整合心脏超声参数的PCI术后再入院风险列线图模型构建

To Construct a Risk Nomogram Model for Readmission after PCI by Integrating Cardiac Ultrasound Parameters

摘要: 目的：探讨经皮冠状动脉介入治疗(PCI)术后1年心血管再入院的危险因素，构建整合心脏超声参数的多维度列线图模型，并明确其临床定位与适用范围。方法：回顾性分析2013年1月~2024年5月青岛大学附属医院7125例PCI患者的临床资料。以术后1年内因心血管相关原因再入院为结局，采用多重插补法处理缺失数据。经LASSO回归筛选变量后按7:3分为训练集(n = 4988)与验证集(n = 2137)，构建多因素Logistic回归列线图。采用ROC曲线、校准曲线及决策曲线分析(DCA)评估模型效能。结果：全组再入院率为26.0%。多因素分析显示，住院次数、Gensini评分、病变节段数、二尖瓣中度及以上反流、总胆固醇、空腹血糖及肌酐为独立危险因素(均P < 0.05)。模型验证AUC 0.665 (95% CI: 0.639~0.691)，提示区分度中等；校准曲线贴合理想对角线(Hosmer-Lemeshow检验P = 0.178)，显示校准度良好；DCA证实其在5%~80%阈概率范围内具有正向临床净获益。结论：本研究首次整合临床–解剖–代谢–心脏结构多维度指标构建列线图模型，其核心创新在于纳入二尖瓣反流、室壁变薄等超声参数以反映心脏结构重构。模型校准性能优异，可作为辅助筛查工具识别PCI术后再入院高危患者，但区分度有限，不支持个体精准预测，临床应用需结合医师经验综合判断，推广前需经多中心前瞻性外部验证。

Abstract: Objective: This study aims to investigate the risk factors associated with cardiovascular readmission within one year following percutaneous coronary intervention (PCI) and to develop a comprehensive nomogram model that incorporates echocardiographic parameters, thereby elucidating its clinical relevance and applicability. Methods: A retrospective analysis was conducted on the clinical data of 7125 patients who underwent PCI at the Affiliated Hospital of Qingdao University between January 2013 and May 2024. Multiple imputation techniques were employed to address missing data, focusing on the outcome of readmission due to cardiovascular-related causes within one year post-surgery. Variables were selected using LASSO regression and subsequently divided into a training set (n = 4988) and a validation set (n = 2137) in a 7:3 ratio. A multivariate logistic regression nomogram was then developed. The model’s performance was assessed using the receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA). Results: The readmission rate was observed to be 26.0%. Multivariate analysis identified the number of hospitalizations, Gensini score, number of diseased segments, moderate or greater mitral regurgitation, total cholesterol, fasting blood glucose, and creatinine levels as independent risk factors, all with statistical significance (P < 0.05). The area under the curve (AUC) for model validation was 0.665 (95% CI: 0.639~0.691), suggesting moderate discriminative ability. The calibration curve closely aligned with the ideal diagonal, as indicated by the Hosmer-Lemeshow test (P = 0.178), demonstrating good calibration. Decision curve analysis (DCA) confirmed a positive net clinical benefit, with threshold probabilities ranging from 5% to 80%. Conclusion: This study represents the inaugural effort to integrate clinical, anatomical, metabolic, and cardiac structural parameters in the development of a nomogram model. A key innovation of this model is the incorporation of echocardiographic parameters, specifically mitral regurgitation and ventricular wall thinning, to effectively capture cardiac structural remodeling. The model demonstrates excellent calibration performance and holds potential as an auxiliary screening tool for identifying high-risk patients for readmission following percutaneous coronary intervention (PCI). However, its discriminatory power is limited, precluding precise individual predictions. Clinical application should be guided by physicians’ expertise, and multi-center prospective external validation is necessary prior to broader implementation.