创伤性颅脑损伤患者凝血功能障碍风险预测模型的构建及验证

Construction and validation of risk prediction model for coagulation disorder in patients with traumatic brain injury

目的探讨创伤性颅脑损伤(TBI)患者凝血功能障碍的危险因素,建立风险预测模型,为早期识别及干预措施提供理论基础。方法回顾性分析2020年1月至2023年12月华中科技大学协和深圳医院急诊科入院的428例TBI患者的临床资料,将2020年1月至2022年12月321例TBI患者数据作为训练集,2023年1月至12月的107例TBI患者作为验证集。依据弥散性血管内凝血评分结果,将患者分为正常组及异常组。分析TBI患者凝血功能异常的影响因素,构建列线图模型,受试者操作特征曲线评估模型的区分度,校准曲线检验模型的校准度,决策曲线分析评估患者获益情况。结果训练集中165例TBI患者出现凝血功能障碍,发生率为51.4%。正常组和异常组入院时脑疝、入院时肺部感染、入院时格拉斯哥昏迷量表(GCS)评分、入院24 h补液量、入院24 h输血史、入院24 h内手术、受伤至入院24 h内应用抗菌药物、吸烟史、肝脏疾病史、年龄、入院时激活部分凝血酶原时间、入院时凝血酶原时间、入院时国际标准化比值、入院时纤维蛋白原、入院时血小板计数比较,差异有统计学意义(P<0.05)。入院时GCS评分(3~8分)、入院24 h内手术、受伤至入院24 h内应用抗菌药物为TBI患者凝血功能异常的危险因素(OR=3.548、4.650、2.749,P<0.05)。构建回归方程为Logit(P)=1.266×入院时GCS评分(3~8分)+1.537×入院24 h内手术+1.011×受伤至入院24 h内应用抗菌药物+0.056。训练集预测TBI患者凝血功能障碍的曲线下面积为0.828,验证集为0.845。Hosmer-Lemeshow检验结果显示,理想曲线与校准曲线重合度高,模型具有良好的校准能力。训练集和验证集阈值概率分别为0.14~0.86和0.13~0.84时,模型获益阈值广泛,具有实用价值。结论入院时GCS评分、入院24 h内手术、受伤至入院24 h内应用抗菌药物是TBI患者凝血功能障碍的主要危险因素,以这些参数构建的列线图预测模型可以较好地预测TBI患者凝血功能障碍的发生风险。

Objective To investigate the risk factors of coagulation disorder in patients with traumatic brain injury(TBI),establish a risk prediction model,and to provide a theoretical basis for early identification and interventions.Methods Clinical data of 428 TBI patients admitted to the Department of Emergency,Union Shenzhen Hospital of Huazhong University of Science and Technology from January 2020 to December 2023 were retrospectively analyzed.A total of 321 cases of TBI patients from January 2020 to December 2022 were used as the training set,while 107 cases from January to December 2023 were used as the validation set.According to diffuse intravascular coagulation score,the patients were divided into normal group and abnormal group.The influencing factors of coagulation disorder in TBI patients were analyzed,a nomogram model was build,receiver operating characteristic curve were evaluated for the degree of differentiation of the model,the calibration curve was tested for the calibration of the model,and the decision curve analysis was evaluated the benefit of the patient.Results Coagulation disorder occurred in 165 TBI patients(51.4%)in training set.There were statistically significant differences in brain hernia at admission,lung infection,Glasgow coma scale(GCS)score at admission,fluid rehydration volume 24 hours of admission,blood transfusion history 24 hours of admission,operation within 24 hours of admission,injury to use of antibiotics within 24 hours of admission,smoking history,liver disease history,age,admission,fibrinogen at admission,and platelet count at admission between normal group and abnormal group(P<0.05).GCS score at admission(three to eight points),operation within 24 hours of admission,and injury to use of antibiotics within 24 hours of admission were risk factors for coagulation disorder in TBI patients(OR=3.548,4.650,2.749,P<0.05).The regression equation was constructed as Logit(P)=1.266×GCS score at admission(three to eight points)+1.537×operation within 24 hours of admission+1.011×injury to use of antibiotics within 24 hours of admission+0.056.The area under the curve for predicting coagulation disorder in TBI patients was 0.828 in the training set and 0.845 in the validation set.The result of Hosmer-Lemeshow test showed that the ideal curve has a high coincidence degree with the calibration curve,and the model has a good calibration ability.When the threshold probabilities of training set and validation set were 0.14-0.86 and 0.13-0.84 respectively,the benefit threshold of the model was wide and had practical value.Conclusion The main risk factors for coagulation disorder in TBI patients are GCS score at admission,operation within 24 hours of admission,and injury to use of antibiotics within 24 hours of admission.The prediction model based on these parameters can better predict the risk of coagulation disorder in TBI patients.