How to get Risk ratio from a Poisson model with count data & interaction terms

[JME1]

Hello!

I'm having a bit of trouble getting RR from a Poisson model using count data and with interaction terms. I have data like this:

data x;

count   time     intervention   status  totalcount;

12         1            1                   0          21

34         2            0                   0          40

12         3            0                   1          20

15         4            1                   1          20

16         5            0                    1         21

run;

*Note: Count is number of patients hospitalized, and total count is total patients.*

I'm interested in knowing how the intervention affected the count, and also how it differed by status so I ran this model.

proc genmod data=x;
class status intervention/param=glm;
model count= time intervention status status*intervention intervention*time/dist=poisson offset=log_tc;
run;

*log_tc*= I took the log of the total count*

I get the coefficients from this output but I want to be able to interpret my output in terms of risk ratios/rate ratios.

However, using the estimate statements/LSMeans does not exactly work in this situation and I'm not sure what to do to get it since I have interaction terms in my model.

If I had just main effect terms in the model, I would take the exponents of the coefficients, but with interaction terms how would I get the estimates? and interpret all of them, including the interactions?

For instance I want to know how the count varied by status (=1) where intervention=1. 

I would appreciate any guidance on this!

[ChrisNZ]

Moved to STATS forum.

[StatDave · Accepted solution]

Your response is apparently binary and then aggregated, resulting in a count of events out of a total count of trials. As such you can simply fit a logistic model using "events/trials" syntax for specifying the observed counts. You can then obtain the risk ratio (relative risk) using the NLMEANS macro as discussed and illustrated in this note. 

[JME1]

Thank you!