add function/logic to pull all measurements for a given codeset and make a derived column

[justaddcoffee]

For the AKI classification ML pipeline we obviously need AKI outcome data. We decide to spend some time today (2021-11-04) looking at how the “AKI in hospital” column in the patient severity table is generated. The logic is spread across a few workbook, but it is possible to follow the trail from OMOP to that column. If we understood correctly, the code pulls all relevant creatinine measurements for each COVID patient from the measurement table, then decides whether the patient had AKI using those data.

We were thinking this might be a good opportunity to pull out and abstract this logic and put it into Template 1A OMOP to cohort, so that we could pull measurements for things other than creatinine, and infer things other than AKI. Essentially this would be a function/template/chunk of code that would take parameters like a codeset to define the measurement of interest, begin and end date of interest, etc and output the measurements

[justaddcoffee]

@JohannaLoomba maybe can comment about whether this is feasible, and how we might implement this in a pattern that is consistent with the existing Template 1A

[justaddcoffee]

@bryanlaraway let's discuss further here

[justaddcoffee]

@bryanlaraway Starting with Bryan's hypertension notebook:

Parameters for function/logic:

Input table of interest:

• exactly one row per patient
• person ID column

OMOP source tables - any/all of:

• Visit Occurrence
• Condition Occurrence
• Procedure Occurrence
• Drug Exposure
• Measurement
• Observation

For each table:

• codeset/string
• table name
• how to reduce cardinality to one (mean, most recent, highest)

Optional flag

• include concept ID descendants (boolean)

Output column name

• value (boolean)
• evidence (table) (optional)