Development And Validation Of A Clinical Prediction Model For Ischemic Stroke In Patients With Nonvalvular Atrial Fibrillation

Background and Objective:Atrial fibrillation is one of the most common tachyarrhythmias and significantly increases the risk of ischemic stroke.To prevent such risks,oral anticoagulants have been widely used for stroke prevention in patients with atrial fibrillation.However,anticoagulant therapy can increase the risk of bleeding to some extent,so it is necessary to accurately predict the occurrence of ischemic stroke to determine whether to use anticoagulant therapy.At present,CHA₂DS₂-VASc scoring is the mainstream method to assess the risk of ischemic stroke in patients with nonvalvular atrial fibrillation,and it is widely used.But the score was moderate in predicting ischemic stroke risk.Therefore,this study analyzed risk factors for ischemic stroke and established an easy-to-use risk prediction model to more accurately predict the risk of ischemic stroke in patients with nonvalvular atrial fibrillation to optimize anticoagulant therapy decision making.Method:This study collected clinical data of 735 patients with atrial fibrillation who were hospitalized in the Department of Cardiology,the First Hospital of Jilin University from January 2018 to December 2018.Based on the inclusion and exclusion criteria,429 patients with nonvalvular atrial fibrillation were selected and randomly divided into a training set and a validation set at 1:1.The dependent variable of this study was whether ischemic stroke occurred,which was a binary variable.In the training set,univariate Logistic regression was used to analyze clinical history,serology and imaging data,and then statistically significant variables were applied to multivariate binary Logistic regression analysis to screen out independent risk factors and construct a clinical prediction model.The visualization of the model is carried out using a Nomogram.Receiver operating characteristic curve（ROC curve）drawing and Area under concentration time curve（AUC）calculation were used to evaluate the differentiation.Calibration curves are used to check the accuracy of the model.Finally,the effectiveness of the model in the validation set and the overall data set was verified,and the ROC curve and AUC values were used to compare the different models.The above data were processed by IBM.SPSS.Statistics 25 and R.4.2.1.Result:（1）In the training set,multivariate binary Logistic regression analysis suggested that age,ischemic stroke,non-paroxysmal AF,and BNP were independent risk factors for ischemic stroke in patients with nonvalvular atrial fibrillation.On this basis,we constructed a clinical risk prediction model named AF-ABSNp with an AUC of 0.857（95%CI,0.7955-0.918）.In addition,the calibration curve is drawn,and the results show that the model has good accuracy.（2）In the validation set and overall data set,the AUC of AF-ABSNp model was 0.763（95%CI,0.692--0.833）and 0.810（95%CI,0.763--0.857）,respectively.The calibration curve is further drawn and it is found that the model has good accuracy.（3）Compared with CHA₂DS₂-VASc,CHADS₂,HELT-E₂S₂,AF-ABSNp model has the maximum AUC value in training set,validation set and overall data set.Conclusion:（1）Age,ischemic stroke,non-paroxysmal atrial fibrillation,and BNP were independent risk factors for ischemic stroke in patients with nonvalvular atrial fibrillation.（2）AF-ABSNp,a novel risk prediction model constructed by combining clinical history including age,ischemic stroke,non-paroxysmal atrial fibrillation,and the biomarker BNP,can effectively predict ischemic stroke in patients with nonvalvular atrial fibrillation with high specificity and sensitivity.（3）Compared with CHA₂DS₂-VASc,CHADS₂,HELT-E₂S₂,AF-ABSNp model has better predictive value.