论文标题
洞悉时间分层的病例交叉研究
Insight into bias in time-stratified case-crossover studies
论文作者
论文摘要
案例交叉设计的使用已在瞬态关联的流行病学和医学研究中广泛存在。但是，最受欢迎的参考选择策略，时间分层的模式，并不是控制案例 - 分解研究中偏差的合适解决方案。为了证明这一点，我们进行了每日臭氧（O3）记录的时间序列分解。审查了时间分层模式控制年度，每月和每周时间趋势的能力；并发现它无法控制每周的时间趋势。基于这一发现，我们提出了一种新的逻辑回归方法，其中我们对每周时间趋势进行了调整。传统模型与提出方法之间的比较是通过仿真进行的。进行了一项实证研究，以探索空气污染物与AMI住院的潜在关联。总而言之，时间分层的架构可有效控制年度和每月的时间趋势，但不能在每周的时间趋势上进行。因此，从传统的后勤回归中进行的估计基本上揭示了每周时间趋势的影响，而不是瞬态效应。相比之下，提出的逻辑回归和调整每周时间趋势可以有效地消除案例分解研究中的系统偏见。
The use of case-crossover designs has become widespread in epidemiological and medical investigations of transient associations. However, the most popular reference-select strategy, the time-stratified schema, is not a suitable solution for controlling bias in case-crossover studies. To prove this, we conducted a time series decomposition for daily ozone (O3) records; scrutinized the ability of the time-stratified schema on controlling the yearly, monthly and weekly time trends; and found it failed on controlling the weekly time trend. Based on this finding, we proposed a new logistic regression approach in which we did adjustment for the weekly time trend. A comparison between the traditional model and the proposed method was done by simulation. An empirical study was conducted to explore potential associations between air pollutants and AMI hospitalizations. In summary, time-stratified schema provide effective control on yearly and monthly time trends but not on weekly time trend. Therefore, the estimation from the traditional logistical regression basically reveals the effect of weekly time trend, instead of the transient effect. In contrast, the proposed logistic regression with adjustment for weekly time trend can effectively eliminate system bias in case-crossover studies.
