Observational Health Data Sciences and Informatics

This is an old revision of the document!

---

CDM Multi-Level Associations

• Proposer: Clair Blacketer

Proposal

Add a table that allows for two or more levels of association or linkage between CONCEPT_IDs. This would mostly benefit data sources that are not claims or EHR based. The specific use case I am thinking of is related to the NHANES dataset; a U.S. based survey that combines both health and nutrition information. I cannot currently store the nutrition information effectively because all the CDM tables have one level of association. Adding in a table with two or more levels would allow me to use the nutritional information effectively. For example:

       Patient_id -> Date of Meal -> Food item consumed -> Calories in Food item
       Patient_id -> Date of Meal -> Food item consumed -> Grams of Fat in Food 
       

This file was taken straight from the NHANES website and shows how the dietary data are constructed. Using the current CDM and assuming they would be routed to the OBSERVATION table 'Respondent sequence number' is PERSON_ID and then 'USDA food code' could be mapped to OBSERVATION_CONCEPT_ID (this also assumes that the food codes would be added to the vocabulary. If not, they food code would go in OBSERVATION_SOURCE_VALUE and OBSERVATION_CONCEPT_ID would be mapped to 0). The problem with this is then I would only be able to see what foods a person ate on a day but I would not know anything about those foods (grams, calories, protein, etc.)