OHDSI Home | Forums | Wiki | Github

NDC codes unrecognizable from claims file

(Chakravarty) #1

I am trying to recognize some of the NDC codes that come through one of vendors. They seem to have 8,9,10,11 codes - even though they should follow the HIPAA requirement of 11-digit character codes. My problem is that the RX claims files that come from RXBenefit has NDC codes that are not being recognized by using SQL search on the NDC database from FDA. And the reason is because of inconsistency in length of NDC that is making it hard for me to figure out where I can place the 0s.
I only get the NDC number and not any other identifier - unsure how to automate this process

(Anna Ostropolets) #2

Here you can find a nice summary of the rules that you can use if you have hyphens. Otherwise, it’s hard, as you can get totally different drugs when placing 0s randomly. Manual curation probably makes some sense if the counts are large.

(Chakravarty) #3

Anna, thanks for your response. My problem is that the codes in the claims files show up without the hyphens; hence, i am unsure of where the 0s are a part of the code and where they have been included to bring it up to a certain format. I feel like if we had a stand-alone database of the “label” part of the NDC code, it might make sense for me to first query against that, to figure out if i am working with a 3,4 or 5-character label length

(Christian Reich) #4


Sounds like a good idea to me!

But careful: The NDC world is messy. The “label” (product, really) part is ambigous. Here is what you will run into: You could extract all the product part of the codes, which are digit 6-9 of a normalized NDC we have in our CONCEPT table. All NDC with the same product code and same digits 1-5 (manufacturer) should map to the same RxNorm concept. Or at least to a similar one (equivalent Branded and Clinical Drugs). But they don’t always. Since NDC code generation is not really overseen, folks issue all sorts of junk and bleed into the last 2 digits (packaging) to identify products.

But give it a try and let us know.

(Feyi110690?) #5

Hi Pavitra, I can help you figure it out. I am a pharmacist and work with NDC number. DM me if you need me,


If dealing with historical data, keep in mind that NDC numbers can be reused. A NDC number that had been used for a discontinued product can be reused for a different product, so a NDC must be combined with a date to identify a medication with certainty.

(Shinjinaka) #7

Do you have a used case? I have seem custom NDC from hospital looks like “00000000001” used for different products, but cannot find actual NDC that were manufacture from a company that used for different product.


I’ve been looking for examples and all the ones I’m finding at the moment are cases where the product description changed without necessarily a clinical difference. E.g. 63807-100-33 (63807010033) was apparently changed from a generic to a branded drug, or NDC 63807-100-11 ( 63807010011 ) got a more specific GPI assigned in March 2015 so the description changed. But there might be cases where the change was more substantial.

I have also seen custom NDCs but that is an unrelated issue.

(Kristin Kostka, MPH) #9

Check out this post:

This is a great example of what we like to call “Bad NDCs”.

(Shinjinaka) #10

I see. This manufacture (51655) is not following the 5-4-2 NDC rule. FYI: (74312) manufacture is also not following the rule.
Thanks for sharing!