R Adams Cowley Shock Trauma Center, University of Maryland School of Medicine, Baltimore, MD 21201, USA.
J Trauma Acute Care Surg. 2012 Apr;72(4):1006-11. doi: 10.1097/TA.0b013e318241c059.
Available trauma scoring systems that predict need for higher echelons of care require data not available in the field. We hypothesized that analysis of continuous vital sign data in comparison to trauma registry data predicts mortality early in trauma patient management.
A real-time vital signs wave form and data capture system collected trauma patient data during prehospital management from Propaq 206E physiologic monitors. Analysis using statistical and mathematical software calculated receiver operator characteristic curves to evaluate the sensitivity and specificity of continuous vital sign waveforms in predicting mortality. The area under the curve (AUC) was calculated to determine nonsurvival by a particular vital sign (oxygen saturation [SpO2], heart rate, and systolic blood pressure) from these data, compared with a single value in the trauma registry, and to standard trauma scoring systems.
The average transport time from field to hospital for all patients was 25 minutes. Eight of 120 patients (7%) died; 5 of 8 patients (62%) died within the first 24 hours. Receiver operator characteristic analysis of mean SpO2 <90% versus mortality yielded an AUC of 0.76 (p = 0.005) with a sensitivity of 62% and specificity of 86% The initial SpO2 <90% measurement from the trauma registry yielded an AUC of 0.59. Preadmission Glasgow Coma Scale score yielded an AUC of 0.74 (p = 0.009). Injury Severity Score and Trauma-Injury Severity Score produced AUCs of 0.91 and 0.96, respectively. Revised Trauma Score gave an AUC of 0.73, no different from automated predictions of mortality from SpO2.
Injury Severity Score and Trauma-Injury Severity Score are predictive of mortality but rely on the inclusion of intra-abdominal and intrathoracic diagnostic data that are not readily available during field assessment. Automated vital signs data collection and analysis from a single noninvasive device with decision support has the potential to alleviate the dual burdens of patient triage and documentation required of the prehospital provider.
现有的创伤评分系统预测需要更高层次的治疗，但这些系统需要的数据在现场无法获得。我们假设，与创伤登记数据相比，连续生命体征数据的分析可以在创伤患者管理早期预测死亡率。
使用实时生命体征波形和数据采集系统从 Propaq 206E 生理监测器收集创伤患者在院前管理期间的数据。使用统计和数学软件进行分析，计算接收器操作特征曲线，以评估连续生命体征波形预测死亡率的敏感性和特异性。计算曲线下面积 (AUC)，以确定特定生命体征（血氧饱和度 [SpO2]、心率和收缩压）与创伤登记数据相比，这些数据与单一值相比，以及与标准创伤评分系统相比，无法生存。
所有患者从现场到医院的平均转运时间为 25 分钟。120 名患者中有 8 名（7%）死亡；8 名患者中有 5 名（62%）在 24 小时内死亡。平均 SpO2<90%与死亡率的接收器操作特征分析产生 AUC 为 0.76（p=0.005），敏感性为 62%，特异性为 86%。创伤登记处初始 SpO2<90%的测量值产生 AUC 为 0.59。入院前格拉斯哥昏迷量表评分产生 AUC 为 0.74（p=0.009）。伤害严重程度评分和创伤伤害严重程度评分分别产生 AUC 为 0.91 和 0.96。修订后的创伤评分产生 AUC 为 0.73，与 SpO2 自动预测死亡率无差异。
伤害严重程度评分和创伤伤害严重程度评分可预测死亡率，但依赖于包含在腹部和胸部诊断数据，这些数据在现场评估期间不易获得。来自单个非侵入性设备的自动生命体征数据采集和分析，以及决策支持，有可能减轻院前提供者在患者分诊和文件记录方面的双重负担。
