目的 构建并验证基于决策树的急性创伤性脑损伤（traumatic brain injury，TBI)患者早期死亡风险分诊模型，为急诊分诊提供准确、直观的新工具。 方法 回顾性收集某院急性TBI患者2287例的临床数据，构建早期死亡风险的分诊模型，采用十字交叉验证，并与改良早期预警评分(modified early waring score，MEWS)、修正创伤评分（revised trauma score,RTS）、改良快速急诊医学评分（the modified rapid emergency medicine score,mREMS）及损伤机制格拉斯哥年龄血压评分（mechanism,Glasgow coma scale,age and arterial pressure score,MGAP）比较预测效能。 结果 2287例急性TBI患者中，24 h内发生死亡者166例。决策树共3层，15个节点，筛选出6个解释变量，分别是瞳孔反应性、RTS、MGAP、MEWS、血氧饱和度及收缩压。决策树模型的受试者工作特征曲线下面积为0.917，高于MEWS、RTS、MGAP评分和mREMS，差异均有统计学意义（均P＜0.05）。 结论 基于决策树构建的急性TBI患者早期死亡风险分诊模型能准确预测急性TBI患者24 h内死亡风险，可作为医护人员分诊的决策依据。

Objective To construct and validate a triage model of early death risk prediction in acute traumatic brain injury (TBI) patients based on the decision tree，and to provide a novel，accurate，intuitive tool for emergency triage. Methods The clinical data of 2,287 acute TBI patients in a hospital were retrospectively collected，and a triage model for early death risk prediction was constructed.Cross validation was used for validation.The prediction efficiency of this model was compared with that of modified early waring score (MEWS)，revised trauma score (RTS)，the modified rapid emergency medicine score (mREMS) and Mechanism，Glasgow coma scale，age and arterial pressure scores (MGAP). Results There were 166 subjects died within 24 hours among the 2287 acute TBI patients.The decision tree had 3 layers and 15 nodes，with 6 explanatory variables，namely，pupillary reactivity，RTS，MGAP，MEWS，oxygen saturation of blood and systolic pressure.The area under the receiver operating characteristic curve of the triage model was 0.917，which was higher than that of MEWS、RTS、MGAP and mREMS （all P＜0.05）. Conclusion The triage model of early death risk prediction in acute TBI patients based on the decision tree can accurately predict acute TBI patients’ death risk within 24 hours which can serve as a decision basis for medical staff triaging.It can provide reference for medical staff’s triage.

R472．2

南通市卫健委指令性课题面上项目(MS2022010)