Poisson regression and estimated percentage decrease in cases with 95% CI

[Martin Mueller]

Dear Statalist

I have a simple dataset with two observations and two variables only i.e. total number of patients that presented with chest pain during the Covid-19 lockdown in 2020 in our country and the number of patients presenting with chest pain at the same large emergency department (ED) one year before the lockdown at the same time period.

clear all
input chest_pain_number covid_lockdown
 120 1
 213 0
end

I am interested if the observed effect of reduced numbers is likely due to chance or an actual effect (that might be explained by multiple factors such as avoidance of an ED, reduced stress levels etc... and have to be further studied).

As the exposure group, our catchment area is likely to be the same in both time periods, I used poisson regression to get a 95% confidence interval (CI) and the incident rate ratio (IRR) with p-value.

poisson chest_pain_number i.covid_lockdown, irr cformat(%9.2g)

I obtain that the IRR is 0.56 (95% CI: 0.45-0.70, p<0.001), thus I conclude, that there is strong evidence against the hypothesis that the numbers in the two periods are the same.

I am further interested in the estimated percentual change (with 95%) from patients presenting with chest pain as the chief complaint in Covid-19 lockdown compared to one year before.
However, I cannot figure out the correct command. Probably, a margin option can do the trick?

Can someone help? Would you agree with my approach?

Best wishes & thanks
Martin

[Maarten Buis]

That is just a linear transformation of the IRR for covid_lockdown: (IRR-1)*100. So the p-value remains unchanged, and you can just apply that transformation to the coefficient and upper and lower bounds: -44% CI: -55%--30%

[Martin Mueller]

Thank you, Maarten!