基于决策树和随机森林算法模型预测醛固酮瘤患者术后血压恢复情况  

Prediction of postoperative blood pressure recovery in aldosterone⁃producing adenoma patients based on decision tree and random forest models

作　　者：关兆娟 薛舒婷 谷婷钰 张岩波 王彦[3] GUAN Zhaojuan;XUE Shuting;GU Tingyu;ZHANG Yanbo;WANG Yan(Department of Health Statistics,School of Public Health,Shanxi Medical University,Taiyuan 030001,China;Department of Endocrinology,First Clinical Medical College of Shanxi Medical University;Department of Endocrinology,First Hospital of Shanxi Medical University)

机构地区：[1]山西医科大学公共卫生学院卫生统计学教研室,太原030001 [2]山西医科大学第一临床医学院内分泌科 [3]山西医科大学第一医院内分泌科

出　　处：《山西医科大学学报》2025年第2期127-133,共7页Journal of Shanxi Medical University

摘　　要：目的 构建决策树和随机森林算法模型,预测醛固酮瘤患者术后血压恢复情况,评估其预测效果,并识别影响术后血压恢复的关键因素。方法 收集211例醛固酮瘤患者的基本信息,按照7∶3的比例将数据集划分为训练集和测试集。在训练集上分别使用决策树和随机森林算法模型构建醛固酮瘤患者术后血压恢复情况的预测模型,并在测试集上进行验证。通过比较两种模型的预测性能,评估其在术后血压恢复预测中的效果。结果 在211例醛固酮瘤患者中,术后血压恢复正常的患者79例,术后血压有所改善但未治愈的患者132例,术后血压治愈率为37.4%。两组患者在年龄、体质量指数、高血压病程及肾小球滤过率等方面差异有统计学意义(P<0.05)。决策树模型的预测准确率为0.75,特异度为0.82,灵敏度为0.64,AUC为0.79,F1分数为0.67;随机森林模型的预测准确率为0.81,特异度为0.87,灵敏度为0.72,AUC为0.87,F1分数为0.75,故随机森林模型的预测性能优于决策树模型。结论 随机森林模型能够更准确地预测醛固酮瘤患者术后血压的恢复情况,有效识别出年龄、体质量指数、高血压病程和肾小球滤过率等影响因素。该模型可为醛固酮瘤患者术后血压的临床治疗和个性化管理提供科学依据。Objective To construct decision tree and random forest models for predicting the postoperative blood pressure recovery in patients with aldosterone-producing adenoma,evaluate their predictive performance,and identify key factors affecting postoperative blood pressure recovery.Methods Clinical data of 211 aldosterone-producing adenoma patients were collected,and the patients were divided into training set and testing set at a ratio of 7∶3.Decision tree and random forest models were built using the training set to predict the postoperative blood pressure recovery in aldosterone-producing adenoma patients,and the models were validated using the testing set.The performance of both models was compared to assess their effectiveness in predicting postoperative blood pressure re-covery.Results Among the 211 aldosterone-producing adenoma patients,79 patients achieved normal blood pressure postopera-tively,while 132 patients showed improvement but did not fully recover.The cure rate of postoperative blood pressure was 37.4%.There were significant differences between the two groups in terms of age,body mass index,duration of hypertension,and estimated glomerular filtration rate(P<0.05).The decision tree model achieved an accuracy of 0.75,a specificity of 0.82,a sensitivity of 0.64,AUC of 0.79,and an F1 score of 0.67.The random forest model achieved an accuracy of 0.81,a specificity of 0.87,a sensitivity of 0.72,AUC of 0.87,and an F1 score of 0.75.Therefore,the random forest model was better than the decision tree model in predictive performance.Conclusion The random forest model can more accurately predict the postoperative blood pressure recovery in aldoste-rone-producing adenoma patients and effectively identify key influencing factors such as age,BMI,duration of hypertension,and eGFR.This model can provide the scientific evidence for clinical treatment and personalized management of postoperative blood pres-sure in aldosterone-producing adenoma patients.

关 键 词：醛固酮瘤 肾上腺切除术 血压 预测模型 决策树模型 随机森林算法模型 影响因素 

分 类 号：R736[医药卫生—肿瘤]