Identifying and preliminary validating patient clusters in coronary artery bypass grafting integrating autonomic function with clinical and demographic data for personalized care

Abstract Aims: This study aims to identify distinct clusters of patients undergoing coronary artery bypass grafting (CABG) based on demographic, clinical, and autonomic function characteristics and to validate these clusters. Methods and results: Our cohort study included 154 subjects aged 18 years and older undergoing CABG, enrolled in Italy, from April 2017 to January 2020. Data were prospectively collected from pre-anaesthesia induction to hospital discharge. Clustering was performed using t-distributed stochastic neighbour embedding (t-SNE) on 23 variables and hierarchical clustering, including pre- and post-anaesthesia autonomic function indices and demographic and clinical data. Two distinct clusters were identified: 'higher risk-responsive group' and 'lower risk-responsive group'. The higher risk-responsive group cluster consisted of older patients with higher co-morbidity rates and worse autonomic function. Validation of clusters through multiple correspondence analysis and Poisson regression demonstrated significant differences in post-operative outcomes. Patients in the lower risk-responsive group cluster had fewer complications (IRR = 0.441, P = 0.004). The analysis indicated that intensive care unit (ICU) stay duration and the power of systolic arterial pressure (SAP) series in low-frequency band derived in the post-anaesthesia phase were significant predictors of complications above and beyond the expected contributions of age and comorbidities, with longer ICU stays and lower low-frequency power of SAP post-anaesthesia induction being associated with higher complication rates. Conclusion: Integrating autonomic function measures and demographic and clinical data could enhance patient monitoring and intervention, improving outcomes if included in future risk stratification tools and early warning score systems. Registration: ClinicalTrials.gov: NCT03169608. Keywords: Cardiovascular nursing; Clustering analysis; Coronary artery bypass grafting; Hierarchical clustering; Patient stratification; Post-operative complications.摘要 研究目的：本研究旨在基于人口统计学、临床及自主神经功能特征，对接受冠状动脉旁路移植术（CABG）的患者进行聚类分型，并对该聚类结果予以验证。 方法与结果：本队列研究于2017年4月至2020年1月间在意大利纳入154名年满18岁、接受冠状动脉旁路移植术的受试者，相关数据自麻醉诱导前至患者出院均前瞻性收集。研究采用t分布邻域嵌入（t-SNE）结合层次聚类法，基于23项变量（涵盖麻醉前后自主神经功能指标、人口统计学及临床数据）开展聚类分析，最终得到两个特征明确的聚类组：“高风险应答组”与“低风险应答组”。其中高风险应答组包含年龄更大、合并症发生率更高、自主神经功能状态更差的患者。通过多重对应分析与泊松回归对聚类结果进行验证，结果显示两组术后结局存在显著差异：低风险应答组患者的并发症发生率更低（IRR = 0.441，P = 0.004）。进一步分析表明，在年龄与合并症的既定影响之外，重症监护病房（ICU）停留时长以及麻醉诱导后收缩动脉压（SAP）序列的低频波段功率，是并发症发生的独立预测因子；更长的ICU停留时间与麻醉诱导后SAP低频功率更低均与更高的并发症发生率相关。 结论：将自主神经功能测量指标与人口统计学、临床数据相结合，可优化患者监测与干预方案；若将该整合模型应用于未来的风险分层工具与早期预警评分系统，有望改善患者术后结局。 注册信息：临床试验注册平台（ClinicalTrials.gov）：NCT03169608。 关键词：心血管护理；聚类分析；冠状动脉旁路移植术；层次聚类；患者分层；术后并发症。