Problems about extracting pregnancy patients from MIMIC-IV v2.2

[YixiZhang001]

Prerequisites

• Put an X between the brackets on this line if you have done all of the following:
  • Checked the online documentation: https://mimic.mit.edu/
  • Checked that your issue isn't already addressed: https://github.com/MIT-LCP/mimic-code/issues?utf8=%E2%9C%93&q=

Description

Greetings! I'm trying to extract pregnant patients diagnosed with particular disease and I have already tried using itemid=225082 (chartevents) to identify those potential pregnant individuals. However, I got puzzled about the corresponding value. According to previous issue proposed by others (#612), could I equal value = 0 to "no pregnancy" and value = 1 to "indeed pregnancy"?

I also noted that chartevents reside in ICU module. Thus, is it possible that some pregnant patients were only documented in hosp but could not be identified by chartevents? In this reason, I'm wondering if there exists other methods to identify pregnancy? I noted your practical response in #1724 but the resulted sample size was still rather limited.

Thanks for your time and support.

[alistairewj]

Greetings! I'm trying to extract pregnant patients diagnosed with particular disease and I have already tried using itemid=225082 (chartevents) to identify those potential pregnant individuals. However, I got puzzled about the corresponding value. According to previous issue proposed by others (#612), could I equal value = 0 to "no pregnancy" and value = 1 to "indeed pregnancy"?

Most patients aren't pregnant, so I wouldn't try to use documentation of 0, since it won't be routine procedure to document it. Instead I'd just assume lack of documentation indicates no pregnancy.

I also noted that chartevents reside in ICU module. Thus, is it possible that some pregnant patients were only documented in hosp but could not be identified by chartevents? In this reason, I'm wondering if there exists other methods to identify pregnancy? I noted your practical response in #1724 but the resulted sample size was still rather limited.

Yes, very possible. I would investigate some combination of ICD diagnosis codes and discharge summary parsing would work (though I think I already suggested it). Keep in mind we would expect this population to be small. Maybe check a national database to get an idea of what proportion of ICU patients are pregnant .. my hunch is it's very small.