OR: Odds ratio 比值比
RR: Risk ratio/ relative risk 相对危险度
HR: Hazard ratio
这三个指标都用来衡量 结局 和 暴露因素 之间的关系
OR=病例组的暴露比值/对照组的暴露比值
病例组的暴露比值（Odds病例）= (a/(a+c)) ÷ (c/(a+c)) = a/c
对照组的暴露比值（Odds对照）= (b/(b+d)) ÷ (d/(b+d)) = b/d
比值比（Odds Ratio）= (病例组的暴露比值(a/c)) ÷ (对照组的暴露比值(b/d)) = ad/bc
与RR相似，OR是两个比值的比，因此一定是正数。OR的解释与RR也类似。
作者：海马
链接：https://www.zhihu.com/question/24436989/answer/38582667
来源：知乎
著作权归作者所有。商业转载请联系作者获得授权，非商业转载请注明出处。
In a longitudinal study, we can use risk ratio (cumulative incidence ratio) as table below;
<img src="https://pic4.zhimg.com/9d6cf5cb7b8f4c378f4188b7ce5c949f_b.jpg" data-rawwidth="493" data-rawheight="259" class="origin_image zh-lightbox-thumb" width="493" data-original="https://pic4.zhimg.com/9d6cf5cb7b8f4c378f4188b7ce5c949f_r.jpg">
However, in a case control study, we use odds ratio instead of risk ratio due to the fact thatthe cases and controls are chosen by the investigator.We can use different matching ratios such as 1:1 or 1:10 to choose cases and controls, which may lead to different results if we use risk ratio. The following is a good example;
In the same source population
If we choose 1:1 matching, the table is shown below;
<img src="https://pic3.zhimg.com/198390c1af156a312b8b60b19b2183ca_b.jpg" data-rawwidth="282" data-rawheight="192" class="content_image" width="282">
=
If we choose 1:10 matching, the table is shown below;
<img src="https://pic1.zhimg.com/900751dbd53212a5b954323742675d98_b.jpg" data-rawwidth="270" data-rawheight="183" class="content_image" width="270">
=
From the table, we can see that the risk ratios (1.83 and 3.08) are different under different matching ratio settings, and the odds ratios (3.5) are the same. Since the source population stays constant, we want constant results no matter what matching ratios we use. That's the reason we choose odds ratio in case control study.
Risk ratio depends on chosen matching ratio in case control.
