Reviewing “Reviewing Top 10 DRGs…”

April 22nd, 2013 / By Ron Mills, PhD

For the Record magazine recently posted an article, “Reviewing Top 10 DRGs, Codes, Insufficient for ICD-10 Prep,” by Valerie Rinkle. AHIMA linked to it from one of their recent emails, so many of you may have already seen it.

The article accurately reflects the current ICD-10 zeitgeist—everyone is concerned about the financial impact of the coming transition to ICD-10—and since 3M also sells the kind of comprehensive analysis it recommends, I ought to be like, “Right on! Go for it!” But I have some reservations that get in the way, so I figure I can share them with you and you can draw your own conclusions.

I quite agree that it is insufficient to “pull the top 10 … DRGs or ICD-9 codes for revenue and volume” and look there for potential trouble. Assuming payers intend, like CMS does, to play by ICD-9 rules until there is enough real ICD-10 data to start making use of its better discrimination, then reimbursement shifts only occur in places where ICD-9 makes distinctions that ICD-10 doesn’t. There are only about thirty of those with significant volume—the top ten of which I discussed here a few months ago. So, while pulling the top 10 DRGs might not help, pulling the top 10 DRG shifts may be enough to see where you stand.

Though if you can afford it, by all means plunk for the full analytics and modeling treatment. After all, the payers with whom you negotiate may have done so—though judging from the methodologies I’ve seen in their RFPs, I’d bet that what they think they know is just as likely to hurt them as you. Besides, (the elephant in the room no one talks about) every model’s conclusions, no matter how good the model, will be swamped by coding uncertainty. How accurate will your coders be?

I can’t really argue with the article’s assertion that “Numerous payers have acknowledged that they intend to process ICD-10 claims by applying the reimbursement maps and then grouping the codes from the maps,” assuming that “numerous” means “maybe more than one.” CMS, 3M, and AHIMA, among others, have been recommending for years that payers convert their systems to process ICD-10 directly. Nearly all of the ones I’ve talked to did so—and that was before they were given an extra year to get ready.

Payment systems based on current version MS-DRGs or APR-DRGs would not need to map—version 32 groupers will process ICD-10 directly. For payment systems based on earlier versions of DRGs that will not be converted to ICD-10, 3M has for years included a “Codemapper” component in its groupers, so that users can code in the current version of the coding system, but still use an older grouper. Come October 2014, Codemapper will be able to convert version 32 ICD-10 to version 31 ICD-9 and then apply its usual 9-to-9 rules to get back to any old grouper still in use. Some grouper vendors license this utility from 3M, others create their own. So in this case there is a discoverable map involved—you merely have to ask your payer to elicit it from the software, or buy the software and discover it yourself. Good luck (and by the way, the financial shifts will still be confined to those areas where ICD-9 makes distinctions that ICD-10 no longer does).

For those of you still intent on full frontal financial forecasting, I’ll finish by focusing on one sentence that I completely agree with—as long as I get to define its terms my way:

“To better isolate the areas of greatest risk requires a tool that can apply the general equivalence mapping (GEM) translation maps taking the ICD-9 codes to their ICD-10 code options and the reimbursement maps that show how payers will likely process the ICD-10 codes.”

With respect to “taking ICD-9 codes to their ICD-10 code options,” translation must be of records, not just individual codes. A correctly coded ICD-9 record tells a story and it is that story that a payment system (especially DRGs) evaluates. Effective translation must produce an ICD-10 record that tells the same story. Analysis of single codes out of context produces nonsense. Further, as I have argued elsewhere, it takes a lot of programming beyond the GEMs to turn a correctly coded ICD-9 inpatient record into a plausible correctly coded ICD-10 equivalent. The best way to do it is to have your coders dual code a few thousand discharges. (This has the side benefit that your analysis now includes an estimate of coder variability.) But if dual coding is too expensive and you have to use software, have your coders scrutinize its performance carefully—the accuracy of the model’s predictions depends critically on how well it preserves the story in the codes.

Finally, I take “reimbursement maps” to mean “pricing logic.” You have to know how a payer will process ICD-10 coded records. That logic may be expressed as a ICD-10 to ICD-9 map, though more likely it will be expressed directly in ICD-10 as a list or a definitions manual or only in software, such as the ICD-10 APR-DRG grouper. In any case, your impact analysis will need to know the pricing logic. To get it, you should be pestering your payers, who in turn should be pestering their vendors. 3M realized this years ago, and so has ICD-10 versions of all its relevant products in the pipeline.

The ICD-10 Financial Impact Analysis market is open for business. There are “solutions” available for every level of budget and concern. Please be a careful shopper.

Ron Mills is a Software Architect for the Clinical & Economic Research department of 3M Health Information Systems.