Modelling excess mortality using Poisson regression

[Nina Afshar]

Hi Paul and everyone,

I am working on a project in which I am interested in modeling excess mortality using Poisson regression.
We usually do this as follows (according to "Estimating and modeling relative survival" by Paul Dickman et al, 2015):

use colon, clear
stset surv_mm, failure(status==1 2) id(id) scale(12)

**Estimating relative survival (Ederer II method)
strs using popmort if stage==1, breaks(0(1)5) mergeby(_year sex _age) ///
by(sex year8594 agegrp) save(replace) notable noshow

**Modeling excess mortality using grouped data
use grouped, clear
glm d i.end i.sex i.year8594 i.agegrp, fam(poisson) link(rs d_star) lnoffset(y) ///
eform nocons nolog baselevels


The codes for modelling excess mortality is only available if we use 'strs' command in Stata i.e. use Ederer II method. What if we use the new method of estimating net survival (Pohar-Perme method)?

stnet using popmort if stage==1, mergeby(_year sex _age) breaks(0(.083333333)5) ///
diagdate(dx) birthdate(dob) standstrata(agegrp) list(n d cns locns upcns secns) listyearly by(sex)

Does anyone know how to model excess mortality when using 'stns' command? I tried to do this by saving the results of net survival analysis similar to what we do for relative survival although the results will not be saved as individual and grouped data. It was not possible to run 'glm' command in this case because there is not appropriate '.ado' file in Stata to link, similar to the existing one 'rs.ado' that we used in the above command link(rs d_star). I was wondering if it is appropriate to report both net survival (Pohar_perme) and excess mortality rate results when publishing a paper considering that Phoar-Perme gives us the weighted estimates while Ederer II does not and we model our excess mortality in frame work of Ederer II (relative survival).

Thank you in advance for your help.

Regards,

Nina

[Nina Afshar]

Thanks to Paul. "The short answer is that we cannot model the Pohar Perme estimates. So, it would be ok to report both net survival and excess mortality rate results."

[Dart Stater]

Excess Relative Risk (ERR) per unit of dose is commonly used in radiation epidemiology (an example of the study can be found on this link) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC558612/
Although specific statistical instruments are used to obtain ERR's, it seems intuitive that ERR calculation can be performed using STATA as well.
Poisson regression realized in Stata seems to be the appropriate tool, but it is unclear for me how to separate the background (baseline?) risk and excess risk to get the ERR per unit of radiation dose.