Zhao Xiao-Gang, Ma Yue-Feng, Zhang Mao, Gan Jian-Xin, Xu Shao-Wen, Jiang Guan-Yu
Department of Emergency Medicine, Second Affiliated Hospital, Medical School of Zhejiang University, Hangzhou 310009, China.
Chin J Traumatol. 2008 Dec;11(6):368-71. doi: 10.1016/s1008-1275(08)60074-7.
To assess whether these characteristics of less misclassification and greater area under receiver operator characteristic (ROC) curve of the new injury severity score (NISS) are better than the injury severity score (ISS) as applying it to our multiple trauma patients registered into the emergency intensive care unit (EICU).
This was a retrospective review of registry data from 2 286 multiple trauma patients consecutively registered into the EICU from January 1,1997 to December 31, 2006 in the Second Affiliated Hospital, Medical School of Zhejiang University in China. Comparisons between ISS and NISS were made using misclassification rates, ROC curve analysis, and the H-L statistics by univariate and multivariate logistic progression model.
Among the 2 286 patients, 176 (7.7%) were excluded because of deaths on arrival or patients less than 16 years of age. The study population therefore comprised 2 110 patients. Mean EICU length of stay (LOS) was 7.8 days ?2.4 days. Compared with the blunt injury group, the penetrating injury group had a higher percentage of male, lower mean EICU LOS and age. The most frequently injured body regions were extremities and head/neck, followed by thorax, face and abdomen in the blunt injury group; whereas, thorax and abdomen were more frequently seen in the penetrating injury group. The minimum misclassification rate for NISS was slightly less than ISS in all groups (4.01% versus 4.49%). However, NISS had more tendency to misclassify in the penetrating injury group. This, we noted, was attributed mainly to a higher false-positive rate (21.04% versus 15.55% for ISS, t equal to 3.310, P less than 0.001), resulting in an overall misclassification rate of 23.57% for NISS versus 18.79% for ISS (t equal to 3.290, P less than 0.001). In the whole sample, NISS presented equivalent discrimination (area under ROC curve: NISS equal to 0.938 versus ISS equal to 0.943). The H-L statistics showed poorer calibration (48.64 versus 32.11, t equal to 3.305, P less than 0.001) in the penetrating injury group.
NISS should not replace ISS because they share similar accuracy and calibration in predicting multiple blunt trauma patients. NISS may be more sensitive but less specific than ISS in predicting mortality in certain penetrating injury patients.
评估新损伤严重程度评分（NISS）的误分类较少及受试者操作特征（ROC）曲线下面积较大这些特性，在应用于入住急诊重症监护病房（EICU）的多发伤患者时，是否优于损伤严重程度评分（ISS）。
这是一项对1997年1月1日至2006年12月31日在中国浙江大学医学院附属第二医院连续入住EICU的2286例多发伤患者的登记数据进行的回顾性研究。采用误分类率、ROC曲线分析以及单因素和多因素逻辑回归模型的H-L统计量对ISS和NISS进行比较。
在2286例患者中，176例（7.7%）因入院时死亡或年龄小于16岁被排除。因此，研究人群包括2110例患者。EICU平均住院时间（LOS）为7.8天±2.4天。与钝性损伤组相比，穿透性损伤组男性比例更高，EICU平均LOS和年龄更低。钝性损伤组中最常受伤的身体部位是四肢和头/颈部，其次是胸部、面部和腹部；而穿透性损伤组中胸部和腹部更常见。所有组中NISS的最小误分类率略低于ISS（4.01%对4.49%）。然而，NISS在穿透性损伤组中误分类的倾向更大。我们注意到，这主要归因于更高的假阳性率（ISS为15.55%，NISS为21.04%，t = 3.310，P < 0.001），导致NISS的总体误分类率为23.57%，而ISS为18.79%（t = 3.290，P < 0.001）。在整个样本中，NISS表现出相当的辨别力（ROC曲线下面积：NISS为0.938，ISS为0.943）。H-L统计量显示穿透性损伤组的校准较差（48.64对32.11，t = 3.305，P < 0.001）。
NISS不应取代ISS，因为它们在预测多发钝性创伤患者时具有相似的准确性和校准。在预测某些穿透性损伤患者的死亡率方面，NISS可能比ISS更敏感但特异性更低。
