Abstract: Objective To explore the relative importance of nutritional and metabolic indicators in blood based on random forest algorithm in overweight and obesity among children aged 1 - 6 years,in order to provide reference for clinical evaluation of childhood obesity and development of nutritional plans. Methods Children aged 1-6 years who were admitted to the Children's Healthcare Department of the Daxing District Maternal and Child Health Hospital of Beijing from February 2021 to May 2022 were selected as study subjects,and were divided into normal group (n=400),overweight group (n=200),and obesity group (n=200).Blood related indicators among the three groups were compared,and multivariate Logistic regression analysis was used to explore the related factors of childhood obesity.The variable relative importance of random forest was obtained through SPSS Modeler,and the variables were sorted. Results The weight and body mass index (BMI) index of children in the obesity group were significantly higher than those in the overweight group and control group,while the weight and BMI index of children in the overweight group were significantly higher than those in the control group (P<0.05).Serum levels of vitamin D,vitamin A,high-density lipoprotein cholesterol (HDL-C) in the obesity group were significantly lower than those in the overweight group and control group (F=63.205,14.753,1.604,P<0.05),while fasting blood glucose (FBG),leptin,promelanocortin (POMC),low-density lipoprotein cholesterol (LDL-C),triglyceride (TG),serum total cholesterol (TC) levels were significantly higher than those in the overweight group and control group (F=79.923,1.637,23.177,46.202,93.157,293.033,P<0.05).Children in the overweight group had higher level of glutamic oxaloacetic transaminase (AST) than control group and obesity group (F=3.919,P=0.020).Multivariate Logistic regression analysis showed that obesity was mainly related to serum vitamin A,vitamin D,leptin,POMC,FBG,HDL-C,LDL-C,TG,cholesterol.After measuring the average Gini index of factors by 100 cycles,it was found that cholesterol,FBG,leptin,TG,LDL-C,POMC,HDL-C,serum vitamin D,and vitamin A were ranked as important factors affecting the occurrence of obesity. Conclusions The relevant factors and ranking of importance influencing childhood obesity among infants and preschool children can be used as auxiliary laboratory indicators for clinical evaluation of childhood obesity.Schools and parents should reasonably distribute micro nutrients intake and adjust the dietary structure of children to avoid an increase in the rate of childhood obesity.

Key words: infants and young children, preschool children, obesity, vitamin A, vitamin D, random forest algorithm

摘要： 目的 探索基于随机森林算法探讨血液中营养与代谢相关指标在婴幼儿及学龄前儿童超重肥胖中的相对重要性,以期为临床评估儿童肥胖和制定营养方案提供参考依据。方法 选取2021年2月-2022年5月就诊于北京市大兴区妇幼保健院儿童保健部的1～6岁儿童作为研究对象,分为正常组(n=400)、超重组(n=200)和肥胖组(n=200)。比较三组儿童血液相关指标,采用多因素Logistic回归分析1～6岁儿童肥胖的相关因素。通过SPSS Modeler 得出随机森林的变量相对重要性并进行排序。结果 肥胖组儿童体重和体质量指数(BMI)显著高于超重组和对照组,超重组儿童体重和BMI显著高于对照组(P<0.05);肥胖组儿童血清25-(OH)D、维生素A、高密度脂蛋白胆固醇水平均低于超重组和对照组(F=63.205、14.753、1.604,P＜0.05),而空腹血糖、瘦素、阿黑皮素原、低密度胆固醇、甘油三酯、胆固醇则显著高于超重组和对照组(F=79.923、1.637、23.177、46.202、93.157、293.033,P＜0.05);在谷草转氨酶水平上,超重组儿童高于对照组和肥胖组(F=3.919,P=0.020)。多因素Logistic回归结果发现,肥胖主要与血清维生素A、维生素D、瘦素、阿黑皮素原、空腹血糖、高密度脂蛋白胆固醇、低密度胆固醇、甘油三酯、胆固醇密切相关;采用平均下降Gini指数衡量因素进行100次循环后发现,影响肥胖发生的变量中其重要性依次为胆固醇、空腹血糖、瘦素、甘油三酯、低密度胆固醇、阿黑皮素原、高密度脂蛋白胆固醇、血清维生素D、维生素A。结论 影响1～6岁儿童肥胖的相关因素及其重要性排序,可作为临床评估儿童肥胖的辅助实验室指标。学校及家长应合理分配宏量营养素摄入,适当调整儿童膳食结构,避免儿童肥胖率增长。

关键词: 婴幼儿, 学龄前儿童, 肥胖, 维生素A, 维生素D, 随机森林算法