Zhang Zhongheng
Department of Critical Care Medicine, Jinhua Municipal Central Hospital, Jinhua Hospital of Zhejiang University, Jinhua 321000, China.
Ann Transl Med. 2016 Sep;4(18):341. doi: 10.21037/atm.2016.05.42.
Case-crossover design is a variation of case-control design that it employs persons' history periods as controls. Case-crossover design can be viewed as the hybrid of case-control study and crossover design. Characteristic confounding that is constant within one person can be well controlled with this method. The relative risk and odds ratio, as well as their 95% confidence intervals (CIs), can be estimated using Cochran-Mantel-Haenszel method. R codes for the calculation are provided in the main text. Readers may adapt these codes to their own task. Conditional logistic regression model is another way to estimate odds ratio of the exposure. Furthermore, it allows for incorporation of other time-varying covariates that are not constant within subjects. The model fitting per se is not technically difficult because there is well developed statistical package. However, it is challenging to convert original dataset obtained from case report form to that suitable to be passed to clogit() function. R code for this task is provided and explained in the text.
病例交叉设计是病例对照设计的一种变体，它将个体的历史时期用作对照。病例交叉设计可被视为病例对照研究与交叉设计的结合。这种方法能够很好地控制个体内部恒定的特征性混杂因素。相对风险和比值比及其95%置信区间（CIs）可使用 Cochr an-Mantel-Haenszel 方法进行估计。正文中提供了计算的R代码。读者可根据自身任务对这些代码进行调整。条件逻辑回归模型是估计暴露比值比的另一种方法。此外，它还允许纳入个体内部不恒定的其他随时间变化的协变量。由于有成熟的统计软件包，模型拟合本身在技术上并不困难。然而，将从病例报告表获得的原始数据集转换为适合传递给clogit()函数的数据集具有挑战性。文中提供并解释了完成此任务的R代码。
