Development and validation of a risk prediction model for femoral artery hematoma following cerebrovascular intervention

WANG Juan; HAO Pei; HE Yangmin; ZHANG Hongchen; XU Suiyun; LI Xia; LU Chuanhao

doi:10.3639/j.issn.1671-2897.2025.04.007

Abstract：Objective To investigate risk factors for hematoma at the femoral artery puncture site after cerebrovascular intervention and to construct and validate a risk prediction model.Methods A retrospective analysis was conducted on the clinical data of 743 patients who underwent cerebrovascular intervention at the Department of Neurosurgery,the First Affiliated Hospital of the Air Force Medical University,from July 2018 to July 2023.These patients were divided into hematoma and non-hematoma groups based on the occurrence of post-operative hematoma.Factors with P<0.05 in univariate analysis were included in logistic regression to construct a hematoma risk prediction model.The model's discrimination and calibration were evaluated using the area under the ROC curve and Hosmer-Lemeshow goodness-of-fit test.Internal validation was performed using the Bootstrap method,and external validation involved 258 patients from the Second Affiliated Hospital of Xi'an Jiaotong University from June 2019 to July 2023.Results Alcohol history,hypertension history,BMI,puncture site,and intraoperative anticoagulant use were independent risk factors.The prediction model formula is Y=-4.487+2.284(alcohol history)+0.984(hypertension history≤5 years)+1.465(hypertension history>5 years)+1.234(BMI>24)+2.617(bilateral puncture site)+2.819(1 anticoagulant)+2.815(≥2 anticoagulants).The area under the ROC curve was 0.881,with sensitivity 0.769,specificity 0.85,and Youden index 0.619.Hosmer-Lemeshow test P=0.374(χ2=8.633)indicated good predictive value.Internal validation showed AUC=0.887(95%CI:0.849-0.925),sensitivity 0.938,specificity 0.638,Hosmer-Lemeshow P=0.791(χ2=4.678).External validation showed AUC 0.921(95%CI:0.847-0.995),sensitivity 0.902,specificity 0.769,Hosmer-Lemeshow P=0.198(χ2=8.590).Calibration curves indicated an average absolute error<0.05 for both internal and external validation.Conclusion The prediction model based on alcohol history,hypertension history,BMI,puncture site,and intraoperative anticoagulant use demonstrates good performance and can facilitate early identification of hematoma at puncture sites post-cerebrovascular intervention,providing a reference for reducing hematoma incidence after femoral artery intervention.