Risk Prediction Of Short-term Poor Prognosis In Patients With Ischemic Stroke Based On Bayesian Network Model

Objective:To explore the risk factors affecting the prognosis of acute ischemic stroke(AIS),and to understand the interaction of various risk factors on short-term poor prognosis by constructing Bayesian network model,so as to provide the basis for reducing the incidence of poor prognosis.Method:A prospective cohort study was designed to investigate the demographic,medical history,laboratory tests,and clinical features of hospitalized patients with acute ischemic stroke from April 2019 to February 2021 in four Grade A hospitals in Nanchang City.Baseline data were collected using electronic medical record reporting forms,and the outcomes were followed up 90 days after discharge.Logistic regression was performed on the statistically significant variables in the univariate analysis,and risk factors with statistically significant differences were selected.Combined with professional knowledge to determine the structure of Bayesian network,the application of R software "bnlearn" package for parameter learning.The predictive effect of the model was evaluated according to sensitivity,specificity and receiver operating characteristic curve(ROC).Results:A total of 1191 effective cases were followed up in this study,including 233 patients in the poor prognosis group 90 days after discharge,the incidence of poor prognosis was 19.6%.Age,smoking,history of stroke,ischemic heart disease,mean red blood cell volume,dementia,MRI diagnosis,head CTA/MRA/DSA examination,NIHSS scale score on admission,mean red blood cell content,absolute value of granulocyte,hemoglobin,uric acid,GCS scale score,fasting blood glucose were correlated with good prognosis group in the difference,and the difference was statistically significant(P < 0.05).Logistic regression analysis showed that the risk factors for poor short-term prognosis of ischemic stroke included age(OR=1.034,95%CI:1.018-1.050),smoking(OR=0.706,95%CI:0.4955-1.008),stroke history(OR=1.622,95%CI:1.129-2.330),NIHSS(1: OR=1.551,95%CI:0.981-2.453,2:OR=3.442,95%CI:2.138-5.541),neutrophil absolute value(OR=1.456,95%CI:1.014-2.088),white bulb ratio(OR=1.582,95%CI:1.087-2.302),high density lipoprotein(OR=0.405,95%CI:0.170-0.968)and fasting blood glucose(OR=1.034,95%CI:1.018-1.050).Bayesian network model showed that age,NIHSS score,fasting blood glucose and high-density lipoprotein had a direct effect on the prognosis of patients with ischemic stroke.In terms of efficiency,the sensitivity,specificity and AUC values of the Bayesian network model were 0.662,0.812 and 0.79,respectively,while the sensitivity,specificity and AUC values of the Logistic regression model were 0.575,0.795 and 0.723,respectively.Probabilistic inference of risk factors and individualized risk prediction of short-term poor prognosis can be made through Bayesian networks.Conclusion:The risk factors affecting the short-term prognosis of AIS were age,smoking,stroke history,NIHSS score,neutrophil absolute value,white bulb ratio,high density lipoprotein and fasting blood glucose.The short-term poor prognosis prediction model constructed by Bayesian network has good prediction ability,can describe the complex network risk mechanism between risk factors and prognosis more intuitively,can conduct probabilistic reasoning on the risk factors in the network,and can predict the short-term poor prognosis risk individually,so as to better guide the medical treatment decision-making.