Open isangil opened 2 years ago
I'll let @KJDonovan-RS21 respond but my rec would be to use null
whenever possible and reserve 0
only for cases where 0
represents an actual code. Looking at the OMOP vocabulary there are only 6 or so entries where 0
is a valid source code so it's really unlikely that we'd see it in the extracts
| a | b | c | d |
| ---- | ----------- | ------------------ | ---------------------------------------------------------------------------- |
| 0 | Procedure | ICD10PCS | Medical and Surgical (Procedure) |
| 00 | Procedure | ICD10PCS | Medical and Surgical @ Central Nervous System and Cranial Nerves (Procedure) |
| 00 | Procedure | ICD9Proc | Procedures and interventions, Not Elsewhere Classified |
| 00.0 | Procedure | ICD9Proc | Therapeutic ultrasound |
| 0 | Observation | MDC | Ungroupable |
| 0 | Visit | UB04 Pt dis status | Unknown Value (but present in data) |
| 0 | Observation | UB04 Typ bill | Non-Payment/Zero |
If no analogous source value exists, just cast as NULL as SOME_COLUMN_NAME.
kk- will put '' (nulls). The zeros are probably inspired on the OMOP strategy on putting zeros when there is no mapping. -- but important is that we have a way forward here - make'em nulls
Original code sets two fields hard coded to zero. Is this correct? ,0 AS OBSERVATION_SOURCE_CONCEPT_ID ,0 AS UNIT_SOURCE_VALUE
OMOP best practices asks for zeroes under circunstances, but these are pre-omop, (LDS).