基于决策树算法的慢性心力衰竭患者中西医预后模型构建及验证

Construction and Validation of Traditional Chinese and Western Medicine Prognosis Model for Patients with Chronic Heart Failure Based on Decision Tree Algorithm

目的运用决策树算法构建慢性心力衰竭患者的中西医预后模型并验证其预测性能。方法运用回顾性队列研究方法,选择山东中医药大学附属医院2017年1月至2019年12月符合研究标准的慢性心力衰竭住院患者作为研究对象,收集人口学信息、实验室指标及中医证候等预后信息,通过病案调查或随访获得患者复合终点事件发生情况。应用单因素Logistic回归分析、共线性诊断并结合医理确定最终的建模变量,将纳入病例按照94的比例随机分为训练集和测试集,运用训练集数据构建决策树模型,测试集数据计算准确度、灵敏度及特异度进行模型性能评估。结果(1)共纳入患者383例,发生复合终点事件者56例(14.6%),未发生复合终点事件327例(85.4%);(2)经单因素Logistic回归分析、共线性诊断并结合医理,最终确定年龄、收缩压、心率、N末端B型利尿钠肽原(NT-pro BNP)、血红蛋白、D-二聚体、冠心病病史、高血压型心脏病病史、气虚血瘀证、阳气亏虚血瘀证等共25项建模变量;(3)建立了包含高血压型心脏病病史、收缩压、NT-pro BNP及阳气亏虚血瘀证在内的决策树预后模型,经验证模型的分类准确度为85.0%,灵敏度为75.0%,特异度为86.5%。结论基于中西医预后因素并采用决策树算法,对于构建慢性心力衰竭患者的中西医预后模型具有较好的适用性,能够为中医院慢性心力衰竭患者的预后评价提供参考。

Objective The decision tree algorithm was used to build a traditional Chinese and western medicine prognosis model for patients with chronic heart failure and verify its predictive performance.Methods Using retrospective cohort study,we selected hospitalized patients with chronic heart failure who met the research criteria from January 2017 to December 2019 in the Affiliated Hospital of Shandong University of Traditional Chinese Medicine as the research objects,collected demographic information,laboratory indicators,TCM syndromes and other prognostic information,and obtained the occurrence of patients composite endpoint events through case investigation or follow-up.Single factor Logistic regression analysis,collinearity diagnosis and medical science were used to determine the final modeling variables.The included cases were randomly divided into training set and test set according to the proportion of 94.The training set data was used to build a decision tree model,and the test set data was used to calculate the accuracy,sensitivity and specificity for model performance evaluation.Results 1.A total of 383 patients were included.56 cases(14.6%)had composite endpoint events,and 327 cases(85.4%)had no composite endpoint events;2.After univariate logistic regression analysis,collinearity diagnosis and medical theory,25 modeling variables were finally determined,such as age,systolic blood pressure,heart rate,NT-pro BNP,hemoglobin,D-dimer,history of coronary heart disease,history of hypertensive heart disease,Qi deficiency and blood stasis syndrome,Yang Qi deficiency and blood stasis syndrome.3.A decision tree prognostic model was established,including history of hypertensive heart disease,systolic blood pressure,NT-pro BNP and yang qi deficiency and blood stasis syndrome,which was verified that the classification accuracy,sensitivity and specificity of the model were 85.0%,75.0%and 86.5%.Conclusion Based on the prognostic factors of traditional Chinese medicine and western medicine,and using the decision tree algorithm,it has good applicability for constructing the traditional Chinese medicine and western medicine prognosis model of patients with chronic heart failure,and can provide a reference for the prognosis evaluation of patients with chronic heart failure in traditional Chinese medicine hospitals.