Friday, November 20, 2009

Health IT Standards - what I would do

I almost never blog about anything that's work related. For example, if you visit my blog page you'll see a "label cloud" with 360 posts on Economics, but I'm no economist.

This post, written as a private citizen, is different. I'm going to write about something that I really do know quite well. It's a sufficiently obscure topic that there are probably only a handful of people who know it as well as I, and I doubt any of them have been invited to participate in the Health and Human Services IT standards process.

I wasn't invited, but I feel a moral obligation to contribute anyway. I can't see a good way to do that, so I'll post my contribution here. Sometimes these posts travel in odd ways.

My unusual expertise is in combining the realms of healthcare "accounting" (ICD-9-CM, HCPCS, CPT) and the realms of industrial ontology (gritty knowledge representation) such as SNOMED. I've been personally grinding these pieces together for over twelve years in various software systems. I know them rather better than I'd like.

The accounting systems matter. Their idiosyncrasies distort health care statistics, change people's insurance, impede and break computerized decision support, dictate care and determine how most clinicians define disorders. They are fashioned in obscure dark rooms, and they alter health care as surely as technical accounting dictates corporate software development.

They matter so much that they are deeply embedded and almost impossible to displace. ICD-9 was obsolete 30 years ago, but it staggers on. ICD-10-CM is a merely improvement that will cost many fortunes to implement.

On the other hand, SNOMED, a language for healthcare, is a very rich tool. Buggy, yes. Imperfect, yes. Even so, it's a powerful tool for anyone who wants to provide cost-effective decision support that will make all health care providers smarter and faster.

So why don't we implement things like SNOMED now? Are there technical issues? Well, there are some technical challenges, but they're not too big. The real problem is the deadweight of ICD-9, CPT and all that layers upon them, such as vast "medical necessity" (LRMP, medical coverage) databases. Since payment is closely bound to ICD and CPT coding, the easiest route to legal maximization of reimbursement is to stay close to ICD and CPT.

I don't think we have the energy to move America quickly to better health care standards like SNOMED CT. Maybe we do, but this kind of change is very hard. Even so, I think we can do it gradually. The trick is to keep the current system in place, while incrementally building up an alternative approach.

For example, consider the "coverage determination" database. This is a reasonably complex set of tables that define relationships between ICD-9-CM (aka "ICD" in the US) codes and CPT codes (AMA owns CPT btw). The tables express rules such as "we will pay for procedure X (CPT) if a patient has condition Y" (ICD).

I think those tables would be simper, and more internally consistent, if the rules were expressed using SNOMED CT. Medicare (CMS) could then publish rules in both SNOMED and, through things called "mappings", ICD-9-CM and CPT too. The transaction systems would still use the ICD and CPT codes of old, but developers could represent the rules internally using SNOMED, thereby facilitating SNOMED use in their clinical systems. This alone would remove a very large hurdle.

State governments could encourage clinicians to include SNOMED CONCEPTIDs (codes) in a new class of public health and/or payor transactions. This would be entirely optional, but transactions could have come with small payments and regulatory rewards.

We could express new ARRA reporting requirements in SNOMED as well as in the traditional ICD and CPT code sets. Again, accept either data set.

Lastly, we could accelerate implementation of SNOMED-founded ICD-11, perhaps even foregoing ICD-10-CM plans and doing an early partial implementation of the full ICD-11 vision.

It's very hard to move things as deeply embedded as ICD-9-CM and CPT. This deadweight is heavy weight. We can't do it all at once, but we could take doable steps that would provide us with better decision support and more portable electronic health records.

We now return you to the regular amateur hour ...
My Google Reader Shared items (feed)

No comments: