Hello everybody;
I'm working on a paper and some parts of the statistical results are vague for me!
I calculated OR using a multiple logistic regression model, to determine the effect of some factors on mortality. And then I calculated HR using the cox proportional hazard model (CPHM) to investigate affective factors on the progression of the disease to death. Now my question is that:
1. Can I report both of them in a paper?
2. And if your answer is Yes, what is the difference in their interpretation and reporting?
Thank you for your help.
Yes. You can report both.
This explanation is extremely simplified, but I think it covers more or less what you asked. Shahab Mohammadebrahimi
OR is the odds ratio at a particular followup time (e.g. at the 12 month mark the A is 2.2 times more associated with the occurrence of B, when compared to the the absence of A). - You only look at the data at the 12 month mark, not before, nor after. It doesn't matter if the events in each group took place very early (1 month mark) or later (11 month mark).
HR is the time dependent "version" of the OR. What it does is it creates a mathematical formula that takes into account the timing of each event. So if events occur earlier, the hazard of that group will be higher than if it only occurred later. Although it is different than the area under the curve (AUC), the interpretation is somewhat similar: it the events occur earlier, there is s steep decrease in the kaplan meier curve, the AUC will be small and the hazard will be big and vice versa. If the AUC are quite different, then the HR will be >1 or
Although the null hypotheses are the same (i.e., OR =1 and HR = 1), they have different interpretations. The readers can get confused if you use both in the same paper as they interpreted differently. You can avoid reporting HR as the interpretation of HR is often complicated (https://dx.doi.org/10.1097%2FEDE.0b013e3181c1ea43). Also, the OR is more intuitive, while the HR varies over time and has an inherent selection bias.
Interpretations are different for "OR" and "HR". However, there is a situation (short follow‐up and low incidence of event occurrence: Article Efficiency of the logistic regression and Cox proportional h...
) when logistic regression and Cox PH model can provide similar results.
Thank you all,
I'll use the good solutions and tips you provided.
- Badges
- Science topic
- Similar topics
- Psychology