Department of Emergency Medicine and Critical Care, National Center for Global Health and Medicine Hospital, Tokyo, Japan.
World J Surg. 2012 Apr;36(4):813-8. doi: 10.1007/s00268-012-1498-z.
The objective of the present study was to identify logistic regression models with better survival prediction than the Trauma and Injury Severity Score (TRISS) method in assessing blunt trauma (BT) victims in Japan and Thailand. An additional aim was to demonstrate the feasibility of probability of survival (Ps) estimation without respiratory rate (RR) on admission, which is often missing or unreliable in Asian countries.
We used BT patient data (n = 15,524) registered in the Japan Trauma Data Bank (JTDB, 2005-2008). We also extracted data on BT patients injured in the Khon Kaen District between January 2005 and December 2008 (n = 6,411) from the Khon Kaen Hospital Trauma Registry. For logistic regression analyses, we chose the Injury Severity Score (ISS), age year (AY), Glasgow Coma Scale (GCS) score, systolic blood pressure (SBP), RR, and their coded values (c) as explanatory variables, as well as the Revised Trauma Score (RTS). We estimated parameters by the method of maximum likelihood estimation, and utilized Akaike's Information Criterion (AIC), the area under the receiver operating characteristic curve (AUROCC), and accuracy for model comparison. A model having the lower AIC is considered to be the better model.
The AIC of the model using AY was lower than that of the model using the coded value for AY (cAY) (used by the TRISS method). The model using ISS, AY and cGCS, cSBP, and cRR instead of the RTS demonstrated the lowest AIC in both data groups. The same trend could be observed in the AUROCCs and the accuracies. In the Khon Kaen data, we found no additional reduction of the AIC in the model using the cRR variable compared to the model without cRR.
For better prediction of Ps, the actual number of the AY should be used as an explanatory variable instead of the coded value (used by the TRISS method). The logistic regression model using the ISS, AY, and coded values of SBP, GCS, and RR estimates the best prediction. Information about RR seems to be unimportant for survival prediction in BT victims in Asian countries.
本研究旨在确定比创伤和损伤严重程度评分（TRISS）方法更能预测钝器伤（BT）患者生存情况的逻辑回归模型，该研究在日本和泰国进行。本研究的另一个目的是展示在亚洲国家，因呼吸频率（RR）缺失或不可靠，而不使用入院时 RR 来估计生存概率（Ps）的可行性。
我们使用了日本创伤数据库（JTDB，2005-2008 年）中登记的 BT 患者数据（n=15524）。我们还从孔敬医院创伤登记处提取了 2005 年 1 月至 2008 年 12 月期间孔敬区受伤的 BT 患者的数据（n=6411）。对于逻辑回归分析，我们选择损伤严重程度评分（ISS）、年龄（AY）、格拉斯哥昏迷评分（GCS）、收缩压（SBP）、RR 及其编码值（c）以及修订创伤评分（RTS）作为解释变量。我们使用最大似然估计法估计参数，并利用赤池信息量准则（AIC）、受试者工作特征曲线下面积（AUROCC）和准确性来进行模型比较。AIC 较低的模型被认为是更好的模型。
使用 AY 的模型的 AIC 低于使用 AY 编码值（TRISS 方法中使用）的模型的 AIC。在两组数据中，使用 ISS、AY 和 cGCS、cSBP 和 cRR 替代 RTS 的模型的 AIC 最低。AUROCC 和准确性也呈现出相同的趋势。在孔敬数据中，与不使用 cRR 的模型相比，使用 cRR 变量的模型的 AIC 没有进一步降低。
为了更好地预测 Ps，应使用 AY 的实际数值作为解释变量，而不是使用编码值（TRISS 方法中使用）。使用 ISS、AY 和 SBP、GCS 和 RR 的编码值的逻辑回归模型可以更好地预测。在亚洲国家，RR 信息对 BT 患者的生存预测似乎不重要。
