Initial federated COVID-rich ICU database documentation.

[danamouk]

This pulls request provides initial documentation to the tables and standard terminologies included as part of MIMIC-Fed, a large, freely available federated COVID-rich ICU database comprising deidentified health-related data from Beth Israel Deaconess Medical Center (BIDMC) and Northwestern Memorial HealthCare (NMHC).

The following tables are included as part of the hospital and icu modules in version 0.1.0 of the federated database:

• Hosp Module: Admissions, Patients, Labevents, D_labitems, Diagnoses_icd, D_icd_diagnosis, Prescriptions, Emar.
• ICU Module: ICU_stays, D_items , Chartevents (vital signs), Procedure events.

[bemoody]

MIMIC-Fed is a large, freely available federated COVID-rich ICU database

I'm not sure that "freely available" is a good description and would take those words out.

In any event this database isn't yet available at all. Please add the following at the top of the page (maybe it should be added to each sub-page too?)

{{% pageinfo %}}
MIMIC-Fed is not yet released and its structure is subject to change.
{{% /pageinfo %}}

It is not clear to me what the long-term relationship will be between this data and "MIMIC-IV" as we know it. I think it would be unfortunate to have two sets of documentation (that both have to be maintained and that would confuse visitors). At the same time we don't want to give the impression that the new COVID stuff is already integrated in MIMIC-IV.

In terms of the website structure, I'd suggest one of two things:

• Inside the MIMIC-IV top-level section, add a subsection (at the end, after Tutorials) labelled Federated (Beta).

• Add a new top-level section (at the end, after MIMIC-II) labelled Federated (Beta).

As this database isn't released (nor is it a replacement for MIMIC-IV), it shouldn't be given "top billing" above MIMIC-IV, and at the same time the hope is that eventually this will be part of MIMIC-IV (it's clearly not separate from MIMIC-IV in the way that MIMIC-III is.)

@tompollard how would you feel about these options?

Additionally, someone suggested that "Federated" might not be a good term since it has a specific technical meaning when talking about databases. Perhaps "Multi-center" could be used instead?

[danamouk]

thanks @bemoody for the suggestions. I have updated the changes detailed in the commits.

[danamouk]

@tompollard let us know if the changes look ok and we can deploy the changes on the website.

[danamouk]

After discussions with @rgmark, @bemoody, @tompollard and @alistairewj, I have restructured the mimic-northwestern multi-center database to the main docs page, and replaced mentions of "federated" with "multi-center" to avoid confusion with another technical term under the term "federated learning".

This is a version that tries to address most of the points raised, though I think the website could be reviewed for any further suggestions either on formatting or naming etc.

[alistairewj]

some comments here and there

[alistairewj]

all of the data for this subset should be ICD-10 - can't imagine NW using ICD-9 in 2020 onward - perhaps make it clear the description of ICD-9 is just for informational purposes?

[danamouk]

yes, added a clarifying statement to indicate ICD 9 codes are just for informational purposes, after mandate of ICD10 in 2015 based on CMS.

[alistairewj]

would be useful to add itemid and the LOINC for these examples

[danamouk]

updated!

[alistairewj]

hmm I don't see it? something wrong with my viewer or did the commit not go through?

[danamouk]

it is actually part of commit 2f8806e.

[danamouk]

After discussions with @rgmark regarding the MIMIC-NW project, MIMIC-NW: Harmonizing Northwestern's ICU Data with MIMIC's Data Structure, I will update the website to emphasize the MIMIC-NW project and adjust the release statement across the main page and modules accordingly. This version will serve as documentation for the data prior to its public release.