Adjusting for Year of Event in Logistic Regression - how to account for incomplete year.

[David Motorniak]

Hello all,

Hopefully this question interests someone enough!

I am new to Stata. I have been working on a dataset of Disease X in the community and have created a multivariate logistic regression to look at adjusted ORs towards survival.

I am keen to incorporate and adjust for the year of diagnosis in my logistic regression, as the current literature tells us that the pandemic years and lockdown influenced rates of this disease.

My issue:
- My dataset begins collecting data from start of April 2019.
- My dataset completed collecting data in end of May 2023.
- So if I try and assign a value to each observation for calendar year, I will inevitably underweigh 2019 and 2023.

Any solutions to get around this?

Thank you!

[Joseph Coveney]

- My dataset begins collecting data from start of April 2019.
- My dataset completed collecting data in end of May 2023.
- So if I try and assign a value to each observation for calendar year, I will inevitably underweigh 2019 and 2023.

Any solutions to get around this?

If your dataset allows, then perhaps you could consider assigning values to a more finely measured interval, for example, months or quarters, instead of years.

On the other hand, although it's not completely clear from your description, it seems that all of the communities were observed in parallel, that is, their temporal exposure was the same. So, depending upon the question you're trying to answer (not mentioned), it might not matter.

If you're looking at survival time since a varying time point of diagnosis, that is, a widely varying observation interval (time-at-risk), then would a binomial likelihood be the best way to model the data-generating process?