OHDSI Home | Forums | Wiki | Github

Titer value mapping in measurement table(Antinuclear antibodies)

#1

Hello everyone,

I am currently working on converting hospital EMR data into a Common Data Model (CDM). During this process, I’ve encountered some challenges with handling test results that are expressed as titer values, such as those for Antinuclear antibodies (ANA), where results may be reported in titers like 1:40 or 1:160. I would like to preserve these exact values in the CDM, but I am uncertain about the best approach.

Would it be possible to represent such values in a format like 1/40 or 1/160 within the OMOP-CDM? Is there any way to store these ratios in a meaningful way without losing the specificity of the data? Alternatively, are we limited to categorizing these results simply as negative or positive?

This issue is not just specific to Antinuclear antibodies. Similar questions arise with other lab results such as virus PCR tests, where titer-based results are common. I am wondering what the best practices are for loading such data into the CDM while maintaining their clinical significance.

Does the OMOP-CDM support storing fractional values for titers, or is there another recommended method to handle this?

#2

Hello @Jinchoi

We do not have any conventions on how or where to store these data in the OMOP CDM. This is something we need to solve in the Themis WG. Do you want to bring your use case to the Themis WG on October 17th at 10:30am Eastern Time?

#3

Good idea to come to @MPhilofsky group indeed.

What I’ve seen is storing ANAs as 1:x as value_source_value. There are no good concept_ids for it and you can’t really store it as number. Funny enough, I just got a question on how to then retrieve and use those values and that was rather straightforward from value_source_value (albeit some sql is needed).

#4

Why not? It’s a float.

#5

I apologize, but due to my upcoming PhD defense, I am unable to allocate time to participate in the Themis WG. If the WG could provide a consensus of the relevant content, I will proceed with the OMOP-CDM conversion accordingly.

#6

Technically you could. Practically ppl are looking for 1:80 or 1:160 not funky floats

#7

True, it’s ugly. But the data is not for the consumption by people, such as individual patient care by a provider. It is for large scale analytics, by a computer. The computer does not care whether you give it 1:80 or 0.0125. We would have to document that, though.

#8

The University of Colorado and other places we are doing the same as @aostropolets

Most ETLs map the non-float data to this field as a default. BUT, as a “source_value” field no standardized analytics will be done. And this field can contain PHI, so we can’t use it in the network. BUT, this is where the data go regardless if they map or not.

True. This is a possible solution.

However, we need the use cases from those who will be using these data for research in order to model these data properly. What question do the data need to answer? How will they be used in cohort creations? And how important is this question? Themis needs a sponsor to move this issue through to ratification. And right now we don’t have a use case or a sponsor.

#9

Do you need guidance on how to put these data in to the CDM? Or do you have a study question relying on these data? If you have a question, what are you asking of the data?

#10

The question I saw is identifying patients with ANA>1:160, which is the diagnostic criteria for SLE. In practice, that means discarding 1:40, 1:80 and whatever else is there and taking 1:160 and 1:320 (the set of possible titers is a short list). That would be an inclusion criteria in the cohort with the dx being an entry (or some other permutations). How important: is important in general but there is no ongoing OHDSI study network that requests it so to speak.

1 Like
#11

Could be the solution. Then we will need the conversion table so that people know how to easily get from floats to what they’re used to (even if this conversion is trivial :slight_smile:).

t