Discrepancy with stratified incidence estimates

Researchers
1

Dear all,

My team is carrying out incidence estimation in ATLAS and a question has come up regarding a discrepancy with the person-years for the incidence estimates. The person-time for the stratified estimates do not sum perfectly to the person-time for the overall incidence, and these numbers are far enough off that rounding cannot explain the discrepancies. For example, the overall person-years for one incidence was 516 years; whereas, for males it was 310 years and 216 years for females (total = 526 years). Is there an explanation for this difference? We see this across many of our incidence analyses.

Bortezomib_ex
Bortezomib_ex904×311 18.7 KB

Thank you very much for your help!

2

Defintely looks like a bug, but what is surprising to me is that the individual subgroups should be calculating time at risk in exactlyt he same way, just with a subset of the population. So it’s very odd that the N and cases sum up between Males and females to the overall total but the TAR does not. Should be running the same code except on a different subset of data. Strange! I will take a look at this.

Can you confirm which version of WebAPI/Atlas you are running?

Thank you for the report.

-Chris

3

Thank you so much Chris! We are running version 2.7.3.

Alexi

4

There has been updates to the incidence rate calculation that is in the 2.8 release:
https://github.com/OHDSI/WebAPI/commit/3c364e0d232982f4f72a50d6fa080ed0553b240c#diff-3e80d9d52e0ecfa7c4da735c93a6635e5e6332f8e2025be592c0e3ff3a7e0a5d

You may want to try to update.

Edit, yes, this is the issue that is describing your problem:
https://github.com/OHDSI/WebAPI/issues/1307

5

Thank you Chris! We will update to the newest release.