Is immortal time bias an issue for outcomes not related to mortality?
If I understand your question correctly, you are thinking about a situation where say the outcome is cancer progression rather than cancer death, and we are examining a similar question to that in the lectures (e.g., post-op ASA use among individuals treated with surgery for CRC, focusing on progression of their CRC). In this scenario, the sequence of events could be: T0 is post-op for CRC, a subject might develop recurrence/progression 1 yr post-op, then you look at their exposure data from a survey obtained 18m after surgery. You make the assumption that usage of ASA doesn't change much over time, so you assign the person as a user back at T0, and the person counts in your analysis as a "user" or "exposed" until the point of recurrence. Is this what you mean?
I think this still violates principles of good study design bc technically, your exposure is now occurring AFTER your outcome, and could have been influenced by that event... I'm not sure it's creating the same Immortal time bias scenario (since you do allow the person to experience the event before the survey is returned and don't force her to "live" until the survey is returned), but it's still creating an uninterpretable situation, as now exposure could come after outcome. I'm not sure we'd call that immortal time bias, but it's is creating a bias... maybe it's more of a "reverse causation" situation.
Great, this is helpful and on point with what I was thinking. Thank you.