Comparison of two models in predicting the risk of thrombosis in elderly patients with CHF complicated with lower respiratory tract infection
Ji Miaomiao
Li Chuanbo
Wang Yuekun
Xu Yong
Abstract：Objective To compare the value of Logistic regression model and XGBoost model in predicting the risk of thrombosis in elderly patients with CHF complicated with LRTI.Methods A total of 138 elderly CHF patients with LRTI admitted to our department from April 2019 to April 2024 were prospectively enrolled,and divided into thrombus group(43 cases)and non-thrombus group(95 cases)according to whether thrombosis occurred.Clinical data of these pa-tients were collected,and two risk prediction models of thrombosis in these patients were con-structed based on logistic regression and XGBoost regression,respectively.The predictive value was compared between the two models.Results The thrombus group had higher neutrophil count,NLR,and CRP,D-D and vWF levels,and increased MA,estimated dissolution percentage,and percentage LY30,but K and R and lower coagulation index than the non-thrombus group(P＜0.01).NLR,CRP,D-D,vWF,LY30,K and R were influencing factors for thrombosis in the elderly CHF patients with LRTI(P＜0.05,P＜0.01).The AUC value of the multivariate logistic regression model and XGBoost model in predicting thrombosis in the patients was 0.915(95％CI:0.861-0.986)and 0.894(95％CI:0.841-0.971),respectively,with a sensitivity of 85.40％and 88.90％and a specificity of 96.50％and 82.30％,respectively.There was no statistical difference in AUC value between the two models(Z=0.573,P=0.678).Hosmer Lemeshow test showed the differences were not significant in the calibration curves of the multivariate logistic regression model and XGBoost model(x2=0.485,P=0.452;x2=0.669,P=0.335).Conclusion Multivari-ate logistic regression model and XGBoost model show equivalent efficacy in predicting thrombo-sis in CHF patients with LRTI.Abnormal levels of NLR,CRP,D-D,vWF,LY30,K,and R are im-portant factors affecting thrombosis in these elderly patients.
Keywords：agedheart failurerespiratory tract infectionsthrombosislogistic modelsforecasting
Publication Date：2025-07-15
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：5( 890-894 )