OMOP CDM Enhancements

[AndyRae]

Our ambition is for Carrot to support mapping the complete OMOP CDM, but this is a journey that will be delivered over time. We may need to update the Carrot CDM tool to support these as well. This is a tracking issue to track, discuss, and prioritise that progress.

Currently Carrot supports the following tables of the CDM:

• person
• measurement
• condition_occurrence
• observation
• drug_exposure
• procedure_occurrence
• specimen

We do not support the following tables, these are a priority:

• [ ] #518 / #646
• [x] #636

Lower priority as they don't really describe mapping:

• [ ] fact_relationship / #625

Likely that Carrot doesn't play a role here:

• [ ] note

• [ ] note_NLP

• [ ] episode

• [ ] episode_event

• [ ] observation_period

• [ ] visit_occurence*

• [ ] visit_detail*

• (complex, might not be done with Carrot, this is likely a pre/post-processing step)

We are open to the order in which we will undertake these, and would like feedback on which would be a priority from our community. Ideally we would tackle areas that are high impact, and low effort first - so help on identifying these is appreciated. If you would like us to focus on a specific area, please provide examples of Scan Reports, and mappings that we can work with to test against.

It is not as simple as "supporting" these tables, but also the intricacies for each table, and domain. For example we see these issues as part of our journey of the core functionality of Carrot, and are a current priority for us:

• [x] #647
• [x] #769
• [x] #648
• [x] #711
• [ ] #406

Domain Support

Currently we support a restricted set of domains, and provided a limited "opinions" of where these should map to. We will explore which specific domains are supported (this is seeded into the DB), and provide further configuration.

[rsippy]

Hello, I'm working with data that have a lot of variable/value pairs as described here. In short, the variable is explicitly paired to a value; in the OMOP CDM the variable has a relationship ("Has answer") with the value. Ours are primarily things like:

1. Mobility disability (variable), Yes or No (value)
2. Category of need (variable), N1 or N2 or N3 or N4 or N5 (value)
3. Test result (variable), Suspected or Not suspected or Declined (value)

Sometimes the variable is a Measurement, and sometimes an Observation (see the description link for more details on this). For these types of paired variable-value concepts, many of the value concepts should be in the Measured Values (I requested this table). However, I suspect that the paired nature of these variable-value concepts means that the fact_relationship table will also need to be implemented in Carrot-Mapper in order for these concepts to be represented.

In our data on Carrot-Mapper, I've attempted to map the variable to its concept, but in many cases have no where to map its value, so it's not clear to me if/how this works in the resulting mapping rules/JSON file.

I'm attaching a ScanReport with many variable-value concepts and Usagi mappings for these (where possible): health_values_mappings.csv health_variables_mappings.csv ScanReport_health_pairs.xlsx .