Construction Of Prediction Model Of Ischemic Stroke Recurrence Based On XGboost Algorithm

Objective: To investigate the recurrence of patients with ischemic stroke within 12 months after the first stroke;to identify important influencing factors of recurrence and to construct a prediction model based on the XGboost algorithm for predicting recurrence after ischemic stroke,in order to provide a predictive tool for identifying patients at high risk of recurrence after ischemic stroke in clinical work.Methods: Based on literature review of risk factors for recurrence after ischemic stroke,from June 1,2017 to June 30,2019,524 patients with first-ever ischenic stroke hospitalized in the Neurology Department were enrolled and followed up 12 months in The fifth of Affiliated Hospital of Zunyi Medical University,Zhu Hai,and data about risk factors of recurrence was retrospectively collected in this study,in order to investigate the recurrence of patients within 3 months,6 months and 12 months after stroke,respectively;SPSS21.0was used for univariate and multivariate analysis to explore factors affecting the recurrence within 3 months,6 months and 12 months after the first ischemic stroke,respectively;XGboost algorithm was used to construct a classification model for predicting recurrence after ischemic stroke.Results: 1.Among the 524 patients with first-ever ischemic stroke included,11 cases(2.1%)had recurrences within 3 months after stroke,33 cases(6.3%)had recurrences within6 months after stroke,and 59 cases(11.3%)had recurrences within 12 months after stroke.2.The results of univariate analysis showed that the recurrence within 3 months after stroke of patients with first-ever ischemic stroke was related to hospitalized systolic blood pressure and fibrinogen(P<0.05);the recurrence within 6 months after stroke was related to smoking,carotid atherosclerosis,and hospitalized diastolic blood pressure,Uric acid,apolipoprotein A1,glycosylated hemoglobin(P<0.05);the recurrence within 12 months after stroke of patients was related to smoking,drinking,hypertension,carotid atherosclerosis,admission to the hospital microcomputer blood sugar,glycosylated hemoglobin(P<0.05);multivariate analysis showed that the independent influencing factor for the recurrence within 3 months after the first ischemic stroke was the level of admission systolic blood pressure(OR=1.036,95%CI=1.002~1.072,P=0.038),and smoking was an independent factor for the recurrence within 6 months(OR=2.453,95%CI=1.015~5.931,P=0.046)and the recurrence within 12 months stroke(OR=2.960,95%CI=1.328~6.595,P=0.008).Glycated hemoglobin(OR=1.435,95%CI=1.071~1.924,P=0.016)is an independent factor of recurrence within 12 months after stroke.The logistic regression model of the data processed by the SMOTE algorithm showed that age(60~74years old:OR=2.552,95%CI=1.364~4.775,P=0.003;≥75 years old:OR=2.134,95%CI=1.073~4.245,P=0.031),hospital stays(OR=1.792,95%CI=1.129~2.847,P=0.013),smoking(OR=3.319,95%CI=1.809~6.090,P<0.001),drinking(OR=1.982,95%CI=1.118~3.513,P=0.019),hypertension(OR=2.442,95%CI=1.404~4.246,P=0.002),glycosylated hemoglobin(OR=1.330,95%CI=1.072~1.651,P=0.009)and carotid artery atherosclerosis(OR=5.648,95%CI=1.623~19.659,P=0.007)are independent risk factors for recurrence within 12 months after the onset of ischemic stroke.In the XGboost model,the top seven variables of the importance of influencing factors are neutrophil percentile,fibrinogen,homocysteine,total cholesterol,uric acid,hospitalized microcomputer blood glucose,and glycosylated hemoglobin.3.Prediction model based on the XGboost algorithm exceeds the traditional Logistic regression model in predicting recurrence after first ischemic stroke.The accuracy,precision,sensitivity,specificity and AUC value of the XGboost model for predicting recurrence within 12 months after first ischemic stroke are 0.97,0.91,1.0,0.95,0.97,respectively.XGboost model has a strong ability to discriminate patients wtih high-risk recurrence after ischemic stroke.Conclusion: 1.The recurrence rate within 12 months after the onset of ischemic stroke was 11.3%,which was within the range of previous research reports.2.There are similarities and differences in the distribution of ischemic stroke recurrence factors in the time dimension.Patients with high admission systolic blood pressure are at high risk of recurrence within 3 months after ischemic stroke.Smokers are at high risk of recurrence within 12 months after ischemic stroke.High glycosylated hemoglobin is at high risk of recurrence within 12 months after stroke.Compared with Logistic regression model,XGboost model can identify important influencing factors that are easy to be ignored.The percentage of neutrophils,fibrinogen,homocysteine,total cholesterol,uric acid,hospitalized microcomputer blood glucose and glycosylated hemoglobin have an important impact on the recurrence of ischemic stroke.In the process of clinical nursing,it is necessary to focus these indicators and implement corresponding intervention measures.3.The model based on the XGboost algorithm for predicting recurrence after first ischemic stroke has a strong ability to identify patients with high-risk recurrence,and XGboost model is better than the traditional Logistic regression model.In the future,we can further consider the prospective inclusion of a large sample of clinical cases to verify the effect of the model and gradually optimize the model,and then develop the system for clinical practice applications,so as to provide practical guidance for identifying patients with high risk of recurrence after ischemic stroke in clinical work.