This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Survival analysis: Why to categorize age?

Hi there!

I wonder ed why scientists usually categorize age in survival analysis? For example, a 51-year-old patient is in the same category with a 60-year-old, but a 61-year old in a different category from the 60? (Supposing dividing the data in 10-years intervals). Why age is not considered continuous? A reference (textbook) will be appreciated as well.

survival
• 2,587 views • link
updated by Kevin Blighe Kevin Blighe • written by english.server english.server

The only reason may be if the function response(age) is complex and can not be easily modeled. Otherwise it is a "mistake" to categorize anything since it leads to the loss of power.

Ok, another reason - clinicians want to have a clear simple cut off, so before age x risk is low, after - high

• 0 views • link
written by German.M.Demidov German.M.Demidov

This was my "feeling".

I also thought that if there is something mysterious leading statisticians/bioinformaticians to categorize data, one may use "sliding windows" 

0-9 ; 10-19; .....

1-10 ; 11-20; ....

.

.

9-18 ; 19-28; .....

and then recheck the results.

• 0 views • link
written by english.server english.server

1 answer

It will depend on the disease under study - many are age-related / confounded by age, including cancer.

Regarding the specifics of the difference between 60 and 61, it's simply a cut-off / threshold. One could just as easily make the same argument about p<0.05 and p<0.051

In other disease areas, 55 may be a single cut-off for age, as it is regarded as the age at which menopause commences, but this obviously differs from individual to individual.

Kevin

• 0 views • link
written by Kevin Blighe Kevin Blighe

Thank you. It seems I still have to study more.

• 0 views • link
written by english.server english.server

Not sure that you need to study more, as such... in certain diseases, there are just well established relationships between age and disease. This is therefore more related to medicine and epidemiology, as opposed to being about bioinformatics.

What German says is correct, too, i.e., that using strict cut-offs and creating categorical variables from continuous variables can result in lost power.

There are no standards in how to deal with this.

• 0 views • link
written by Kevin Blighe Kevin Blighe

Log in to answer this question.

More posts like this