Thursday, October 20, 2011

Computer-Assisted Coding for ICD-10

Note: I posted this on LinkedIn in response to a question from a fellow group member.  After I realized I spent more than a few mintues responding, I fugured I may as well post this here since I rarely post topics like Computer Assisted Coding that are largely in the realm of healthcare providers - as opposed to my payer bailiwick.

Here's what I posted on LinkedIn - Unadulterated:

I’m not real familiar about the functional details and better or less-better vendors in emerging computer-assisted-coding applications (CAC); mainly because I’m a payer-side, administration, financial and care management guy. But from what I’ve come to know about this topic via my natural curiosity over the last year, I'll offer this response for your consideration:

{For sure, I’m venturing just a bit outside of my bailiwick here so consider the following information subject to correction and due the payment I am requesting; nothing! :)}


Besides the anatomical and physiological capabilities that must be identified and processed by a CAC (or even a human for that matter,) I’ve learned the following “data points” are required for accurate ICD-10 diagnosis coding:

1. Type of encounter (initial or subsequent or sequelae?)

2. Applied specificity (did the patient lose consciousness?)

3. Acute versus chronic

4. Relief or non-relief (intractable versus non-intractable?)

5. External cause (was it caused by an accident? – i.e. Other Party Liability?)

6. Activity (what was the patient doing when ‘injured?’)

7. Location (where was the patient when injured?)

Many of these data points should be currently available in a provider’s EHR, medical records and notes. So a good NLP-based CAC application should be able to consistently, accurately identify and categorize all of the above data points into their clinical documentation process over the next 24 months – well in advance of the October 1, 2013 ICD-10 compliance date.

Integration and access to multiple data stores - structured and unstructured will be key features of any CAC application! But that's what NLP is all about, right?

Some vendors I’ve noticed in this emerging CAC space include - in no specific order or testament as to capability or value – include Quadramed, OptumInsight, 3M, Nuance, Ingenix, Wolters Kluwer, Precyse, and Dolbey. But I don’t know much about their detailed CAC capabilities. Again, I'm not focused on this topic - yet interested for sure!

And, to be clear, I don’t have any financial or emotional connection to any of the above vendors. Well, maybe Quadramed. :') I’ve just heard and/or read about all these vendors CAC efforts more than a few times lately; and all of them seem to be moving in the right/straight direction. (As opposed to a number of firms I’ve run across who seem totally lost and out of the ballpark.)

Here are some potential useful references that may be of use to you with your quest for CAC knowledge:

Top 10 Questions for CAC Vendors:
http://journal.ahima.org/2011/07/01/ten-questions-for-cac-vendors/

Evaluating Computer Assisted Coding Systems & ICD-10 Readiness: http://www.mediregs.com/files/1007-1/WKLBEvaluatingCADICD10.pdf
{I’m not endorsing WK, I just think this is a pretty good, clear overview of CAC. I’d include the same info from other vendors if I could find such from their sites}

Wow! Isn't speech to text translation cool? I just rattled all this off in about 10-15 minutes (with, admittedly the majority of that time editing mistakes and inserting white space).

Take care Corrie,et al

Steve/ShimCode

My Thoughts on CAC and new ICD-10 data points:

No comments:

Post a Comment