Division of Genetics and Development, Toronto Western Research Institute, Toronto, Ont., Canada.
Neuroepidemiology. 2011;37(2):96-101. doi: 10.1159/000330835. Epub 2011 Sep 15.
Miscoding is a common source of error in population-based registries. Given this, we performed a validation study comparing the Canadian National Trauma Registry (NTR) data based on the 10th Revision of the International Classification of Diseases coding with clinical data from an institutional database.
All patients with acute spine trauma who were admitted to Toronto Western Hospital from May 2003 to April 2007 were included. Accuracy, sensitivity and specificity were estimated having chart data abstraction as the gold standard.
There were 92 patients with spine trauma (50 males, 42 females; ages from 16 to 102 years). The use of the NTR as a spine trauma database has an accuracy of 87%, sensitivity of 89.8% and specificity of 25%. If the same database is considered as a spinal cord injury (complete motor injury) database, there will be a decrease in the precision with an accuracy of 32.6%, sensitivity of 81.3% and specificity of 6.7%.
Our results indicate that the NTR may be relatively more precise when used as a database of spine trauma in comparison with its use as a spinal cord injury database. However, the low specificity suggests that the NTR should be comprehensively validated using data from the other institutions that contribute with data collection for the NTR.
在基于人群的注册中心，编码错误是常见的错误来源。有鉴于此，我们进行了一项验证研究，比较了基于国际疾病分类第 10 版编码的加拿大国家创伤登记处 (NTR) 数据与机构数据库中的临床数据。
纳入 2003 年 5 月至 2007 年 4 月期间入住多伦多西部医院的所有急性脊柱创伤患者。以图表数据提取作为金标准来估计准确性、敏感性和特异性。
共有 92 例脊柱创伤患者（50 名男性，42 名女性；年龄 16 至 102 岁）。将 NTR 用作脊柱创伤数据库的准确性为 87%，敏感性为 89.8%，特异性为 25%。如果将同一数据库视为脊髓损伤（完全运动损伤）数据库，则准确性会降低至 32.6%，敏感性为 81.3%，特异性为 6.7%。
我们的结果表明，与将 NTR 用作脊髓损伤数据库相比，将其用作脊柱创伤数据库时可能更精确。但是，特异性较低表明 NTR 应使用来自为 NTR 数据收集做出贡献的其他机构的数据进行全面验证。
