目的 通过分析儿童创伤性颅脑损伤（traumatic brain injury， TBI）患者临床输血影响因素，建立其临床输血预测模型，以指导临床输血决策。 方法 回顾性分析4家医疗机构2015年1月1日至2022年12月31日收治的1 535例儿童TBI患者病历资料，根据患者院内是否输注红细胞将患者分为输血组和未输血组，分析两组患者间临床资料，通过logistic回归筛选患者院内输血相关危险因素，并采用列线图建立预测模型，利用ROC曲线评价模型的预测性能。 结果 输血组与未输血组患者基本资料、临床指征和实验室检测指标间均存在差异（P < 0.05）。输血组患者院内死亡率、并发症发生率、机械通气、ICU入住率、住院时间均明显高于未输血组（P < 0.05）。多因素logistic回归分析显示心率、其他骨折、治疗方式、血红蛋白（Hb）、血小板计数（Plt）、活化部分凝血活酶时间（APTT）、D-二聚体7个影响因素是TBI患者临床输血的独立危险因素。根据独立危险因素构建的临床输血预测模型ROC曲线下面积为0.95（95%CI： 0.94 ~ 0.97）；校准曲线和决策曲线亦显示该模型具有良好的预测效果。 结论 心率、其他骨折、治疗方式、Hb、Plt、APTT、D-二聚体是TBI患者临床输血的独立危险因素，据此构建的预测模型具有较好预测效果，可指导临床医师输血决策，有助于提高患者救治成功率。

Objective To develop a predictive model for guiding blood transfusion decisions in pediatric patients with traumatic brain injury （TBI） by identifying and analyzing key factors that influence blood transfusion requirements. Methods A retrospective analysis was conducted on the clinical data of 1，535 pediatric patients with TBI admitted to four medical institutions from January 1， 2015， to December 31， 2022. Patients were divided into two groups： those who received red blood cell transfusions during hospitalization and those who did not. Comparative analyses were performed on demographic， clinical， and laboratory data between these two groups. Logistic regression analysis was used to identify risk factors associated with in-hospital blood transfusion， and a predictive model was developed using a nomogram. The performance of this model was evaluated using a receiver operating characteristic （ROC） curve. Results Significant differences were observed between the blood transfusion and non-blood transfusion groups in terms of baseline demographics， clinical indicators， and laboratory test results （all P < 0.05）. Patients in the blood transfusion group exhibited significantly higher in-hospital mortality， complication rates， use of mechanical ventilation， ICU admission rates， and length of stay compared to those in the non-blood transfusion group （all P < 0.05）. Multivariate logistic regression analysis identified heart rate， presence of other fractures， treatment methods， hemoglobin （Hb）， platelet count （Plt）， activated partial thromboplastin time （APTT）， and D-dimer levels as independent risk factors for blood transfusion in TBI patients. The area under the ROC curve for the blood transfusion prediction model， based on these independent risk factors， was 0.95 （95%CI： 0.94 ~ 0.97）， indicating excellent predictive accuracy. Calibration and decision curves further validated the robustness and reliability of the model's predictive capacity. Conclusions Heart rate， presence of other fractures， treatment methods， Hb， Plt count， APTT， and D-dimer levels serve as independent risk factors for blood transfusion in TBI patients. The prediction model developed based on these factors demonstrates excellent predictive performance， thereby guiding clinicians in making informed blood transfusion decisions and enhancing the success rate of patient outcomes.