how to compute Incidence rate and incidence rate ratios with test for significance.

[GreenTree1]

Hi,

Can someone please help me with this question about how to compute IRR.

From my dataset, I have calculated  crude incidence rates for cases and controls using (numbers of events/ total person-years for each group).

And now I want to know if this association is significant.

I have a dataset with cases and controls according to disease status. For each ID I have

1.case/control status (1/0),
2. person-years contributed by each id
3. log person-years contributed by each ID
4. and number of events per ID.

Can someone please explain how can I calculate IRR and test for significance or if any more steps are needed? I have the log person-years variable because I was trying to run proc genmod but couldn't find information on how to run it in context of incidence rates.

I will be happy to provide more detail if needed.

Thankyou.

[Ksharp]

There is a function FINANCE('irr',  ) .

Or could try older function IRR() .

[GreenTree1]

Thanks, Ksharp. I am looking into this as well

[StatDave]

See this note which discusses estimating rates and rate ratios using a log-linked Poisson model and LS-means in PROC GENMOD.

[GreenTree1]

Thank you, Statdave

I went over the link that you had sent. Can you please clarify what is "ln" in the offset statement?  In the document, it is mentioned that 

"ln = log(n)"

In which case is it the log of the total person-time contributed by the entire cohort (cases and controls)? or the logs of the total person time contributed by each individual?

Here is the code that I have changed according to my data.

proc genmod data=mydata;
         class disease exposure;
         model outcome = disease exposure / dist=poisson link=log offset=(what is this?);
         run;

 I understand that I have to make some assumptions about the distribution of the outcome before I run this.

I appreciate your help.

[StatDave · Accepted solution]

As shown in the note, the OFFSET= option specifies a variable in the DATA= data set that is the log of the person-time for each observation. The beginning of the note shows how using the log of person-time as an offset in the model allows you to model a rate.

[GreenTree1]

It worked, thank you for the clarification.