测概率与实际发生概率的一致性,评估模型的校准性能;临床 决策曲线从风险阈值角度分析模型的净获益,以明确其临床应用价值。 研究结果 1、在接受 V-A ECMO 的 192 例患者中,3 期 AKI 组 114 例(59.4%),非 3 期 AKI 组 78 例(40.6%)。 2、经 LASSO 回归、多因素 Logistic 回归分析显示 ECMO 置入前乳酸水平 (OR=1.116,95%CI:1.037~1.210)、ECMO 置入前血红蛋白水平(OR=0.985,95%CI: 0.972~0.997) 、ECMO 前 SOFA 评分(OR=1.898,95%CI:1.560~2.3868)及初始 ECMO 转数(OR=1.002,95%CI:1.001~1.004)是 3 期 AKI 发生的独立危险因素。 3、在构建预测模型时,将上述四个独立影响因素纳入其中,并借助列线图 实现其可视化。通过对建模人群数据应用 Bootstrap 重抽样法,进行 1000 次重复 抽样后,计算得出一致性指数 (concordance index, C-index)为 0.881。通过绘制 ROC 曲线,计算模型的 AUC 为 0.897(95%CI:0.852-0.941),对应于 82.5%的 敏感性、79.5%的特异性以及 81.25%的准确性,该预测模型具有良好的区分性能; 校准曲线表明模型的预测结果与实际观察结果基本一致,校准度良好;临床决策 曲线分析证实了该模型在临床应用中的有效性。 研究结论 1、ECMO 置入前乳酸水平、ECMO 置入前血红蛋白水平、ECMO 前 SOFA 评分以及初始ECMO转数是V-A ECMO支持患者发生3期AKI的独立影响因素。 2、基于 ECMO 置入前乳酸水平、ECMO 置入前血红蛋白水平、ECMO 前 SOFA 评分以及初始 ECMO 转数建立的预测模型具有良好的预测能力。 3、对于接受 V-A ECMO 支持的患者,基于预测模型绘制的列线图可以帮助 临床医师早期识别潜在的 3 期 AKI 患者,从而为早期干预提供支持。 |
论文文摘(外文): |
Background Venous-arterial extracorporeal membrane oxygenation (V-A ECMO) is a critical support modality for patients with severe cardiopulmonary failure but is frequently complicated by acute kidney injury (AKI), particularly stage 3 AKI, which is strongly associated with poor prognosis. The risk factors for stage 3 AKI are complex and multifactorial, yet limited studies have focused on predictive models for stage 3 AKI in V-A ECMO patients. Identifying high-risk patients through risk factor analysis is essential for early intervention and precision therapy. Objective To analyze clinical data from V-A ECMO-supported patients, identify independent risk factors for stage 3 AKI in adults receiving V-A ECMO, and establish and validate a nomogram prediction model to facilitate early risk stratification and improve outcomes. Methods Clinical data from 192 adult patients who underwent V-A ECMO at Fuwai Hospital, Chinese Academy of Medical Sciences, between January 2016 and August 2024 were retrospectively analyzed. Patients were categorized into stage 3 AKI (n=114, 59.4%) and non-stage 3 AKI (n=78, 40.6%) groups. Predictors were selected using LASSO regression, followed by multivariate logistic regression to identify independent risk factors. A nomogram was constructed for visualization. Internal validation was performed using 1,000 bootstrap resamples. Model discrimination was evaluated via ROC curve analysis (AUC), calibration curves assessed accuracy, and decision curve analysis (DCA) determined clinical utility. Results 1.Multivariate analysis identified four independent risk factors for stage 3 AKI: pre-ECMO lactate levels (OR=1.116, 95% CI: 1.037–1.210), pre-ECMO hemoglobin levels (OR=0.985, 95% CI: 0.972–0.997), pre-ECMO SOFA score (OR=1.898, 95% CI: 1.560–2.386), and ECMO initial speed (OR=1.002, 95% CI: 1.001–1.004). 2.The nomogram incorporating these factors demonstrated excellent discrimination (AUC=0.897, 95%CI:0.852–0.941; sensitivity=82.5%, specificity=79.5%, accuracy= 81.25%)and a concordance index (C-index) of 0.881. Calibration curves indicated strong agreement between predicted and observed probabilities, and DCA confirme favorable clinical utility. Conclusion 1. Pre-ECMO hyperlactatemia, anemia, elevated SOFA score, and higher ECMO initial speed are independent risk factors for stage 3 AKI in V-A ECMO patients. 2. The nomogram prediction model based on these factors exhibits robust predictive performance. 3. This model enables clinicians to early identify high-risk patients, guiding timely interventions to mitigate AKI progression and improve prognosis |
开放日期: | 2025-06-06 |