Hearing Test in Observation or Measurement Domain?

Vocabulary Users
1

We are converting hearing test data from Epic into OMOP format and looking for a standard mapping strategy. Here are two questions we encountered so far:

  1. Should the thresholds of hearing test data mapped to observation domain or measurement domain? What’s the best practice to map these thresholds?

  • We found there are two vocabularies for hearing test, one is SNOMED (Threshold of hearing: Athena) and another is in LOINC (Pure tone threshold audiometry panel: Athena). The SNOMED has a hierarchy while LOINC are contained in a panel. The LOINC codes have 2 more frequencies than SNOMED codes, i.e. 750 Hz and 5000 Hz. Both vocabularies have the lateral info.

  • We also found in this paper Creating a health informatics data resource for hearing health research - PMC they mapped test outcomes from PTA to OMOP by having a [Procedure_Occurrence] record for each test performed. The most appropriate OMOP procedure concept (from the Systematized Nomenclature of Medicine - SNOMED - vocabulary) was used in each case (e.g. 4091134 = Pure tone audiometry, or 4091877 = Soundfield audiometry). Multiple records from the [Observation] table were then related to the [Procedure_Occurrence], to describe results for each frequency (and ear) tested. Concepts such as masking level, air conduction or bone conduction could then be related to the [Observation], whereas concepts such as headphone (worn over the ear) or inserts (worn in the ear) could be related to the [Procedure Occurrence]. Fact relationship tables were used to communicate these relations.

  1. The current vocabulary doesn’t cover all levels of hearing test frequencies found in our Epic instance. Is this a potential vocabulary update in SNOMED or LOINC?
  • From Epic, we saw multiple thresholds of hearing test which are not found in either SNOMED or LOINC. The bolded frequencies are missing from the vocabulary. Will they be added into the vocabulary?
SOUND_FREQ_C
125
250
500
750
1000
1500
2000
3000
4000
6000
8000
9000
10000
11200
12500
14000
16000
18000
20000

Adding @karthik and @Jungmi for awareness.

2

Hi @xj2193, It seems you’ve nailed the problem. So please tell us what concepts are missing and what the design should be.
It could be a submission to LOINC/SNOMED, but the release cycle could be substantially longer than a community contribution to OHDSI Vocabularies.

Is it a matter of dozens of concepts or more a new model as smoking?

3

I think there are two questions: 1) should we use LOINC or SNOMED? and 2) should the current overlapping concepts b/w LOINC and SNOMED be in the same domain? currently one is in measurement and the other is in observation based on @xj2193 's first bullet.

I agree this might be a longer fix upstream for the missing codes, but I think there are some short term fixes of aligning the domain and/or making one standard over the other. (Latter might require more discussion.)

Tagging @aostropolets and @MPhilofsky for their input…