OHDSI Home | Forums | Wiki | Github

ICD10CM start and end dates for retired and new concepts

I want to start discussion about tracking ICD10CM codes (new and retired).

See the code below using Vocab version from ‘2021-04-21’ (well 21-APR-2021 :slight_smile: )

d=concept %>% filter(vocabulary_id=='ICD10CM')

> d %>% count(invalid_reason) 
  invalid_reason     n
1                95358
2              D  1566
> d %>% count(valid_start_date) %>% arrange(1)
   valid_start_date     n
1          19700101 63966
2          20061230  1403
3          20070101  6474
4          20091230     6
5          20100101    26
6          20101230     2
7          20110101     6
8          20120101  9077
9          20140101 11934
10         20140701    44
11         20170101  2305
12         20180101   418
13         20190101   317
14         20200101   328
15         20200401     2
16         20210101   616
> d %>% count(valid_end_date)%>% arrange(1)
   valid_end_date     n
1        20081231     4
2        20091231   902
3        20101231   157
4        20111231   103
5        20121231    21
6        20131231     9
7        20161231   212
8        20171231   122
9        20181231     3
10       20191231     7
11       20201231    26
12       20991231 95358

ICD10CM went into effect Oct 1 2015.

I have following topics to discuss

  • There seem to be codes ending Dec 31 (and starting Jan 1) but some websites seem to imply that new codes end Sep 31 and start Oc 1. Who is right?
    CMS website says: These 2020 ICD-10-CM codes are to be used for discharges occurring from October 1, 2019 through September 30, 2020 and for patient encounters occurring from October 1, 2019 through September 30, 2020.
  • Valid start and end dates - should they reflect “tweaking the terminology” or “validity for coding”? (there is new and retired prior the big bang date of oct 1 2015. Which view is Athena adopting (billing validity or creation of codes as they evolved?)

Hi @Vojtech_Huser, thanks for bringing this up. It is associated with the general versioning discussion we have (and have had) here.
So, looking at the CDM definition, we find the following statements:
valid_start_date: The date when the Concept was first recorded. The default value is 1-Jan-1970, meaning, the Concept has no (known) date of inception.
valid_end_date: The date when the Concept became invalid because it was deleted or superseded (updated) by a new concept. The default value is 31-Dec-2099, meaning, the Concept is valid until it becomes deprecated.

As of ICD10-CM: yes, the official start date of a yearly edition is actually in October of the previous year and that version ends with the month of September of the year the version is supposed to be valid for. I can only assume that this has been creating confusion in the past (I learned it only with the recent update) and that this made us choose the actual start and end of year as validity dates. we should probably fix this to better and more accurately reflect the validity of each respective code change.
For the validity question: it is my view that start and end date should reflect validity of coding so that in the ETL process this additional information could be used for QA reasons. Ideally your primary source / system for capturing these billing codes should have taken care of that, but I would feel better if we had that reflected for versioning purposes as well (e.g. if a code remains the same but the meaning / description is changed). Now the large number of codes with a start date in 1970 is presumably due to the fact that when adopting this terminology, we did not have a proper start date in the source data (and did not set one manually). The ones prior to 2015 again are most likely taken from preliminary versions before the big bang of officially making ICD10 coding mandatory and / or are taken from parent ICD10-WHO codes and their validity (but this is speculation).
You do see individual codes showing up, not having a “regular” start date. Like it was necessary in the recent crisis, these are additions to the vocabulary for particular reasons that needed to be added outside the normal refresh cycle.
So, going forward I pledge to use the correct start (Oct.) and end date (Sep.) for all new additions and deprecations. The question remains if for the existing entries, a correction would provide us with enough benefit to justify the effort. Please weigh in!
Cheers - Mik

1 Like

I think fixing it for 2015-2021 to clearly show start and stop for each code would be nice. But I am happy to be outvoted. For my informatics article, I was responding to a comment by reviewer and wanted to show the might of Athena - but I was not able to use Athena for that since it was not so mighty after all…