OHDSI Home | Forums | Wiki | Github

[NDC Questions] 9-Digit vs 10-Digit?

Vocabulary Users
#1

I’m posting this for @anthonysena.

Hi @Christian_Reich ,

I had a few questions related to NDC codes.

We were investigating a drug ingredient Tacrolimus. We are working with a partner who needed to use NDCs, so we used the Vocabulary to find those mapped to Tacrolimus. They’ve since come back with a few questions and identified some gaps where they had NDC codes for Tacrolimus that were not present in our list (and I cannot find them in the Vocabulary).

When we provided them with the list of NDC codes, we gave them both the 9 digit and 11 digit codes. This caused some confusion and I wasn’t sure why we had both in the Vocabulary? Erica mentioned to me that we use the 9-digit code when we cannot find an exact match using the 11 digit; we essentially chop the 11 digit code down to a 9 digit and try to match that way. However, when I did a quick query of NDC codes in the Vocabulary, I found that not all 11 digit codes have a 9 digit equivalent (when I drop the last 2 digits). Should all 11 digit NDCs have a cooresponding 9-digit?

It was noted that we were missing NDCs from our list. I took a look on the OHDSI wiki: http://www.ohdsi.org/web/wiki/doku.php?id=documentation:vocabulary:ndc and when I read this it sounds like we source the NDCs from RxNorm via NLM? If so, it could be due to the fact that this other party was looking up NDC codes on the FDA website: http://www.fda.gov/Drugs/InformationOnDrugs/ucm142438.htm. Here is an example of an NDC code that was noted as missing (perhaps it is in the one coming on 1/17?):

Tacrolimus | 63629-5423-1 | CAPSULE, GELATIN COATED | 1 mg/1 Product NDC: 63629-5423 Proprietary Name: Tacrolimus Non-Proprietary Name: Tacrolimus Product Type Name: HUMAN PRESCRIPTION DRUG Market Category Name : ANDA Application Number: ANDA090802 Route Name: ORAL Substance Name: TACROLIMUS Package Description : 30 CAPSULE, GELATIN COATED in 1 BOTTLE (63629-5423-1) Pharm Class: N/A
DEA: N/A
Labeler Name: Bryant Ranch Prepack
Start date: 10-30-2012 / End date: N/A

#2

@anthonysena; @ericaVoss:

Bring them on. We will add them. There is no complete list in the world. Only the Lord in Heaven knows them. It’s called National Drug Codes, but there is nothing National about it. So, it’s an endless business of improving and catching up. But I think we have by far the best collection: Better than the FDA or any other commercial source.

Yes, welcome to this world. We should probably document this with some better examples. The problem is this: the 10th and 11th digit should only designate the package size. The first 9 digits should unambiguously identify the manufacturer (first 5 digits) and the product (4 digits). So, any two 11-digit NDCs with matching 9 digits should be the same product, and since we don’t care about the package size we would only need the first 9-digit portion. But since there is no authority to enforce that, it gets violated all the time, and we need to keep the 11-digit NDCs and the 9-digit NDCs.

So, here is what you do: Take all 11 digit NDC and chop off the 2 digits. If they all map to the same RxNorm Concept things are good. If not, stick with the 11 digits. If there is no 11-digit, only a 9, take that one. If you have an 11-digit that is not in the vocabulary, but the first 9 digits are (either in other in a 11-digit or a 9-digit NDC), do the same: figure out whether the 11-digit are unambiguous and use them. If not, you are a little out of luck.

And submit all NDCs you can’t find.

From RxNorm (coming from FDB, GPI, Multum, FDA), from RxNav and from the FDA SPL on the website and on Dailymed. It’s a big mill.

[quote=“ericaVoss, post:1, topic:1072”]
Here is an example of an NDC code that was noted as missing:[/quote]

It’s right there: 46252418 “Tacrolimus 1 MG Oral Capsule”. The last “two” digits in your description was shortened to one. The NDC code should be 63629542301. It it is nicely mapped to the right RxNorm. So, all is good. :smile:

Have fun with this stuff.

#3

Thanks Christian – this helps and gives me even more appreciation for the vocabulary.

I think part of the challenge that I’m facing is that I’m investigating this drug with the intention of providing NDCs back to the partner. With that in mind, I took 2 approaches:

  1. I looked at the NDC codes that are mapped to the RxNorm Ingredient for Tacrolimus and then used that list to do some matching between 9 and 11 digit NDCs

  2. I then did a basic string search for “tacrolimus” against the concept_name field to see if there were any other NDC codes that matched the description that were not mapped to RxNorm.

In approach #1, I found that there are 138 distinct 11-digit NDC codes and 60 distinct 9-digit NDC codes that map to the RxNorm ingredient Tacrolimus (concept_id = 950637). Within this list of codes, we attempted to match every 11-digit to a corresponding 9-digit by dropping the last 2 digits as you mentioned. Doing this left us with 23 distinct 11-digit codes that did not have a match to a 9-digit code. Within that list of 23 11-digit codes we had 14 unique 9-digit stem codes. I have attached a list (tacro-missing-9-digit-counterpart.xlsx).

When we take a deeper look at this list, only 5 of the 23 have a valid end date in the vocab and when we looked at our in-house data, we found even fewer in use (2 of the 23). So I’m concluding that it is very unlikely that we’ll encounter these codes. Do you agree?

In approach #2, I used the following query to identify NDC codes with the term “Tacrolimus” and then to see if any concepts were not mapped to a concept in RxNorm:

select c.*
from
(
select *
from concept
where concept_name like ‘%TACROLIMUS%’
and vocabulary_id = ‘NDC’
and invalid_reason is null
) c
left join concept_relationship cr ON cr.concept_id_2 = c.concept_id
left join concept c2 ON cr.concept_id_1 = c2.concept_id
where c2.concept_id is null

This yielded a list of concepts (attached: tacro-missing-mapping.xlsx) that are not mapped to a concept in RxNorm. I’d like to see if these mappings can be added?

In light of this work and the needs to provide these NDCs to a partner, I have some additional follow up thoughts/questions:

  • Should valid 11-digit NDC codes always have a corresponding 9-digit code in the vocabulary?
  • Where we have identified an 11-Digit NDC code in the vocabulary that does not have a corresponding 9-digit code, it is advisable to still use the first 9 digits when matching codes?

Thanks!

tacro-missing-9-digit-counterpart.xlsx (12.4 KB) tacro-missing-mapping.xlsx (12.1 KB)

t