目的：构建基于多维度指标的预测模型，评估早期非小细胞肺癌（NSCLC）患者胸腔镜肺叶切除术后心肺并发症的风险。

方法：第一部分：回顾性分析2023年1月—2023年9月我院600例接受胸腔镜肺叶切除术的早期NSCLC患者的临床资料。第二部分：回顾性分析2024年5月—2024年10月我院260例接受胸腔镜肺叶切除术的早期NSCLC患者的临床资料。收集患者病历资料，将第一部分资料作为训练集（n=600），第二部分资料作为验证集（n=260）。根据术后围术期是否发生肺不张、肺栓塞、呼吸衰竭、乳糜胸、大量胸腔积液、低氧血症、心房颤动等并发症，将训练集患者分为并发症组（84例）和无并发症组（516例）。训练集用于模型训练及参数调整,验证集用于模型效果评价。采用多因素Logistic回归分析影响患者术后心肺并发症的危险因素并构建预测模型，通过Nomogram可视化，对模型进行评价及验证，评估模型的临床决策价值。

结果：多因素Logistic回归分析显示，年龄、慢性阻塞性肺疾病（COPD）病史、Karnofsky功能状态（KPS）评分、第一秒用力呼气容积占预计值百分比（FEV1%pred）、一氧化碳弥散量占预计值百分比（DLCO%pred）、术中出血量及单肺通气时间是术后心肺并发症的独立影响因素。对独立影响因素进行列线图构建，结果显示ROC曲线的评估结果在训练集以及验证集均表现出较好的预测能力，校准曲线及DCA曲线分析也显示出本研究所构建的模型具有较好的预测能力。

结论：早期NSCLC患者胸腔镜术后心肺并发症的风险预测模型具有良好的预测性能。

Objective: To develop a multivariable prediction model for assessing the risk of cardiopulmonary complications after thoracoscopic lobectomy in patients with early-stage non-small cell lung cancer (NSCLC).

Methods: ‌Part 1:‌ Retrospective analysis was performed on the clinical data of 600 patients with early-stage non-small cell lung cancer (NSCLC) who underwent thoracoscopic lobectomy at our hospital between January and September 2023.‌ Part 2:‌ Retrospective analysis was performed on the clinical data of 260 patients with early-stage NSCLC who underwent thoracoscopic lobectomy at our hospital between May and October 2024.Patient medical records were collected. The data from Part 1 were designated as the training set (n=600), and the data from Part 2 were designated as the validation set (n=260).Based on the occurrence of postoperative perioperative complications such as atelectasis, pulmonary embolism, respiratory failure, chylothorax, massive pleural effusion, hypoxemia, or atrial fibrillation, patients in the training set were divided into a complication group (84 cases) and a no-complication group (516 cases).The training set was used for model training and parameter adjustment, while the validation set was used to evaluate the model's performance. Multivariate logistic regression analysis was employed to identify risk factors influencing postoperative cardiopulmonary complications and to construct a prediction model. The model was visualized using a nomogram, subsequently evaluated and validated, and its clinical decision value was assessed.

Results: ‌Multivariate logistic regression analysis‌ revealed that ‌age, history of chronic obstructive pulmonary disease (COPD), Karnofsky Performance Status (KPS) score, percentage of predicted forced expiratory volume in one second (FEV1% pred), percentage of predicted diffusing capacity of the lung for carbon monoxide (DLCO% pred), intraoperative blood loss, and one-lung ventilation time‌ were ‌independent risk factors‌ for postoperative cardiopulmonary complications. A ‌nomogram‌ was constructed based on these independent risk factors. ‌Evaluation using Receiver Operating Characteristic (ROC) curves‌ demonstrated ‌favorable predictive performance‌ of the nomogram in both the ‌training set and the validation set‌. Furthermore, analyses using ‌calibration curves and Decision Curve Analysis (DCA) curves‌ consistently indicated that the prediction model developed in this study possessed ‌good predictive capability‌.

Conclusion: The multivariable prediction model for cardiopulmonary complications after thoracoscopic lobectomy in early-stage NSCLC patients exhibits strong predictive performance.