Survey weights (statistical weights) in CDM

Implementers
1

For the Nationwide Inpatient Sample and NHANES and other national survey data, there are statistical weights that we would like to store in the CDM along with the data (sometimes there is more than 1). Has anyone done this already? If so, did you put the weight in the Observation table? Did you find a usable LOINC code? I would be interested in what others may have done to get this data into the CDM.

Thanks.

2

Hi Mark: For NHANES, we put the weights in the OBSERVATION table. For
HCUP, we didn’t do it, but should have and will in a future ETL revision,
since we needed it for a recent analysis. We did not standardize the
weights to a common CONCEPT_ID, but if we reached a community consensus
around that, I think it’d be wonderful.

3

Can we use UMLS concept IDs? If so, they have one that might fit the bill:
C0237953 = statistical weight

Not sure what the process is for adding these things, or whether it would be better to ask LOINC to add to their vocabulary.

4

Mark,

Like Patrick, I put survey weights into the Observation table and mapped them to the following concept:

4178502,Body weight measure,Observation,SNOMED,Observable Entity,S,363808001,19700101,20991231,

Bill

5

Thanks for the response @wstephens! However, I believe that CONCEPT_ID is specific to a patient’s body weight. The statistical weight @Mark_Danese is referring to is from surveys. It is more like a weight assigned to a record to help create national estimates. For example, if the survey administrators over-sampled a particular under-represented population, like cancer patients, any record with a cancer patient will have a smaller weight assigned to it, compared to other non-cancer patient records, to help create accurate national estimates with proper population proportions.

I found the following CONCEPT_IDs with a domain = Observation:
44804059 - Office of National Statistics area classification - National Public Health Classification
44803895 - National statistics socioeconomic classification - National Public Health Classification
44803896 - Nomenclature of units for territorial statistics - National Public Health Classification
44804316 - Statistical data - National Public Health Classification
4260191 - Population statistic

These are not that specific. But they might work? Thoughts? I’m also tagging @Christian_Reich as we are having a similar discussion on the following thread: observation_type_concept_id for HCUP-NIS ETL

6

Jen,

Yeah, I read through the HCUP doc and found that out. I had a similar issue in attempting to conceptually map child height and weight growth percentages according to the WHO standards. WHO Z-score concepts are missing, too.

Bill

7

Friends:

Let’s discuss this here and close this one.