You should only put standard concepts in condition_occurrence, so if the SNOMED is the standard concept, then use the SNOMED concept. The concept_hierarchy table just tries to roll it up in a friendly way.
It does appear that there’s a dependency on MedDRA vocabulary to build the concept_hierarchy. If that’s a problem, could someone suggest a standard concept hiearchy rollup? This logic for building the concept_hiearchy into 4 distinct levels is very legacy code, and while works OK for things like drug (ingredient->ATC5-ATC3-ATC1) it is showing signs of stress in the condition domain…and procedures are even more unstructured. If we could get a standard definition for these 4-levels (and, requiring 4 levels is again an old ACHILLES reporting requirement) then we could get something that will work license-free.
-Chris