目的 基于Logistic回归和决策树探究终末期肾病患者发生轻度认知功能障碍的影响因素。方法 选取就诊于北京协和医院、北京朝阳医院及中国人民解放军总医院进行血液透析和腹膜透析的终末期肾病患者。基于多因素Logistic回归分析和决策树的CHAID算法分析终末期肾病患者发生轻度认知功能障碍的影响因素。 结果 共纳入337例患者，轻度认知功能障碍发生率为44.2%。多因素 Logistic回归分析结果显示：相比年龄≤44岁的患者，≥60岁发生轻度认知功能障碍的风险更高    (OR=5.637，95% CI：2.843～11.175，P＜0.001)；相比受教育时间6年及以下的患者，7～12年(OR=0.138，95% CI：0.023～0.827，P=0.030)和12年以上(OR=0.076，95% CI：0.013～0.454，P=0.005)发生轻度认知功能障碍的风险更低；伴有糖尿病(OR=2.009，95% CI：1.147～3.521，P =0.015)和有脑血管病史(OR =5.039，95% CI：2.281～11.129，P＜0.001)的患者发生轻度认知功能障碍的风险更高。共6个变量进入决策树模型，由上到下分别为年龄、脑血管病史、糖尿病、受教育时间、透析方式、肌酐。 结论 决策树模型与传统Logistic模型的结合应用可以科学充分地探究终末期肾病患者发生轻度认知功能障碍的影响因素。其影响因素涉及患者一般情况、合并症、治疗方式及检查指标各个方面，临床中应给予关注和重视。
【关键词】终末期肾病；轻度认知功能障碍；影响因素；Logistic回归；决策树

Objective  To analyze the influencing factors of mild cognitive impairment based on logistic regression and decision tree.  Methods   Patients were selected from three third-class and first-class hospitals in Beijing, who were diagnosed with end-stage renal disease and underwent hemodialysis or peritoneal dialysis. Based on multivariate logistic regression analysis and CHAID algorithm of decision tree, the influencing factors of mild cognitive impairment in patients with end-stage renal disease were analyzed.  Results   A total of 337 patients with end-stage renal disease were included, the incidence of mild cognitive impairment  was 44.2%. The results of multivariate logistic regression analysis showed that compared with patients aged≤44 years, the risk of mild cognitive impairment at≥60 years is higher (OR=5.637, 95% CI: 2.843～11.175, P<0.001). Compared with patients with education for 6 years or less, those with education for 7 to 12 years (OR=0.138, 95% CI: 0.023～0.827, P=0.030) and more than 12 years (OR=0.076, 95% CI: 0.013～0.454, P=0.004) have lower risk of mild cognitive impairment. Patients with diabetes (OR=2.009, 95% CI: 1.147～3.521, P=0.015) and history of cerebrovascular disease (OR=5.039, 95% CI: 2.281～11.129, P＜0.001) are at higher risk of mild cognitive impairment. A total of 6 variables entered the decision tree model, from top to bottom were age, history of cerebrovascular disease, diabetes, time of education, mode of dialysis and creatinine. Both models have high efficiency.  Conclusion   The combination of decision tree model and traditional logistic model can scientifically and fully explore the influencing factors of mild cognitive impairment in patients with end-stage renal disease. The influencing factors are related to the general condition of patients, complicated disease, treatment methods and examination indicators, which should be paid attention to in clinic.