Calculating confidence intervals for predicted incidence rates

[Joe Ward]

Hello all
I am using data on counts of children with abscesses per day within each year.

Using the trend in the years prior to the pandemic (2013-2019), I would like to predict the incidence of these events in the years during and since the pandemic (2020-2022), to see if the observed incidence now is following the pre-pandemic trend.

Because there is variation in the number of days of observation each year I have created the variable "counts_per_day"

I have used national population data to offset the model

Firstly i have run

poisson event_per_day year if year < 2020, exposure(population)

then

predict abscess_ir, ir

This gives the incidence rate for all years (abscess_ir). I would like to plot this with 95% confidence intervals, - does anyone know how to calculate confidence intervals for this?

Any help would be fantastic!

Best wishes

Joe

[Joe Ward]

Graph of counts per day per 100,000.pdf

Attached is an example of what I would like to create - with observed counts per 100,000 in green and predicted counts per 100,000 as the line with grey confidence intervals.

I created this using

gen count_per_day_incidence = (event_per_day/ population) * 100000

twoway lfitci count_per_day_incidence year if (year < 2020), range(2013 2023) || scatter (count_per_day_incidence year), ytitle("counts per day per 100,000") ylabel(, angle(horizontal)) xscale(range(2013 2023)) xlabel(#10) xmtick(none) legend(off) graphregion(fcolor(white) ifcolor(white))

But I understand lfitci is fitting a linear model, wheras I ideally would like to fit a poisson model to these data

Thanks again!

Joe