field_fkclass_drug_strength_ingredient_concept_id does not consider "Precise Ingredient" concept_class_id

[james-cockayne]

Hi,

We've started to see failures today in our OMOP 5.4 database for the new field_fkclass_drug_strength_ingredient_concept_id rule.

Checking the query, it defines violating rules with the following query

SELECT 
    'DRUG_STRENGTH.INGREDIENT_CONCEPT_ID' AS violating_field,
    cdmTable.* 
FROM cdm.DRUG_STRENGTH cdmTable
    LEFT JOIN cdm.concept co
        ON cdmTable.INGREDIENT_CONCEPT_ID = co.concept_id
    
WHERE co.concept_id != 0 
    AND (co.concept_class_id != 'Ingredient') 
/*violatedRowsEnd*/

On inspection of our violated rows, I can see all of our invalid rows refer to Precise Ingredient rather than just Ingredient.

Looking at the 5.4 specification, I think that category probably should be included too

We are using the following vocabularies in our deployment https://answerdigital.github.io/oxford-omop-data-mapper/docs/database-setup.html

I had a quick look at the code to see if I could put a PR in for you, but it seems to be templated code and I'm not quite sure in which way you would want to alter it. We might need another field_fk_class.sql that supports a collection of allowed classes.

James

[katy-sadowski]

@clairblacketer, do you know if the issue here is with the DRUG_STRENGTH table from the vocabulary (i.e., concepts with the wrong class are being used in that table), or if we should also allow Precise Ingredient as an fkClass for this column in DQD?

@james-cockayne, the reason this just started failing for you is that due to an issue with how check statuses were calculated, this failure was previously categorized as Not Applicable. i am seeing the same in my CDMs (since we all use the same DRUG_STRENGTH table 😃 ). feel free to disable the fkClass check on this field in your threshold file for now if you'd like the failure to go away until we decide on the solution!

[james-cockayne]

Thanks @katy-sadowski we will supress the rule for now.

[EhsanBitaraf]

I can confirm the same behavior in my environment.

Setup:

• CDM schema: phems_cdm
• Vocabulary schema: demo_cdm
• DRUG_STRENGTH exists in demo_cdm

DQD reports 3,642 violated rows out of 3,020,774 rows for field_fkclass_drug_strength_ingredient_concept_id, corresponding to 0.120565%.

When grouping the violated rows by concept_class_id, all violations are Precise Ingredient:

Precise Ingredient | 3642 | 0.120565%

Example row:

• drug_concept_id: 41430738
• drug_concept_name: 180 ML Sodium Oxybate 500 MG/ML Oral Solution Box of 1
• ingredient_concept_id: 769935
• ingredient_concept_name: sodium oxybate
• ingredient_concept_class_id: Precise Ingredient
• vocabulary_id: RxNorm
• domain_id: Drug

After changing the logic to allow both Ingredient and Precise Ingredient, the number of violated rows becomes 0:

num_violated_rows = 0
pct_violated_rows = 0
num_denominator_rows = 3020774

This seems to confirm that the current check is producing false positives because it only allows Ingredient and does not allow Precise Ingredient for DRUG_STRENGTH.INGREDIENT_CONCEPT_ID.