Construction Of A Risk Assessment Model For Venous Thromboembolism In Stroke Patients

ObjectivesTo construct a risk assessment model for stroke patients with venous thromboembolism(VTE), provide a specialized risk assessment model for stroke patients with VTE,and provide guidance for the prevention and management of VTE risk.Methods1.A research group was established to retrospectively analyze the literature related to venous thromboembolism in stroke published in China and abroad before December 31,2021,screen out the relevant risk factors,combining the risk factors of the Padua Risk Assessment Model(RAM),and initially compile risk factor questionnaire entries and conduct expert correspondence after group discussion.2.Clinical and laboratory data and Padua risk scores of stroke patients from May 1,2021 to April 30,2022 were retrospectively collected from a hospital,and all data collected were statistically analyzed to determine independent predictors of venous thromboembolism formation in stroke using univariate and multifactorial logistic regression analysis.The included risk factors were analyzed by SPSS and R software and the nomogram of the risk of venous thromboembolism in stroke were plotted.Conduct internal validation using Bootstrap resampling method,ROC curves were plotted to evaluate the predictive performance of the model and compared with the predictive performance of the Padua RAM,and the nomogram calibration was evaluated by Hosmer-Lemeshow test.Draw a DCA diagram to evaluate the clinical benefits of the model.Results1.Through literature search and two rounds of correspondence from 20 experts,the preliminary forms include ongoing hormone treatment,diabetes,age≥65,coronary heart disease,hyperlipidemia, platelet count,deep vein catheterization(≤1 month),hypertension,BMI≥25kg/m~2,atrial fibrillation,heart failure and/or respiratory failure,active cancer,NIHSS score at admission,fibrinogen,acute infection and/or rheumatic disease,recent trauma or surgery(≤1 month)Acute myocardial infarction,activity of daily living,D-dimer,bed rest or immobilization time≥72 hours,VTE history,and other 21 risk factors related to VTE in stroke.The authority coefficients of the two rounds of expert correspondence were 0.830and 0.870,respectively,with positive coefficients were 100%,and the Kendall’s W ranged from 0.375 to0.448 for the two rounds of expert correspondence,with a statistically significant difference(P<0.05).2.Univariate analysis showed that factors such as age≥65 years old,previous VTE history, acute infection and/or rheumatic disease,acute myocardial infarction,D-dimer,NIHSS score on admission,ADL,deep venous catheterization(≤1 month),smoking,and other factors had statistical significance with the risk of VTE(P<0.05).BMI≥25Kg/m~2,active cancer,hypertension,atrial fibrillation,diabetes,heart failure and/or respiratory failure,hyperlipidemia,coronary heart disease,platelet count,fibrinogen,bed rest or braking≥72h,recent trauma or surgery(≤1 month),ongoing hormone therapy and VTE risk were not statistically significant(P>0.05).3.Multifactorial logistic regression analysis showed a statistically significant difference between age≥65 years,previous history of VTE,ADL,D-dimer and risk of VTE(P<0.05).The above four factors were independent risk factors for the formation of VTE in stroke patients.The logistic regression equation modeled was Logit(P)=-3.699+1.065(age≥65 years)+3.045(previous history of VTE)+1.019(D-dimer)+0.429(ADL),whereby the nomogram were plotted.4.The AUC of the stroke VTE risk prediction model was 0.781(95%CI:0.712-0.851),After1000 resamples using the Bootstrap method,the AUC was 0.741,when the maximum Youden index was0.554,the sensitivity was 76.39%,the specificity 67.35%,and the predictive accuracy was 75.31%,it shows that the model has good recognition ability.In contrast,the AUC of Padua RAM was 0.596(95%CI:0.516-0.675),when the maximum Youden index was 0.147,the sensitivity was 48.98%,the specificity66.94%,and the predictive accuracy was 64.79%,the prediction performance is lower than the VTE risk prediction model.The Hosmer-Lemeshow test showed that P=0.993(>0.05),indicating that there was no significant difference between the predicted results of the new model and the real results.ConclusionTwenty-one risk factors associated with VTE in stroke were screened by literature review and combined with the Padua risk assessment model and expert correspondence.A nomogram of the model for assessing the risk of venous thromboembolism in stroke patients was drawn by univariate and multifactorial regression analysis and tested for predictive efficiency and goodness of fit.he model has high sensitivity and specificity,good calibration curve fit,and high predictive efficacy,and higher than the Padua RAM,which can be better applied to clinical practice,but large sample validation is still needed.The DCA chart shows that the net income of the model is good.