OHDSI Home | Forums | Wiki | Github

Inconsistent Vocabulary Options for Observation


So we have items we are trying to map and I am seeing inconsistencies in the vocabulary where there is a Yes and No option for chew tobacco.

For example, there is 4036559 for does not chew tobacco and 4218741 for chews tobacco. And for alcohol use, there is No or denies alcohol use 45772705 and 45766930 for admits alcohol use.

However, there is not a ‘no’ option for:

cigar smoker
pipe smoker
intravenous drug user
illicit drug user

Vocabulary is inconsistent with having some items for no and some items for yes. Requesting to add ‘No’ concepts in the observation domain for the above 5.

(Anna Ostropolets) #2

You brought up a great point. @MPhilofsky has also mentioned it a while ago.
We do not have a standard way to represent smoking (drug users may be another but less common group).
The vocabulary team has been thinking of the following design:
To have an artificial pre-coordinated scheme that will allow you to classify any tobacco/marijuana/ other substances based on the 4 axis:

  1. Substance (tobacco/marijuana etc)
  2. Type (pipe, cigarette, hookah etc )
  3. Amount (1/day, 10-20/day etc)
  4. Time (year, 5 years etc)
    Each axis will support synonyms to facilitate ETL (e.g. >20/day = heavy smoker) and will have an internal hierarchy so that you can map the concepts with any granularity (e.g. nicotine substance -> tobacco-> loose leaf tobacco).

The end product will look like a permutation of these axes, e.g:
Consumption of tobacco, cigarette, >20/day for 5 years

Probably we will also include ‘unknown’ attributes like:
Consumption of tobacco, unknown type, duration and amount.

All the SNOMED concepts will subsequently be mapped to the corresponding terms within this scheme.

That’s the preliminary plan (@Christian_Reich may be willing to add more).

(Melanie Philofsky) #3

I like the plan, @aostropolets and @Christian_Reich!


Would this also include ‘No’ answers like denies tobacco use or does not use tobacco?

(Christian Reich) #5

That’s what we discussed.

We may not need that. Just use the combination of the attribute of the axes you do have information on.

Causes immediate bellyache, but in this case we may have to bite the bullet.

(Don Torok) #6

How will the 4 axis be implemented. For example will an observation record have an observation concept id indicating ‘Substance’ and then value as concept indicate the what substance?

(Anna Ostropolets) #7

No, just one concept_id to be put in observation_concept_id

(Don Torok) #8

Then is that one concept a combination of the 4 axes or will it require and observation record for each axis.

(Anna Ostropolets) #9

It will be one pre-coordinated concept_id with a combination of 4 axes as in the example above.
Or, rather, the plan is to make it this way. We are open to suggestions :slight_smile: