Avoid ICD-10! Yes you can!

Lots of news/talk about ICD-10 these days.  Most organizations are spending time and money training care providers on it.  Software developers are busy implementing it – often by changing diagnosis selection search menus from ICD-9 to ICD-10.

They're missing a fantastic opportunity.

ICD-9-CM and ICD-10-CM are administrative coding systems.  They're used to code diagnoses. Clinicians have (unfortunately) been forced to learn many ICD-9 codes and are being told that we need to shift to ICD-10.  Some of our colleagues are hoping that they can just use ICD-9 and "someone else" will convert ICD-9 to ICD-10 but of course this can't happen.  ICD-10 is much more granular, and often requires additional information.  It's like the vet requiring one to specify your animal's breed:  ICD-9 allowed for "dog, cat, aardvark."  ICD-10 requires:  "Golden Retriever, Persian, O. a. lademanni ."  Nobody can translate to the more precise term if you hadn't recorded sufficient information in the first place.

"But how can we avoid ICD-10?  That's the title of your blog post!"  You say.  "How?  Why?"  ICD-10 (and ICD-9) are administrative coding systems, weren't designed by or for clinicians.  We don't think that way.  There are (much) better alternatives.  When ONC made SNOMED-CT required for recording diagnoses in certified EHRs in 2012 (effective for the 2014 certification criteria) I thought it would be obvious that the combination of SNOMED-CT for recording of diagnosis – combined with the free ICD-10 to SNOMED CT mapping tools that NLM published at the same time would meet the needs of organizations to RECORD SNOMED-CT and yet DELIVER ICD-10 to those who required it – primarily CMS and other payers.  Why capture SNOMED-CT and then (again) capture the same information in ICD-10?  I was sure that everyone would "get" the hint.  Commercial solutions like IMO and HLI offer even more elegant methods of capturing interface terms (terms that are customized to the user) and then mapping to the proper code:  SNOMED-CT for clinical data recording and transmission, and ICD-for administrative transactions.

It wasn't obvious.  Many (but not all) health IT developers ignored the opportunity to insulate clinicians once and for all from administrative codes.  Hospitals and other care delivery organizations spent millions on consultants to develop and implement training and "go-live" strategies to teach clinicians ICD-10.  I implored folks in both communities to think past the veneer of the federal regulations, read the preamble of the ONC Certification criteria (where we explained much of this) and think outside of the box.  Innovation?  Nope.  Folks have read only the veneer of federal regulations from both CMS and ONC, avoided creative thinking, and implemented solutions that check the regulatory box, blame the feds for it, and impose massive pain on a generation of clinicians.

It could have been avoided.  

Naysayers will insist .. "but what about the extra information that ICD-10 requires such as laterality?"  And my answer is that this information can and should be captured without ever exposing a clinician to an ICD-10 code.  Some organizations are already doing this.  Some EHR developers are already doing this.  If yours isn't, then you should ask them why not.  

The requirement is that ICD-10 be delivered.  There is no requirement that ICD-10 be entered into the computer (or paper) by the clinician.  When I order a diagnostic test such as imaging or blood work, those doing the testing will likely require ICD-10 so that they can pass it along to those who will pay them for the service (I say "may" because again – the requirements of them are to pass along ICD-10 to those who will pay.  But they have passed on this burden to the clinician without careful thought:  they, too could insulate the clinician from the burden and perform the translation from a clinical question ("why is this test being ordered?") to a billing transaction ("what is the ICD-10 code for which this test was ordered?")  Technology should capture the diagnosis in a terminology that I understand – MY language (HLI, IMO or SNOMED-CT) and if additional data is required – I should always be prompted for it – in the most elegant manner possible.  The information that I capture can/should then be stored in the patient's problem list if it's not already there (and of course if it IS already there – it should be offered as an initial selection to avoid replicating work that was already done!) and then translated in the background into the administrative code.  This should be opaque to the user.  Accessible?  Yes – sure.  Just as I can "view source" in my browser to see the HTML.  But really – who wants to do that?  Not me (most of the time).  Not you.  Nor will I need to see the ICD-10 code 99% of the time.

Don't burden your clinicians with ICD-10!  Avoid it.  Yes you can.  And you should.  Anything less is irresponsible.  Yes – some Who have been "educated" by high-priced consultants will ask for it.  But you shouldn't give them a faster horse.  Give them what they need.