Create cohort-based standard table 1

[schuemie]

The current createTable1() function relies on concept-based covariates. We should create an alternative table 1 based on cohorts. For example, instead of 'Diabetes type 2' being based on the occurrence of a single condition concept, it should be based on a more complex definition, possibly involving antidiabetes treatments, lab values, etc.

I propose to use the same set of items as the current table 1 (see specs here), since those were selected before to be 'good enough'. For each item (except possible demographics) we should find a corresponding cohort definition. We could then

1. load those cohort definitions into FeatureExtraction.
2. add a function in FeatureExtraction to retrieve all these cohort definitions (in a format suitable for CohortGenerator),
3. add a function for creating cohort-based covariates specifically for this set of cohorts.
4. adapt the createTable1() function for working with this new table 1.

@gowthamrao: I'd imagine the cohort definitions would come from the OHDSI Phenotype Library.

Also looping in @pbr6cornell .

[schuemie]

I have extracted the concepts used in Table 1 here, so people know which cohort definitions we need.

[gowthamrao]

Yes - here is a link to candidate cohorts

https://docs.google.com/spreadsheets/d/1pdGYoH2bwr_fjFvvAWUiSJIVzY1_7FTdfkDHq1sO5dk/edit?usp=sharing

(updated link)

[gowthamrao]

The version 3.1.1 of PhenotypeLibrary R package has the above referenced candidate cohorts. We can use those cohorts with FeatureExtraction to proof test this idea.

[schuemie]

Hi @gowthamrao. The Excel sheet you referenced contains a lot of entries labeled 'not started' or status = 'yes'. Would it be possible to create a clean table. where one column contains the cohort IDs as found the PhenotypeLibrary package?

[gowthamrao]

@schuemie yes

Since building and maintaining these cohorts are now the responsibility of OHDSI Phenotype Development and Evaluation workgroup using the OHDSI Peer review process - i suggest we move this issue to https://github.com/OHDSI/PhenotypeLibrary

We can circle back to technical implementation once peer reviewed cohorts are available

Thoughts?

[gowthamrao]

FYI - Kevin Haynes and @gowthamrao are revisiting this effort now

[anthonysena]

I think we'll want to get #167 merged into FeatureExtraction and use that work to help drive the Table 1 functionality described in this issue.

I suppose once this is in place, we could potentially have a dependency on the PhenotypeLibrary (PL) package such that defining the Table 1 specs simply references the cohort(s) in the PL. Hoping this is what you had in mind @gowthamrao but let me know if I've misunderstood here.