Around the end of February of this year, I started receiving a higher-than-normal rate of requests for my sampling and extrapolation services. Providers wanted me to create a statistically valid random sample of some set of claims and then conduct a four-year retroactive extrapolation with the results.
While I have done this type of work in the past, most of the time I am defending providers against sampling and extrapolation errors committed by government and private payers; this felt like “opposite day.” At the time, I wasn’t really aware of what is now called the Final 60-day Payback Rule, which the Centers for Medicare & Medicaid Services (CMS) released on Feb. 12 before it went into effect on March 13.
In general, the rule indicates that once you discover or have reliable information that an overpayment has occurred, you have 60 days to pay the money back to the government. I thought that was pretty much it, and likely would have continued thinking so, were it not for all of the extrapolation requests. CMS puts it this way: “Under this final rule, overpayments must be reported and returned only if a person identifies the overpayment within six years of the date the overpayment was received. Specifying the length and other parameters of the lookback period provides additional clarity for providers and suppliers who have identified an overpayment that is covered by the provisions of 1128J(d).”
This still didn’t explain all the extrapolation requests, so I did a little more digging and started focusing on the lookback period. As cited, the 60-day rule provides for a lookback period of six years, meaning that CMS can review (or require a review) of claims related to those claims subjected to repayment for up to six years back from the initial date of service (or discovery, I think). Prior to this point, the lookback period was four years, and so the panic associated with getting the sampling and extrapolation engagements finished by March 12 clearly was to limit the extrapolation to four years rather than the new six years.
To get a better perspective on this, I spoke with one client who told me that, in their interpretation, CMS was requiring all providers to conduct an internal audit, and when they found claims that were overpaid, to conduct an extrapolation for those claims over the prior six years, to extrapolate the overpayment, and then to pay that extrapolated amount back to the government. I’m not an attorney, but that really didn’t make a lot of sense to me. So I did what every layperson should do when faced with legal uncertainty; I called a lawyer. In fact, I called several, including two of my favorite healthcare lawyers: former Assistant U.S. Attorney (AUSA) Robert Liles of Liles Parker and our own David Glaser of Frederickson and Byron. I asked them the questions that my clients asked me, and what I discovered were more questions than answers. In fact, the whole exercise made me question whether what CMS published was really a clarification, by its definition. Because in the end, I got more “I’m not sures” than the types of answers I was seeking.
In general, my clients wanted to know when to extrapolate; in essence, while the 60-day rule was relatively clear, when did the six-year lookback rule apply, particularly with regard to extrapolation?
To begin, what constitutes “credible information” about an overpayment? Can the Comprehensive Error Rate Testing (CERT) study be considered “credible information?” For 2015, CERT reported that somewhere around a third of all initial hospital visits are overpaid. Does that mean that every practice that bills for initial hospital visits should be subject to a review and six-year extrapolation because they now have credible information of overpayment? How about a Recovery Auditor (RA) audit; if the RA auditor finds a 30-percent overpayment rate, does that constitute “credible information” of an overpayment?
Considering their toxic error rate, I wouldn’t believe anything that a RA audit produces, but that’s just me. And whether findings can be defined as “credible information” can be subject to interpretation. How about a probe audit? When do the results of a probe audit constitute “credible information” for an overpayment? Remember, the issue now is not just whether a repayment needs to be made within the statutory time; it is whether a six-year retroactive extrapolation is in order.
I assume that, in order for an extrapolation to be considered, there needs to be some pattern of overpayment, but just what defines what is or is not a pattern? For example, if you pull five charts for 99223 and find that three of the five are over-coded, does that constitute a pattern? Well, you would think that three out of five is a 60-percent error rate; however, if you are using this to “infer” whether there is a broader pattern of error, then statistically speaking, it is really somewhere between 15 and 95 percent, which is hardly a reliable “pattern” per se. How about five out of five? At first blush, it certainly looks like a 100-percent error rate, but in reality, when applying extrapolation statistics, the lower boundary of a 90-percent confidence interval is 60 percent, meaning that the precision would be worse than that acceptable by the U.S. Department of Health and Human Services (HHS) Office of Inspector General (OIG) under its rules. You would actually need nearly 10 out of 10 being overpaid to even get close to establishing what I would consider a reasonable pattern of overpayment.
In summary, based on my reading and conversations with attorneys, CMS has done nothing more than muddy up the waters on this issue. I get the 60-day part, but the lookback, in my opinion, remains anything but clear. Maybe it’s me, but when I think of something being “clarified,” I think of it being, well, clarified. What else can I say? At least from a statistician’s perspective, my recommendation to providers is to have a serious talk with your healthcare attorney before jumping into the “statistically valid random sample and extrapolation” club. You may find out that, while the 60-day rule applies, the six-year lookback does not, and as with any extrapolation, if you are wrong, then you can expect to be really wrong – and that’s not a favorable position by any stretch of the imagination.
And that’s the world according to Frank.
About the Author
Frank Cohen is the Director of Analytics and Business Intelligence for DoctorsManagement, a Knoxville, TN consulting firm. Mr. Cohen’s specializes in data mining, applied statistics, practice analytics, decision support and process improvement.
Contact the Author
Comment on this Article