OHDSI Home | Forums | Wiki | Github

Themis: subgroup: measurement, procedure

I decided to start a separate thread for the subgroup.

This thread reports results of one of the subgroups of themis initiative. Namely the Measurement (and few other tables)

Key link is https://github.com/OHDSI/StudyProtocolSandbox/blob/master/themis/

I am pleased to say that we have data for 2 sites as of now.

Please report issues with the study to this new separate thread.

To start - I am replying to this post:

Yes. That would be both great contributions.
We hope to arrive at set of top labs and their most common units. (among other things)
.

I am pleased to say that we now have data from 2 sites (and 8+ datasets). We are still looking for more real world data. (EHR sites). @Rijnbeek Sites with multiple datasets are preferred. @Patrick_Ryan

Here is a peak:

(the usual suspects show up)

Looks like “real world” agrees most on measuring weight in kg.

I’ll run this for Janssen. @Vojtech_Huser, is there a specific version of the athena concept.csv file that I should use?

1 Like

Hi Vojtech, we are interested. We have only recently started to map to the OMOP CDM. Can we participate in this study without performing the full ETL?

Interim results for one themis substudy (aggregated for the network) can be found at this link https://github.com/OHDSI/StudyProtocolSandbox/blob/master/themis/extras/partial_results/units-with-tests.csv

The next step (besides more input from the network) is to consider how Achilles Heel can help with facilitating better measurement data given a RWD-driven convention.

Based on the themis group suggestions – I have revised the query (for themis network study) to include measurements with unmapped units to be part of the extract.
(revised notes.md file with SQL code) https://github.com/OHDSI/StudyProtocolSandbox/blob/master/themis/extras/notes.md

I posted updated results (we have 5+ datasets now!) to this folder (that I described during the meeting) (filtered to rows with n>=2)

We still could benefit from more sites (e.g., from EU, Japan). Please consider joining the study.

@Vojtech_Huser
Thank you for great work!
As I mentioned earlier, I plan to release the top 100~ 300 measurement/observation concept_id s. How can my work help you?

This recent paper http://www.ncbi.nlm.nih.gov/pubmed/29369797 by ColumbiaU group provides nice examples of labs we would expect to see in the resulting ThemisUnits database.

1 Like

@Vojtech_Huser
This is Hyunki Woo from Samsung Medical Center in Korea.
Currently our team is mapping the laboratory measurement of our hospital to LOINC code.

We have plan to release the top measurement code list (LOINC).
(plan to release sequentially rather than total release at one time)

I hope this work contributes to the OHDSI network.

What do you think about this idea?
Could you give me some guidance on how to release the results?

There are several updates on the ThemisMeasurements study (and ThemisUnits).

  1. AMIA poster was accepted
  2. Last week we met with Arachne creators and we are improving the code for the study to run in Arachne. (e.g., hint: use extension .ohdsi.sql to make your SQL analysis be recognized as parametized SQL)
  3. We are planning implementation to Achilles Heel (and Achilles Themis?)
  4. SqlRender is struggling with subqueries that have top X logic (issue filed https://github.com/OHDSI/SqlRender/issues/126 )

See study at https://github.com/OHDSI/StudyProtocolSandbox/blob/master/themis/extras/notes.md

ThemisUnits and ThemisMeasurements updates

  1. preferred single units for 375 tests can be viewed here StudyProtocolSandbox/S4-preferred_units-A.csv at master · OHDSI/StudyProtocolSandbox · GitHub
    (we are working of more preferred units (e.g., if lab is present in 4 sites and there is a “tie” (2 sites have cm and 2 sites have m for height). The ratios will decide the tie.

  2. there are manuscript updates for the larger publication

  3. updated SQL code for the network study is now working in Arachne !!! Use the study named. You can just install Arachne localy and not expose your dataset at all on the OHDSI Arachne server. Simply import the study analysis file (MAIN analysis) from the OHDSI public Archne server to your local server. To see the study and results on SynPuf -simply go to https://www.arachnenetwork.com/ and request a username (if you have Athena login you can use it; the same login works for both)

1 Like

Hi all,

Actually, LOINC itself gives us the default units of measurement for each test if available.

For example,
https://search.loinc.org/searchLOINC/search.zul?query=body+weight
“exUCUMunits”, “exUnits” are the fields with these units.

We can add them to the vocabulary relationship in some of the next releases.

What do you think?

t