基于随机森林算法的糖尿病肾病患者血液透析中发生低血压预测模型的建立  

Establishment of a Prediction Model for Hypotension in Hemodialysis Patients with Diabetic Nephropathy Based on Random Forest Algorithm

作　　者：傅华珍[1] 丁小容[1] 陈丹 姜蕾[1] 任慧 宋长帅 阳新星 隋红霞 FU Huazhen;DING Xiaorong;CHEN Dan(Department of Blood Purification Room in Nephrology of Peking University Shenzhen Hospital,Shenzhen,518035)

机构地区：[1]北京大学深圳医院肾内科血液净化室,深圳518035 [2]中国科学院大学深圳医院血液净化室,深圳518106 [3]深圳百霖透析中心,深圳518000

出　　处：《中国中西医结合肾病杂志》2023年第6期493-496,共4页Chinese Journal of Integrated Traditional and Western Nephrology

基　　金：深圳市三名工程基金资助项目(No.SZSM201812097)。

摘　　要：目的:建立预测糖尿病肾病患者血液透析中发生低血压的随机森林模型。方法:选取2020年01月—2021年12月进入我市三个透析中心进行血液透析的200例糖尿病肾病患者作为研究对象,采用Logistic回归分析筛选糖尿病肾病患者血液透析中发生低血压的危险因素,采用R软件建立预测糖尿病肾病患者血液透析中发生低血压的随机森林模型。结果:Logistic回归分析结果显示,年龄≥70岁、贫血、低白蛋白血症、身高<160 cm、透析液温度≥36℃、超滤量≥5%、干体重不合理及透析龄>5年等是糖尿病肾病患者血液透析中发生低血压的危险因素(P<0.05)。随机森林模型结果显示,影响糖尿病肾病患者血液透析中发生低血压的因素重要性排序依次低白蛋白血症、身高、透析龄、年龄、干体重、透析液温度、超滤量及贫血。随机森林模型预测糖尿病肾病患者血液透析中发生低血压的AUC为0.834。结论:低白蛋白血症、身高<160 cm、透析龄>5年、年龄≥70岁、干体重不合理、透析液温度≥36℃、超滤量≥5%及贫血等是糖尿病肾病患者血液透析中发生低血压的危险因素,随机森林模型有助于临床简单快速进行糖尿病肾病患者血液透析中发生低血压风险分析和及早制定相关的防治措施。Objective:To establish a random forest model for predicting hypotension in hemodialysis patients with diabetic nephropathy.Methods:A total of 200 diabetic nephropathy patients who entered three dialysis centers in our city for hemodialysis from January 2020 to December 2021 were selected as the research subjects.Logistic regression analysis was used to screen the risk factors of hypotension during hemodialysis in diabetic nephropathy patients.R The software established a random forest model to predict the occurrence of hypotension in hemodialysis patients with diabetic nephropathy.Results:Logistic regression analysis showed that age≥70 years old,anemia,hypoalbuminemia,height<160 cm,dialysate temperature≥36℃,ultrafiltration volume≥5%,unreasonable dry body weight,and dialysis age>5 years were diabetes mellitus Risk factors of hypotension during hemodialysis in patients with renal disease(P<0.05).The results of random forest model showed that the factors affecting the occurrence of hypotension during hemodialysis in diabetic nephropathy patients were ranked in order of importance in order of hypoalbuminemia,height,dialysis age,age,dry weight,dialysate temperature,ultrafiltration volume and anemia.The random forest model predicted the AUC of hypotension in hemodialysis patients with diabetic nephropathy to be 0.834.Conclusion:Hypoalbuminemia,height<160 cm,dialysis age>5 years,age≥70 years,unreasonable dry weight,dialysate temperature≥36℃,ultrafiltration volume≥5%,and anemia are common causes of hemodialysis in patients with diabetic nephropathy.The risk factors of hypotension and random forest model are helpful for simple and rapid clinical analysis of the risk of hypotension during hemodialysis in patients with diabetic nephropathy and early development of relevant preventive measures.

关 键 词：糖尿病肾病 血液透析 低血压 危险因素 随机森林模型 预测模型 

分 类 号：R587.2[医药卫生—内分泌] R692.9[医药卫生—内科学] R544.2[医药卫生—临床医学]