OHDSI Home | Forums | Wiki | Github

COVID lab result mapping

(Hongjue Wang) #1


For the All of Us Research Program, we are looking to help sites map their COVID-19 related test results (numeric values) to categorical values. For example, for the below tests, we are considering mapping all values lower than 1 to 45884092/Nonreactive and values above or equal to 1 to 45881802/Reactive. Would this be appropriate? Any input/comment would be much appreciated.

94505-5 SARS-CoV-2 (COVID19) IgG Ab [Units/volume] in Serum or Plasma by Immunoassay
94506-3 SARS-CoV-2 (COVID19) IgM Ab [Units/volume] in Serum or Plasma by Immunoassay
94769-7 SARS-CoV-2 (COVID19) Ab [Units/volume] in Serum or Plasma by Immunoassay


(Alexander Davydov) #2

Is it a titer? Or some arbitrary index?

For categorical results, it’s better to use [Presence] property, e.g. SARS-CoV-2 (COVID-19) IgG Ab [Presence] in Serum or Plasma by Immunoassay.

Positive / Negative tend to be used much more common for both Immunological and Genetic tests.

As well as they specified as predefined answers by LOINC:

(Melanie Philofsky) #3


No, I do not believe this is appropriate. The tests you mention:

are antibody tests. Check with a lab expert at your institution, but these tests could be testing the amount of SARS-CoV-2 antibody in a Person’s blood. If this is true, then a numeric result is appropriate. Take the numeric values from above and put them in the Measurement.value_as_number field.

My basic rule of thumb is that if something is affecting >1% of your data, you should investigate. If the above constitutes 20 records out of half a million, don’t worry about it. Also, what’s the use case for COVID-19 antibody tests?

(Alexander Davydov) #4

Not in this case since the arbitrary units are not useful. If you’re able to convert to units/volume, do it. If not, the only choice is to interpret them as mentioned above.

(Melanie Philofsky) #5

Good point, @Alexdavv!

What is the unit_concept_id and unit_source_value for the labs above? We have some COVID-19 antibody Measurements with unit_concept_id = 8763, units per milliliter

(Alexander Davydov) #6

Depending on the assay used, it may be any of index, index of reactivity, score, signal to cutoff ratio, relative light units, fluorescence intensity unit. They all are not-convertible to units or grams since the test is qualitative by design.

These tests are qualitative, BUT at this point, the units are arbitrary rather than international ones meaning that you still cannot compare/convert the results obtained from different assays.

(Hongjue Wang) #7

Thank you both! @MPhilofsky @Alexdavv
These are great points and I agree with your suggestions. The units in this case seem to be index and we think it would be most appropriate to ask the sites to make the mapping decision according to their own clinical interpretation.