search

PRiSMA-Study / PRISMA-Public

PRISMA public repository where files will be shared with individuals/sites.
GNU General Public License v3.0
0 stars 4 forks source link

Weird skip pattern #41

Open Sfomalley opened 2 months ago

Sfomalley commented 2 months ago

https://github.com/PRiSMA-Study/PRISMA-Public/blob/e50fb738efbe51ea633bc854902cec39bedfdf8a/PRISMA-Maternal-Constructed-Outcomes/Infection/Maternal-Infection-Variables.R#L212

Not sure if this matters, but the skip pattern for M04_HIV_EVER_MHOCCUR doesn't seem to be applied correctly by sites (it's mostly at Ghana, but also at CMC and SAS) So participants are getting asked this question even if it's not an enrollment visit

sloisate2 commented 2 months ago

I think this is OK since these are all "or" statements + each of the variables have already been filtered for the enrollment visit so there shouldn't be a risk of double counting. Are you seeing instances in Ghana where this wasn't answered at enrollment but was answered in a subsequent visit?

Sfomalley commented 2 months ago

Confirming this data set is only type_visit==1 ? I'm comparing to line 162 which looks like visit types 1-5 but I probably just missed it!!

I'm seeing some situations where the mother is being asked this question repeatedly even if she's already answered it. Some women are asked up to 4 times during pregnancy. Currently each of these situations is HIV negative although who knows in the future data uploads.

But you're right, probably not be an issue with all "or" statements if it's restricted to enrollment visits only.

sloisate2 commented 2 months ago

Oops you are so right - not sure what I was thinking there! These datasets allow for any timepoint during ANC, but the enrollment specific infections require VISIT_TYPE =1. Just did a quick crosstab and it does look like sites are asking this question at several visits but thankfully the only reports of M04_HIV_EVER_MHOCCUR=1 occur at enrollment.

I'll be sure to keep my eye out on this one