摘要
目的评估随机森林模型对肾病综合征(NS)患者5年心血管疾病风险的预测价值。方法选取陆军军医大学第一附属医院就诊的350例NS患者随访5年的诊疗资料,按照约7∶3的比例划分为训练集和测试集。模型纳入28个预测变量,通过训练集进行随机森林模型构建,测试集数据进行模型验证,选取最优节点值和决策树数目,观察变量的预测重要性并评价模型预测性能。结果随机森林模型最佳节点值为6、最佳决策树数目为446。模型中预测因子重要性排序依次为:肾小球滤过率(eGFR)、年龄、高密度脂蛋白胆固醇(HDL-C)、载脂蛋白B(apoB)、清蛋白(ALB)、载脂蛋白A1(apoA1)、纤维蛋白原(Fib)、血尿酸(UA)、低密度脂蛋白胆固醇(LDL-C)。模型预测的准确率为0.919、精确率为0.935、召回率为0.829,AUC及95%CI为0.899(0.832~0.966)。结论随机森林分类算法确定的重要预测因子可为预测NS患者5年心血管风险提供有用的信息,模型预测性能良好。
Objective To evaluate the value of the random forest model in predicting the risk of cardiovascular disease in patients with nephrotic syndrome(NS)over five years.Methods The medical data of 350 patients with NS who were followed up for five years in the hospital were collected.The patients were divided into the train set and the test set according to the ratio of nearly 7∶3.A total of 28 predictive variables were incorporated into the model,the train set was used for the random forest model construction,and the test set for model verification.The optimal node values and the number of decision-making tree were selected to observe the predictive importance of variables and evaluate the model′s predictive performance.Results The optimal node values and the number of decision-making tree of the random forest model were 6 and 446.The order of importance of predictors in this model was glomerular filtration rate(eGFR),age,high-density lipoprotein cholesterol(HDL-C),apolipoprotein B(apoB),albumins(ALB),apolipoprotein A1(apoA1),Fibrinogen(Fib),uric acid(UA),low-density lipoprotein cholesterol(LDL-C).The accuracy rate of the random forest model was 0.919,the precision rate was 0.935,the recall rate was 0.829,and the AUC and confidence interval was 0.899(0.832-0.966).Conclusion The important predictors determined by the random forest classification algorithm may provide helpful information for predicting the five-year cardiovascular risk of the NS patients.The model has good predictive performance.
作者
邹新亮
郑万香
何国祥
景涛
ZOU Xinliang;ZHENG Wanxiang;HE Guoxiang;JING Tao(Department of Cardiology,the Southwest Hospital of Army Medical University,Chongqing 400038,China;Department of Cardiology,Guiqian International General Hospital,Guiyang,Guizhou 550000,China)
出处
《重庆医学》
CAS
2022年第3期393-397,共5页
Chongqing medicine
基金
重庆市科卫联合医学科研重点项目(2022ZDXM005)
重庆市卫生适宜技术推广项目(2018jstg036)
重庆市研究生科研创新项目(CYS19371)。
关键词
心血管风险
肾病综合征
随机森林
预测模型
cardiovascular risk
nephrotic syndrome
random forest
predictive model
