基于XGBoost模型的缺血性脑卒中患者中医药治疗复发预测研究

Prediction of recurrence in ischemic stroke patients treated with traditional Chinese medicine based on XGBoost model

目的:研究中医药康复治疗干预下,发现复发的风险因子及预测模型,为首发IS患者的复发预防干预提供参考。方法:纳入2019年3月至2022年6月我院缺血性脑卒中患者485例,分为复发组和未复发组。比较两组患者各项指标差异,分别构建复发风险预测模型,筛选出预测性能最好的模型。结果:6个月内是否复发的预测结果,LR模型预测准确率最低,XGBoost模型预测准确率最高。六种模型中XGBoost模型在6个月和12个月内是否复发预测均性能最优。XGBoost模型变量特征重要性分析结果显示,6个月内预测入院NIHSS评分、总胆固醇、D-二聚体、高密度脂蛋白、中性粒细胞比是前五位风险因子。12个月内预测居民类型、颈内动脉狭窄度、肥胖、年龄、总胆固醇是前五位风险因子;6个月内与12个月内复发风险因子存在一定差异。结论:中医康复治疗下,XGBoost模型对6个月内和12个月内是否复发均有较好的预测性能,可用于首发IS患者复发风险预测,为首发IS患者复发预防干预提供参考。

Objective:To find the risk factors and prediction models of recurrence under the intervention of rehabilitation therapy of traditional Chinese medicine(TCM),and to provide reference for the prevention and intervention of recurrence in the first-episode IS patients.Methods:From March 2019 to June 2022,485 patients with ischemic stroke were divided into recurrent group and non-recurrent group.To compare the difference of each index between the two groups of patients,to construct the recurrence risk prediction model,and to screen out the model with the best prediction performance.Results:The LR model had the lowest prediction accuracy and the XGBOOST model had the highest prediction accuracy for recurrence within 6 months.Of the six models,the XGBOOST model performed best in predicting recurrence at both June and 12 months.In the XGBOOST model,NIHSS score,total cholesterol,d-dimer,high-density lipoprotein,and neutrophil were the top five risk factors for predicting admission within 6 months.Type of residents,internal carotid artery stenosis,obesity,age and total cholesterol were the top five risk factors within 12 months;There were some differences in recurrence risk factors between 6 months and 12 months.Conclusion:XGBOOST model can predict whether the patients will relapse within 6 months and 12 months,which can be used to predict the relapse risk of the first-episode IS patients.