基于LASSO回归的心脏神经官能症刚虚证(肝肾阴亏、肝阳上亢证)关联因素筛选及诊断模型构建  

Study on the State Identification Model of Cardiac Neurosis Patients with Rigid-Deficiency Symptom Based on LASSO Regression

作　　者：王瑞婷 杨曜嘉 张慧[2] 原晨 柳红良 李娅[1] 王韵涵 赵鹏[1,5] WANG Ruiting;YANG Yaojia;ZHANG Hui;YUAN Chen;LIU Hongliang;LI Ya;WANG Yunhan;ZHAO Peng(Dongzhimen Hospital,Beijing University of Chinese Medicine,Beijing 100700,China;Guang'anmen Hospital,China Academy of Chinese Medical Sciences,Beijing 100053,China;Shunyi Hospital,Beijing Traditional Chinese Medicine Hospital,Beijing 101399,China;Aviation General Hospital,Beijing 100012,China;Luoyang Hospital of Chinese Medicine,Luoyang Henan 471099,China)

机构地区：[1]北京中医药大学东直门医院,北京100700 [2]中国中医科学院广安门医院,北京100053 [3]北京中医医院顺义医院,北京101399 [4]航空总医院,北京100012 [5]洛阳市中医院,河南洛阳471099

出　　处：《中医药导报》2024年第7期71-76,共6页Guiding Journal of Traditional Chinese Medicine and Pharmacy

基　　金：“十二五”国家科技支撑项目(2013BAI02B09);河南省中医药科学研究专项课题(2023ZY2174);北京市级中医药专家学术经验继承人(第六批)。

摘　　要：目的:利用LASSO回归联合Nomogram构建心脏神经官能症刚虚证(肝肾阴亏、肝阳上亢证)的诊断模型。方法:采用单中心前瞻性研究,收集141例心脏神经官能症患者的临床资料,纳入分析变量包括年龄、民族、婚姻、教育程度、脑力/体力工作、体质量指数(BMI)、中医症状、中医五态人格量表各因子得分、汉密尔顿焦虑量表-14项、汉密尔顿抑郁量表-24项、症状自测评量表各因子均分。通过LASSO回归筛选与刚虚证诊断显著相关的影响因素,纳入二元多因素Logistic回归分析构建诊断模型,并对模型预测区分度及校准度评价,利用10重交叉验证进行内部验证,最后对模型进行Nomogram可视化,并根据ROC曲线确定诊断阈值。结果:共纳入刚虚证患者70例,非刚虚证患者71例。LASSO回归筛选出与刚虚证诊断相关性最显著的5个变量为女性、年龄、疲乏无力、善嗳气、五态人格中少阳积分。模型AUC为0.85,H-L检验为2.94(P=0.9824),提示模型区分度及校准度较好,10重交叉内部验证结果提示AUC为0.82。结论:LASSO回归联合Nomogram构建的诊断模型可协助诊断心脏神经官能症刚虚证,但研究结果的准确性尚待大样本临床研究进一步验证。Objective:To construct a diagnostic model for cardiac neurosis with rigid-deficiency symptoms using LASSO regression combined with Nomogram.Methods:A single-center prospective study was conducted to collect the clinical data of 141 patients with cardiac neurosis.Variables included in the analysis were age,ethnicity,marriage,education,mental/physical work,BMI,traditional Chinese medicine symptoms,and the scores of Chinese medicine five-state personality,the Hamilton Anxiety Inventory-14-item scale,the Hamilton Depression Inventory-24-item scale and the Symptom Checklist-90.The influencing factors significantly associated with the diagnosis of rigid-deficiency symptoms were screened by LASSO regression,incorporated into a binary multi-factor Logistic regression analysis to construct a diagnostic model,and the model was evaluated for predictive discrimination and calibration,internally validated using 10-fold cross-validation.Finally,the model was visualized by Nomogram,and the diagnostic threshold was determined according to the ROC curve.Results:A total of 70 patients with rigid deficiency syndrome and 71 patients with non-rigid deficiency syndrome were included.The LASSO regression screened the five most significant variables associated with the diagnosis of rigid-deficiency symptoms as female,age,fatigue,belching,and Shaoyang points in the five personality states.The model AUC was 0.85 and the H-L test was 2.94(P=0.9824),suggesting good model differentiation and calibration,and the 10-fold cross-over internal validation results suggested an AUC of 0.82.Conclusion:In this study,the diagnostic model constructed by LASSO regression combined with Nomogram can help clinicians to rapidly diagnose patients with cardiac neurosis with rigid-deficiency symptoms,but the accuracy of the results needs to be further verified by large-sample clinical studies.

关 键 词：心脏神经官能症 刚虚证 刚柔辨证 临床诊断模型 LASSO回归 列线图 

分 类 号：R241.6[医药卫生—中医诊断学]