Development and validation of a nomogram prediction model for venous thromboembolism in trauma patients

He Lingxiao; Du Ying; Ouyang Chaowei; Wang Yiyan; Yi Min; Wang Guanglin; Liao Dengbin

doi:10.3969/j.issn.1009-4237.2025.10.011

Abstract：Objective To develop and validate a risk prediction model for venous thromboembolism(VTE)in trauma patients based on clinical risk factors,thus providing reference for precise intervention of high-risk VTE trauma patients.Methods A total of 996 hospitalized trauma patients admitted to West China Hospital of Si-chuan University from Jun.to Dec.2021 were retrospectively analyzed.Patient information and possible risk factors were collected through the hospital information system.Factors with statistically significant differences between VTE and non-VTE patients were selected for further multivariable logistic regression analysis using the Boruta method,to obtain the final model and draw a nomogram to visualize the risk.Internal validation was performed using the boot-strap method,and the discrimination,calibration,and clinical capability of the nomogram prediction model were eval-uated using the area under the curve(AUC),calibration curve,and decision curve analysis.Results Among the 996 patients,252 cases had VTE,with the incidence being 25.30%(95%CI:0.227-0.281).The final prediction model included 10 risk factors:age,body mass index,injury severity score,upper limb injury,lower limb injury,pel-vic injury,central venous catheter,blood transfusion,length of surgery,and admission to ICU.Internal validation showed that the AUC of the model was 0.796(95%CI:0.767-0.825),with a sensitivity of 0.790,and a specificity of 0.653.The Hosmer-Lemeshow test revealed that χ2=5.559 and P=0.697.The model's Brier score was0.146,and the decision curve analysis showed that the model had positive net benefit within the threshold probability range of 0.01-0.96.Conclusion This study constructs a10-factor VTE risk prediction model for trauma patients and visu-alizes it using a nomogram.The model has good discriminative ability and calibration,and a good clinical capability range,which can be clinically applied in combination with interventions in the future.