Period effect within binary logistic regression


We would like to compare two different procedures used to treat some condition, but these were administered during different time periods (old treatment was at an earlier period: 2014-15, new treatment at later period: 2017-19).

As this is not a time-to-event study, we use a binary logistic regression. Is there any way, one can handle (quantify, adjust, etc), to some extent, the period effect?

Appreciate any ideas.


Time is completely confounded with treatment in your sample. There is no way to salvage that.

I guess this shall be a limitation we can’t take care of. Thank you Frank

That’s right. The data will not be usable for your purpose unless you can show from data and from much extra-data information that the time trend is flat.