OHDSI Home | Forums | Wiki | Github

Mapping drug identifiers (NDC or RxNorm) to on-label drug indications using ICD10 codes

Does a mapping between drug identifiers like NDC or RxNorm to valid drug indications, per the label, using ICD10 codes or other disease/diagnosis descriptors exist. I know that there are some commercial data vendors that have attempted this but wasn’t sure if there was anything in the public domain?

1 Like

@Ilya_Katz:

There is a link from RxNorm to FDB indications, and from there to SNOMED. And there was a similar link from RxNorm to NDF-RT indication, and from there to SNOMED. But NDF-RT is no longer available, and the replacement MED-RT we have not put in, and we may never do that.

This idea sounds nice, but it actually doesn’t work very well. These indications are more like hints. You can display them on an e-ordering system when the physician is browsing products. But for our purposes the idea that you somehow can get all the ICD-10 codes that represent the indication and then find the patient with the indication - pretty much a fallacy. First, most indications are wishy-washy, and not strict ICD-10 or SNOMED material. Image things like “general anxiety”, “difficulty sleeping through”, “fever”, “papulo-vesicular skin lesions”. Second, the links would have to be very broad (and then collecting too many unrelated Conditions, or narrow (missing a lot). In other words the sensitivity/specificity curve is not very favorable.

I would stay away from that. What’s your use case?

Hi @Christian_Reich

For me, I am attempting to figure out the drugs that are used in a given therapeutic area by first using the ICD-code of interest (Let’s say C61 prostate cancer), and obtaining a list of NDC codes that I can use to run analytics on the prostate cancer drug usage. Our current way relies on manual bibliography which is very time-consuming.

@vincenzodelavega:

Yes, would be nice. But it doesn’t work that way. Diseases don’t have a direct street to treatments. It’s a much more complicated equation, and it takes 4 years of Med School to get a basic understanding of it. FDB does a good job of covering some of the obvious cases, but the purpose of FDB is to make suggestions to physicians in the EHR for convenience. You certainly cannot just blindly rely on it.

For drugs, there are drug classes, some of which have a therapeutic intention, like ATC 3rd level. They are actually defined with this in mind, but again, there is no blind one-to-one.

Also, usually associations between conditions and treatments (indications or outcomes) are the result of OHDSI studies. In other words, you should do this data driven.

t