Poisson regression with internal standardisation

[Rebekka Bakke]

Hi,

I have a data set which contains the number of observations "patients", outcome "cancer" and several covariates "gender", "work" and "smoking". I would like to calculate the cancer rate for women with internal standardisation for work and smoking. How do I do this?

Any help would be greatly appreciated by a State-beginner.

Thanks!

[Ken Chui]

Welcome to Statalist.

Please follow the FAQ (http://www.statalist.org/forums/help) and use -dataex- to provide some sample data. They don't need to be the real data, just couple ten rows to show the structure and format is fine. That way users here can test the codes before posting it.

Two other things that the question can use some clarification:

• By "patients" do you mean you data are summary count data and not an individual level data?
• This is mostly for my own lack of knowledge. What did you mean by "internal standardization?" I have heard of standardization, which I am not sure why it is needed here given you have one sample and nothing else to standardize to. And I have never heard of it being "internal".

[Clyde Schechter]

Like, Ken Chui, I'm curious what "internal standardization" is. I'm an epidemiologist, and I work with the kind of data described in #1 all the time. I'm very familiar with direct and indirect standardization, but I've never heard of internal standardization.

I also strongly endorse Ken Chui's comment about using -dataex- to provide sample data. Frankly, without that, it's unlikely that anybody can help you. At best, somebody might attempt writing code that will work on imaginary data but will fail in your real data.

[Tim Hofer]

Not really Stata related but I recently was a reviewer for an article that used both sets of terms and in the revising process the authors put accessible definitions of all of them in the paper (tailored to SMRs). They did a good job I think and I quote from the published article below:
(I recommend the article as well if you use SMRs https://doi.org/10.1371/journal.pone.0257003 )

1) Direct vs. indirect standardization:

• Direct standardization applies the stratum-specific mortality rates of each hospital to the case mix of the same “reference”/“standard” hospital.
• In contrast, the SMR as a measure of indirect standardization applies stratum-specific expected mortality rates to the specific case mix of each hospital...

2) External vs. internal standardization: This distinction refers to the way in which the expected mortality rates are derived:

• External standardization: Expected mortality rates may be derived from data that is not included in the analysis of the hospitals under consideration, e.g. from a dataset of hospitals from a different geographical region. This approach is refereed to as external standardization.
• Internal standardization: Alternatively, expected mortality rates may be derived from the same dataset used to calculate the SMRs of the considered hospitals. In this case, the performance of the hospitals usually is evaluated against their average performance in terms of mortality rates. This approach is referred to as internal standardization.