OHDSI Home | Forums | Wiki | Github

Help Request in MIMIC-IV to OMOP CDM conversion

General
#1

Hi guys! I’ve been working on the OMOP CDM conversion of MIMIC-IV recently, based on the GitHub repo (https://github.com/OHDSI/MIMIC). I have just finished the conversion and realized that my Measurement table nor some of the Observation table are not mapped. I acknowledge that this is a significant issue as these two tables are very important for a lot of research applications. I have checked that voc_concept table does indeed contain concept_ids for the two tables (Measurement and Observation), but I am not sure why this error is happening.

On another note, it says on the repository that “POC_3 data is loaded and prepared manually (todo: provide both manual scripts)” for waveform_collecting. I don’t have the manual scripts for these. Is this the potential reason why some of my final tables are incorrect? If so, could someone please provide the POC_data scripts?

Thank you so much.

1 Like
#2

Hi @jiwoonoh - glad to hear about your MIMIC-IV work. Are you saying that your measurement table comes back empty? That is indeed unexpected. Can you elaborate?

In this repository we made a small “proof of concept” for data derived from waveform recordings (blood pressure, ECG monitor), so that we can show one possible approach for that. However, I would recommend to skip that part, if you try out for yourself. @mkwong has worked (and is working) on that waveform data and could provide you with more background.

Cheers - mik

#3

If you are working with the MIMIC-IV WFDB data, I can provide you some mapping I’ve been using for both the CSV and derived measurements from the waveforms. Regardless of using the clinical vs WFDB data - you should have data that can map into measurements and observations. Please elaborate what data you are trying to convert as well as specific examples if possible.

1 Like
#4

Thank you for your reply! I’m assuming that waveform collecting process doesn’t really have any connections with the broken Measurement table.

I should’ve elaborated better – all of the concept_ids for my Measurement table is not populated correctly (nor some of the concept_ids for observation table). I’ve realized that my lk_measurement_mapped table also has concept_ids of 0 (after populating data from lk_measurement_clean tables), so it might be a problem somewhere around there. All the other tables are fine, but it’s just that all my measurement table has concept_id of zero, and it says the concept_ids are ‘unmapped’ as well (although I’m sure I ran all the processes well). Could you please provide any potential reason for this problem and a solution for it? Thank you so much.

image
image1228×1000 237 KB

#5

I’m just using the full MIMIC-IV data. In that case, do you mean that I won’t need to run the waveform collecting process?

#7

If you are loading the “MIMIC-IV data”, typically that means the ICU patient EHR care data. Odd you are getting measurement.measurement_concept_id = 0. There must be a problem in the load ETL if its populating the other fields. Anyway - you do not then have to concern yourself with the MIMIC-IV WFDB data.

#8

image
image1714×1048 138 KB

Yes, this is what I get from cdm_measurement table. I’m not sure what problems there could be. I am quite in rush with the data – is it possible for me to obtain the correct full data just for measurement and observation tables?

#9

Sorry for jumping into the discussion. I’m just curious: does it mean we’re able to map our obtained MIMIC-IV dataset into OMOP-CDM? I’m probably interested in mapping my data too.

#10

Yes, it’s definitely possible!

1 Like
#11

Hi, guys. I’ve been working on the OMOP CDM conversion of MIMIC-III recently, based on the GitHub repo (GitHub - MIT-LCP/mimic-omop: Mapping the MIMIC-III database to the OMOP schema). At one step, it required me to have NIH licence (which I have no idea how to get sooner) I have applied a forum and all. I am a researcher in Computing Science and would be very helpful to get recommendation on how to fasten up the process of convesion.
Is it possible to covert with some easier way?

#12

Hi Divya. I implemented that MIMIC-III code with some modifications and it worked well. Let me pull out my notes and look at what I did. Was a couple years ago. Will be back in touch, Gigi

#13

That will be really helpful :slight_smile:

t