search

EpiModel / CombPrevNet

The Role of HIV Partner Services in the Modern Biomedical HIV Prevention Era
GNU General Public License v3.0
2 stars 0 forks source link

partner services precisions #59

Closed AdrienLeGuillou closed 2 years ago

AdrienLeGuillou commented 3 years ago

@smjenness following up on our discussion about partner services probabilities, as the code does not reflect our discussion I will outline the dilemma here

I will use part.hiv.test.rate as an example, with the target value of eligible for screening (identified and without a diagnosis) should be 39.4% in Atlanta.

At time-step at are eligible for partner screening an individual without an hiv diagnosis that was identified at at (https://github.com/EpiModel/EpiModelHIV-p/blob/CombPrevNet/R/mod.hivtest.R#L54)

This means that and identified partner has only one chance of being screened, upon identification. If I understood correctly, that's not what you had in mind. I thought you wanted an identified partner to be eligible multiple time-steps after identification. Should I modify the code to reflect this?

Also, the time since last negative test tsincelntst is not used for identified partner. This means that an individual can potentially be identified and screened twice in a row. Is this the desired behavior?

The other services behaves similarly.

The advantage of the current code is it's simplicity. Upon identification you are tested or not and that's it. part.hiv.test.rate is thus 0.394. The downside is that everything starts at once: upon identification the partner is either screened or not and if so he can potentially start PrEP or ART (depending on the screening result) all in one time step.

Changing the code to allow for partner services to start later will imply calculating the rates (like part.hiv.test.rate) over a window of time (we said 4 weeks).

Tell me what you want me to do about this

smjenness commented 3 years ago

@AdrienLeGuillou : thanks for outlining this issue. I think the most straightforward way is to handle the events within the time (and not have them time-extended across time steps). Are PrEP and ART reengagement also currently set up like this in the code? While not perfect, it probably won't make a huge difference to the overall epi dynamics whether X number of people start a service on Week 1 or X/4 start on each of Weeks 1, 2, 3, 4.... Does that make sense and do you agree with that? Does this decision make the coding simplified (I believe you say yes above)? Thanks!

AdrienLeGuillou commented 3 years ago

It is setup this way for all partner services.

I don't think it will make a huge difference either and the code will definitely be simpler.