05-11-2012 07:57 PM
I have data with around 60,000 observations. I need to calculate age adjusted mortality rates from it. I have variable Died (1 0) and AGE as continuous variable. I also have variable WEIGHT. how do I calculate age adjusted mortality rate?
can somebody please give me an example code or guide to appropriate link.
Appreciate your help in advance!
05-11-2012 11:12 PM
That's a million dollar question for life insurers! Read the introduction to SAS survival analysis procedures at :
05-14-2012 10:27 AM
Unless you also have a time-to-death variable, you need to be using logistic regression rather than the survival analysis that PG pointed to. The syntax is in the documentation and Paul Allison has an excellent BBU with the how-to. Interpreting risk adjusted estimates can be challenging at times.
If your WEIGHT variable is how heavy a person is, then it is just another control variable. If it is a sampling weight, then there is a WEIGHT statement to address it. IF you have results from a formal sample-survey design, you may be better served by using PROC SURVEYLOGISTIC (and by consulting with an expert in sample survey analysis!).
05-15-2012 11:14 PM
Thank you for reply.
Sorry but I don't know what you mean by "BBU" (I am new to forum). I am not sure how to get percentage rate from logistic regression (it gives odds ratio. right?)
I am looking for death from disease "X". I dont have time variable. I need to get age and sex adjusted mortality rate.
Let me know if following is the correct way.
10/100 people died from disease "X" . so crude mortality is 10%.
now I do logistic regression (model died= diseaseX age gender) and if I get odds ratio of 1.25.
So age and gender adjustment adds around 25% additional mortality. so 25% of crude mortality (10%) will be 2.5%.
So the age and gender adjusted mortality from disease X is 10% +2.5% = 12.5%
IS THIS CORRECT OR AM I COMPLETELY WRONG?
05-16-2012 12:22 PM
At least I think so.
Check this link and see if it makes sense.
Basically you don't age adjust to your population but to some standard population so all age adjusted ratio's can be compared across locations/regions etc.
If you're running a model and calculating mortality rates for a paper and would like to adjust for the age factor, ie age adjusted rates then the you would add age into the model. But I don't think your calculation of how to adjust is accurate either. You'll need to reference a logisitic regression book for that one.
Or I could be wrong entirely. It depends on your question and we need some more info on what you're trying to do to help answer that.
05-16-2012 12:31 PM
Seconding Reeza -- you need to read a good book on Logistic Regression, rather than asking for a tutorial on the statistical principles here.
"BBU" stands for "Books By Users" and you can find them on support.sas.com in the bookstore. Paul Allison has a useful ones on Logistic Regression (there are others, I've read his) and he also does live training.
12-20-2014 02:38 PM
just to add: if "adjustment" in the sense of "standardization" is meant, then proc stdrate could be of help, see
which is a very nice thing I discovered recently.