Saturday, August 02, 2014

Transition from ICD9 to ICD10 Codes: Strategic effect on reimbursements?

This fall, the US government requires pretty much all health care providers to transition from the old set of codes for medical diagnoses, known as ICD9, to a new larger set of codes, ICD10.  According to the Center for Medicare and Medicaid Services, ICD10 allows for 7 digits while ICD9 allowed only for 5.

So under the new system, we could have up to 9,999,999 different codes while under the old there was only the possibility for 99,999 -- and as the ICD10 is alphanumeric, this is an understatement.  I guess however they never maxed out on usage of the total possibilities, as the article referenced below says ICD9 had only 13,000 distinct codes while ICD10 will have only 68,000.  Quite a difference to be sure, and still room to grow.  CMS provides a fact sheet if you want to learn a bit more.

The key thing is that these diagnostic codes are used for billing purposes.  Submit a code to Medicare, and you get the payment associated with that code.

Naturally the pundits are having a grand time with this one.  Here are the "16 most absurd ICD10 codes."  Here are just a couple:

  • W55.41XA: Bitten by pig, initial encounter​.
  • W220.2XD: Walked into lamppost, subsequent encounter. 
  • Y93.D: V91.07XD: Burn due to water-skis on fire, subsequent encounter​.
  • W61.12XA: Struck by macaw, initial encounter. 
These actually seem too bizarre to be true, but who knows.

Anyway, I have a serious question:  How will the move to a more-granular coding system affect billings?

What I have in mind here is thinking about the hospitals as trying to maximize their revenues given the procedures they did to a patient -- strategic coding of procedures to maximize revenue.  (I am actually on some email list that sends me announcements for seminars to teach me how to code "properly," so strategic coding is certainly not a crazy idea.)

I expect that many first answers would be that a more accurate (more granular) coding system would reduce billings.  Why -- I don't really know.  More precision avoids mis-classification.  Sure, but what is the effect on revenue?

My knee-jerk reaction is that billings will likely increase. Why?  Well, suppose there was an old ICD9 code that had some average reimbursement attached to it, say $1000.  Now there are two subcodes for that old diagnosis, and the reimbursement for one code is $500 while the other is $1500.  These payments were set on the thinking that half of the old diagnoses were of one new type while the other half were of the other new type.  Assuming there is some wiggle room in the new codes (coding systems, like contracts generally, cannot delineate every possible outcome) the hospitals will shift whatever procedures they can into the higher rate category.  So while there was an even split of sub-diagnoses in the population, there will be some strategic over-billing that happens.

This is admittedly incomplete, for we have to think of the whole coding system, and we have to specify a bit more about who knows what and what the constraints are.

Another way to think about it is to go in reverse:  Suppose we reduce the number of codes, and take an extreme case of going down to just one code (capitation, sort of!).  Then if there was any strategic overbilling going on previously, going to one code has to reduce the total reimbursements, it would seem.  A single code is actually more accurate, as we get the true average rather than an average biased by strategic coding.

This might be an interesting project to work on.  There was a Rand study in 2004 that looked at the transition to ICD10 but it doesn't really take this perspective.