Stein Sherman C, Fabbri Andrea, Servadei Franco, Glick Henry A
Department of Neurosurgery, Hospital of the University of Pennsylvania, Philadelphia, PA 19106, USA.
Ann Emerg Med. 2009 Feb;53(2):180-8. doi: 10.1016/j.annemergmed.2008.01.002. Epub 2008 Mar 14.
A number of clinical decision aids have been introduced to limit unnecessary computed tomographic scans in patients with mild traumatic brain injury. These aids differ in the risk factors they use to recommend a scan. We compare the instruments according to their sensitivity and specificity and recommend ones based on incremental benefit of correctly classifying patients as having surgical, nonsurgical, or no intracranial lesions.
We performed a secondary analysis of prospectively collected database from 7,955 patients aged 10 years or older with mild traumatic brain injury to compare sensitivity and specificity of 6 common clinical decision strategies: the Canadian CT Head Rule, the Neurotraumatology Committee of the World Federation of Neurosurgical Societies, the New Orleans, the National Emergency X-Radiography Utilization Study II (NEXUS-II), the National Institute of Clinical Excellence guideline, and the Scandinavian Neurotrauma Committee guideline. Excluded from the database were patients for whom the history of trauma was unclear, the initial Glasgow Coma Scale score was less than 14, the injury was penetrating, vital signs were unstable, or who refused diagnostic tests. Patients revisiting the emergency department within 7 days were counted only once.
The percentage of scans that would have been required by applying each of the 6 aids were Canadian CT head rule (high risk only) 53%, Canadian (medium & high risk) 56%, the Neurotraumatology Committee of the World Federation of Neurosurgical Societies 56%, New Orleans 69%, NEXUS-II 56%, National Institute of Clinical Excellence 71%, and the Scandinavian 50%. The 6 decision aids' sensitivities for surgical hematomas could not be distinguished statistically (P>.05). Sensitivity was 100% (95% confidence interval [CI] 96% to 100%) for NEXUS-II, 98.1% (95% CI 93% to 100%) for National Institute of Clinical Excellence, and 99.1% (95% CI 94% to 100%) for the other 4 clinical decision instruments. Sensitivity for any intracranial lesion ranged from 95.7% (95% CI 93% to 97%) (Scandinavian) to 100% (95% CI 98% to 100%) (National Institute of Clinical Excellence). In contrast, specificities varied between 30.9% (95% CI 30% to 32%) (National Institute of Clinical Excellence) and 52.9% (95% CI 52% to 54) (Scandinavian).
NEXUS-II and the Scandinavian clinical decision aids displayed the best combination of sensitivity and specificity in this patient population. However, we cannot demonstrate that the higher sensitivity of NEXUS-II for surgical hematomas is statistically significant. Therefore, choosing which of the 2 clinical decision instruments to use must be based on decisionmakers' attitudes toward risk.
已引入多种临床决策辅助工具，以限制轻度创伤性脑损伤患者进行不必要的计算机断层扫描。这些辅助工具在用于推荐扫描的风险因素方面存在差异。我们根据其敏感性和特异性对这些工具进行比较，并基于将患者正确分类为有手术指征、非手术指征或无颅内病变的增量效益来推荐工具。
我们对前瞻性收集的7955例10岁及以上轻度创伤性脑损伤患者的数据库进行了二次分析，以比较6种常见临床决策策略的敏感性和特异性：加拿大头颅CT规则、世界神经外科协会联合会神经创伤学委员会、新奥尔良标准、国家急诊X线摄影利用研究II（NEXUS-II）、国家临床优化研究所指南和斯堪的纳维亚神经创伤委员会指南。数据库中排除了创伤史不明、初始格拉斯哥昏迷量表评分低于14分、损伤为穿透性、生命体征不稳定或拒绝诊断检查的患者。7天内再次就诊急诊科的患者仅计数一次。
应用6种辅助工具所需扫描的百分比分别为：加拿大头颅CT规则（仅高风险）53%，加拿大（中高风险）56%，世界神经外科协会联合会神经创伤学委员会56%，新奥尔良69%，NEXUS-II 56%，国家临床优化研究所71%，斯堪的纳维亚5０%。6种决策辅助工具对手术性血肿的敏感性在统计学上无法区分（P>0.05）。NEXUS-II的敏感性为100%（95%置信区间[CI]96%至100%），国家临床优化研究所为98.1%（95%CI 93%至100%），其他4种临床决策工具为99.1%（95%CI 94%至100%）。对任何颅内病变的敏感性范围为95.7%（95%CI 93%至97%）（斯堪的纳维亚）至100%（95%CI 98%至100%）（国家临床优化研究所）。相比之下，特异性在30.9%（95%CI 30%至32%）（国家临床优化研究所）和52.9%（95%CI 52%至54%）（斯堪的纳维亚）之间。
在该患者群体中，NEXUS-II和斯堪的纳维亚临床决策辅助工具在敏感性和特异性方面表现出最佳组合。然而，我们无法证明NEXUS-II对手术性血肿的较高敏感性具有统计学意义。因此，选择使用这两种临床决策工具中的哪一种必须基于决策者对风险的态度。
