A Standardized Algorithm for Determining the Underlying Cause of Death in HIV Infection as AIDS or non-AIDS Related: Results from the EuroSIDA Study
A Standardized Algorithm for Determining the Underlying Cause of Death in HIV Infection as AIDS or non-AIDS Related: Results from the EuroSIDA Study
复制标题
DOI：
10.1310/hct1202-109
复制
发表时间：
2011-03-01
影响因子：
--
通讯作者：
Kirk, Ole
中科院分区：
文献类型：
--
作者：
Kowalska, Justyna D.;Mocroft, Amanda;Kirk, Ole
目的：分析死亡原因随时间的变化对于了解艾滋病毒感染者死亡率的新趋势至关重要，但关于死亡原因的数据往往缺失。这就造成了分析上的限制，纵向研究收集数据的方法也在改变，这是该领域认识和知识增加的自然结果。为了监测和分析死亡率随时间的变化，我们在EuroSIDA研究中探讨了这一问题，并提出了一个标准化的方案，统一收集的数据，并允许将所有死亡分类为艾滋病或非艾滋病相关，包括死亡原因缺失的事件。研究方法：比较了EuroSIDA研究中AIDS或非AIDS相关的潜在死因的几种分类：基于外部标准化方法（CoDe程序）的中心分类（CC-参考组）、研究中心研究者报告的当地队列分类（LCC）和基于特定AIDS事件后生存时间创建的4种算法（ALG）。结果：共发生2，783例死亡，收集了540份CoDe表，488份用于评估协议。CC和LCC之间的一致性是实质性的（kappa = 0.7），CC和ALG之间的一致性是中等的（kappa < 0.6）。因此，推导出一种逐步算法，优先考虑CC而不是LCC，在没有可用信息的患者中，优先考虑最佳拟合ALG。使用这一算法，1 332例（47.9%）死亡被归类为艾滋病，1 451例（52.1%）与艾滋病无关。结论：我们提出的逐步算法分类死亡为未来的研究提供了一个有价值的工具，但在另一个设置验证是必要的。
Objectives: Analyzing changes in causes of death over time is essential for understanding the emerging trends in HIV population mortality, yet data on cause of death are often missing. This poses analytic limitations, as does the changing approach in data collection by longitudinal studies, which are a natural consequence of an increased awareness and knowledge in the field. To monitor and analyze changes in mortality over time, we have explored this issue within the EuroSIDA study and propose a standardized protocol unifying data collected and allowing for classification of all deaths as AIDS or non-AIDS related, including events with missing cause of death. Methods: Several classifications of the underlying cause of death as AIDS or non-AIDS related within the EuroSIDA study were compared: central classification (CC-reference group) based on an externally standardised method (the CoDe procedures), local cohort classification (LCC) as reported by the site investigator, and 4 algorithms (ALG) created based on survival times after specific AIDS events. Results: A total of 2,783 deaths occurred, 540 CoDe forms were collected, and 488 were used to evaluate agreements. The agreement between CC and LCC was substantial (kappa = 0.7) and the agreement between CC and ALG was moderate (kappa < 0.6). Consequently, a stepwise algorithm was derived prioritizing CC over LCC and, in patients with no information available, best-fit ALG. Using this algorithm, 1,332 (47.9%) deaths were classified as AIDS and 1,451(52.1%) as non-AIDS related. Conclusions: Our proposed stepwise algorithm for classifying deaths provides a valuable tool for future research, however validation in another setting is warranted.