I have the following (fictitious) output from a Poisson model (count ~ AgeGroup + other variables):

                  IRR    %95 CI
Age Group 20-39   1.4    1.5, 1.7
Age Group 40-59   1.5    1.4, 1.6
Age Group 60+     1.8    1.5, 2.0

Reference age group is < 20.

I'm writing a conference abstract. I know I should always report the confidence intervals along with the estimates, but the word count is tight (250).

Among the results I'm interested in reporting is the general contrast in incidence between those < 20 (the reference group) with all of those >= 20 together.

I still need to keep the detailed age groups in the model for the purpose of the full results.

Is it acceptable to say something like: "People aged 20 or more have in general higher incidence" and report the lower and higher IRRs for age groups without reporting CI's (i.e. "IRR's ranged from 1.4 to 1.8")?

Or is it better to report the smallest lowest CI limit and the greatest highest limit for the age groups (i.e. "CI ranged from 1.5 to 2.0")?

Is there a better/safer approach to report, in numbers, the contrast of interest with a minimal number of words? Should I redo the analysis on combined age groups (<20, 20+)?


This may vary between fields. In economics, we would not report the confidence interval, but we would put the key word: "significant". That is: "Incidence is significantly higher for people older than 20"

If you are in the medical field, I have the feeling the convention is indeed to report the confidence interval. In this case, I would say that the best/ideal approach is to re-do the analysis on combined age groups (<20, 20+), to put it somewhere in the paper, and to report something like: "People aged 20 or more have in general a higher incidence (IRR = 1.6 ; 95% CI 1.4–1.9)"

PS: By googling it, IRR might be "incident rate ratio", but this is not something obvious to your reader. For my suggestion, I supposed it has already been defined in the abstract - or, like "CI", it is obvious to your reader in your field.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.