基础医学与临床 ›› 2022, Vol. 42 ›› Issue (8): 1169-1175.doi: 10.16352/j.issn.1001-6325.2022.08.1169

• 研究论文 •    下一篇

基于中医证候和机器学习构建慢性心力衰竭中西医结合预后模型

樊佳赛1,2, 杜艺菲1,2, 许佳颖1,2, 陈思臻1,2, 高永桧1,2, 任景怡2*   

  1. 1.北京中医药大学 临床医学院,北京 100105;
    2.中日友好医院 心脏科 心衰中心,北京 100029
  • 收稿日期:2022-06-03 修回日期:2022-06-13 出版日期:2022-08-05 发布日期:2022-08-01
  • 通讯作者: *renjingyi1213@hotmail.com
  • 基金资助:
    国家自然科学基金面上项目(81770359);北京市卫生健康科技成果和适宜技术推广项目(BHTPP202004);中日友好医院“菁英计划”人才培育工程(ZRJY2021-BJ01)

Construction of a TCM and Western combination model for prognostic evaluation of chronic heart failure based on TCM syndrome elements and machine learning

FAN Jia-sai1,2, DU Yi-fei1,2, XU Jia-ying1,2, CHEN Si-zhen1,2, GAO Yong-hui1,2, REN Jing-yi2*   

  1. 1. School of Clinical Medicine, Beijing University of Chinese Medicine, Beijing 100105;
    2. Heart Failure Center, Department of Cardiology, China-Japan Friendship Hospital, Beijing 100029, China
  • Received:2022-06-03 Revised:2022-06-13 Online:2022-08-05 Published:2022-08-01
  • Contact: *renjingyi1213@hotmail.com

摘要: 目的 应用机器学习方法探讨中医证候要素对慢性心力衰竭(CHF)预后模型的意义,并建立以中医证候要素为基础的中西医结合CHF预后模型。方法 纳入2018年1月1日至2021年4月30日收入中日友好医院心内科住院治疗的CHF患者,收集患者人口统计学资料、共病、检验检查和中医证候要素等信息。主要终点为1年内因心力衰竭再住院或心血管死亡复合事件的发生。采用适用于高维数据筛选的最小绝对收缩和选择算子(LASSO)方法从数据集中选择最有用的预测变量,通过Cox多因素回归分析确定最终的独立危险因素并建立可视化列线图。结果 本研究最终纳入164人,平均年龄(72.23±14.16)岁,男性占37.2%。运用LASSO机器学习方法从临床变量中共筛选出9个因素,包括冠心病、高血压、尿酸、N末端B型利钠肽原(NT-proBNP)、左心室射血分数(LVEF)、肌酸激酶同工酶MB、肌红蛋白、气虚和阴虚。经Cox多因素回归分析后发现,NT-proBNP、LVEF、气虚、高血压和冠心病5个因素与CHF患者预后独立相关。结论 中医证候要素气虚是CHF患者1年内因心力衰竭再住院或心血管死亡事件的独立预测因子;中西医结合CHF预后模型具有较高的准确性。

关键词: 慢性心力衰竭, 证候要素, 中西医结合, 机器学习, 预后模型

Abstract: Objective To construct a prognostic model of chronic heart failure (CHF) by traditional Chinese medicine (TCM) syndrome elements by machine learning method. Methods Patients with CHF admitted to the Department of Cardiology of China-Japan Friendship Hospital from January 1, 2018 to April 30, 2021 were included, and their demographic data, vital signs, co-morbidities, laboratory tests, echo-cardiographic indicators, TCM syndrome elements and treatment information were collected. The primary end point for this analysis was a model to predict cardiovascular death or hospitalization because of heart failure in one year follow-up. Least absolute shrinkage, selection operator(LASSO) regression and Cox multivariate analysis were used to screen independent risk factors that potentially affect the prognosis of CHF. A nomogram was used to establish a risk prediction model based on TCM syndrome elements. Results Totally 164 patients with an average age of (72.23±14.16) years old and 37.2% male were included in this study. The LASSO screened 9 factors from clinical variables, including coronary heart disease, hypertension, uric acid, N-terminal pro-B type natriuretic peptide (NT-proBNP), left ventricular ejection fraction (LVEF), creatine kinase-myocardial band, myoglobin, Qi deficiency and Yin deficiency. Cox multivariate regression analysis showed that Qi deficiency, hypertension, coronary heart disease, NT-proBNP and LVEF were associated with prognosis in patients with CHF. Conclusions Qi deficiency was an independent predictor of cardiovascular death or heart failure readmission in CHF patients within 1 year. The prognostic model of CHF with integrated Chinese and Western medicine has demonstrated a high accuracy.

Key words: chronic heart failure, syndrome elements, integrated Chinese and Western medicines, machine learning, prognostic model

中图分类号: