基于Logistic回归与决策树模型的老年食管癌 患者术后认知功能障碍影响因素的探究

Analysis of Influencing Factors for Postoperative Cognitive Dysfunction in Elderly Patients with Esophageal Cancer Based on Logistic Regression and Decision Tree Algorithms

摘要: 目的：探讨老年食管癌患者术后认知功能障碍(postoperative cognitive dysfunction, POCD)的影响因素，为临床防治提供参考依据。方法：回顾性选取2020年1月至2025年1月就诊于康复大学青岛中心医院并接受食管癌切除术的老年患者为研究对象。本研究运用多因素Logistic回归分析与决策树模型，系统性探究老年食管癌患者术后认知功能障碍发生的相关影响因素。结果：共纳入符合标准的患者161例，术后POCD发生率为21.7%。多因素Logistic回归分析显示，年龄 ≥ 60岁(OR = 1.57, 95%CI: 1.007~2.453)、身体质量指数(BMI) < 18.5 kg/m² (OR = 2.60, 95%CI: 1.437~4.692)、既往合并脑血管疾病病史(OR = 3.20, 95%CI: 1.415~7.25)、术前白细胞计数(WBC) > 10 × 10⁹/L (OR = 6.51, 95%CI: 1.76~24.109)、术前血红蛋白(Hb) < 120 g/L (OR = 2.06, 95%CI: 1.161~3.649)、手术过程失血量 > 400 mL (OR = 4.19, 95%CI: 1.375~12.791)、手术持续时间 > 8 h (OR = 5.88, 95%CI: 1.914~18.092)，均为老年食管癌患者术后认知功能障碍(POCD)发生的独立风险因子，各项指标的差异均具有统计学意义(均P < 0.05)。决策树模型分析显示，年龄、脑血管疾病史、术前WBC、术前Hb4个变量成功进入模型。结论：将决策树模型与传统Logistic回归模型联合应用，可更科学、全面地筛选老年食管癌患者术后POCD的影响因素。这些影响因素涵盖患者一般状况、合并症、诊疗相关指标等多个维度，需引起临床高度关注，为制定针对性防治策略提供理论支撑。

Abstract: Objective: To explore the influencing factors of postoperative cognitive dysfunction (POCD) in elderly patients with esophageal cancer and provide a reference for clinical prevention and treatment. Methods: Elderly patients who were admitted to Qingdao Central Hospital, University of Health and Rehabilitation Sciences (Qingdao Central Hospital) and underwent esophagectomy from January 2020 to January 2025 were retrospectively selected as the research objects. A combined application of multivariate Logistic regression analysis and the decision tree algorithm was adopted to conduct an in-depth, systematic exploration of the factors relevant to the occurrence of mild postoperative cognitive dysfunction (POCD) in elderly patients who had undergone esophageal cancer surgery. Results: A total of 161 eligible patients were included, with an incidence of postoperative POCD of 21.7%. Multivariate Logistic regression analysis showed that age ≥ 60 years old (OR = 1.57, 95%CI: 1.007~2.453), body mass index (BMI) < 18.5 kg/m² (OR = 2.60, 95%CI: 1.437~4.692), history of cerebrovascular disease (OR = 3.20, 95%CI: 1.415~7.25), preoperative white blood cell count (WBC) > 10 × 10⁹/L (OR = 6.51, 95%CI: 1.76~24.109), preoperative hemoglobin (Hb) < 120 g/L (OR = 2.06, 95%CI: 1.161~3.649), intraoperative blood loss > 400 mL (OR = 4.19, 95%CI: 1.375~12.791), and operation time > 8 hours (OR = 5.88, 95%CI: 1.914~18.092) were independent risk factors for POCD in elderly patients after esophagectomy (all P < 0.05). Decision tree model analysis showed that 4 variables, including age, history of cerebrovascular disease, preoperative WBC, and preoperative Hb, successfully entered the model. Conclusion: The integrated use of the decision tree model and conventional Logistic regression model enables a more scientific and thorough identification of the factors contributing to postoperative cognitive dysfunction (POCD) in elderly patients who have undergone esophageal cancer surgery. These influencing factors cover multiple dimensions such as the patient’s general condition, comorbidities, and diagnosis and treatment-related indicators, which need to attract great clinical attention and provide theoretical support for formulating targeted prevention and treatment strategies.