Al-Jaishi Ahmed A, Moist Louise M, Oliver Matthew J, Nash Danielle M, Fleet Jamie L, Garg Amit X, Lok Charmaine E
1 Institute for Clinical Evaluative Sciences, Toronto, ON, Canada.
2 Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, ON, Canada.
J Vasc Access. 2018 Nov;19(6):561-568. doi: 10.1177/1129729818762008. Epub 2018 Mar 12.
: We assessed the validity of physician billing codes and hospital admission using International Classification of Diseases 10th revision codes to identify vascular access placement, secondary patency, and surgical revisions in administrative data.
: We included adults (≥18 years) with a vascular access placed between 1 April 2004 and 31 March 2013 at the University Health Network, Toronto. Our reference standard was a prospective vascular access database (VASPRO) that contains information on vascular access type and dates of placement, dates for failure, and any revisions. We used VASPRO to assess the validity of different administrative coding algorithms by calculating the sensitivity, specificity, and positive predictive values of vascular access events.
: The sensitivity (95% confidence interval) of the best performing algorithm to identify arteriovenous access placement was 86% (83%, 89%) and specificity was 92% (89%, 93%). The corresponding numbers to identify catheter insertion were 84% (82%, 86%) and 84% (80%, 87%), respectively. The sensitivity of the best performing coding algorithm to identify arteriovenous access surgical revisions was 81% (67%, 90%) and specificity was 89% (87%, 90%). The algorithm capturing arteriovenous access placement and catheter insertion had a positive predictive value greater than 90% and arteriovenous access surgical revisions had a positive predictive value of 20%. The duration of arteriovenous access secondary patency was on average 578 (553, 603) days in VASPRO and 555 (530, 580) days in administrative databases.
: Administrative data algorithms have fair to good operating characteristics to identify vascular access placement and arteriovenous access secondary patency. Low positive predictive values for surgical revisions algorithm suggest that administrative data should only be used to rule out the occurrence of an event.
我们使用国际疾病分类第十版编码评估医师计费代码和医院入院情况，以在管理数据中识别血管通路置入、二级通畅情况和手术修正。
我们纳入了2004年4月1日至2013年3月31日期间在多伦多大学健康网络置入血管通路的成年人（≥18岁）。我们的参考标准是一个前瞻性血管通路数据库（VASPRO），其中包含血管通路类型、置入日期、失败日期以及任何修正的信息。我们使用VASPRO通过计算血管通路事件的敏感性、特异性和阳性预测值来评估不同管理编码算法的有效性。
识别动静脉通路置入的最佳算法的敏感性（95%置信区间）为86%（83%，89%），特异性为92%（89%，93%）。识别导管插入的相应数字分别为84%（82%，86%）和84%（80%，87%）。识别动静脉通路手术修正的最佳编码算法的敏感性为81%（67%，90%），特异性为89%（87%，90%）。捕获动静脉通路置入和导管插入的算法的阳性预测值大于90%，而动静脉通路手术修正的阳性预测值为20%。在VASPRO中，动静脉通路二级通畅的平均持续时间为578（553，603）天，在管理数据库中为555（530，580）天。
管理数据算法在识别血管通路置入和动静脉通路二级通畅方面具有中等至良好的操作特征。手术修正算法的低阳性预测值表明管理数据仅应用于排除事件的发生。
