This article concerns the use of self-audits. –Ed.
Written by Marla Durben Hirsch
If practices uncover large or systemic billing problems that led them to receive overpayments, they should conduct statistical sampling to quantify the total potential amount of overpayments received.
Doing so will reduce the time and resources some practices will need to spend dealing with overpayments in the wake of the final rule HHS released Feb. 11 (MPCA 3/16). Practices that conduct statistical sampling won’t have to review every claim that may have been overpaid, says attorney Ross Burris with Polsinelli in Atlanta.
Statistical sampling — also known as extrapolation — can be particularly helpful when complying with the rule because failing to meet the deadline to return overpayments is a violation of the False Claims Act. Penalties currently include fines of $5,500-$11,000 per claim, treble damages (damages times three) and/or exclusion from the Medicare and Medicaid programs.
The Accountable Care Act requires overpayments to be returned within 60 days of being identified. The rule clarifies how providers go about doing so.
Now that HHS issued the rule, providers are likely to see an increase in enforcement of the requirement that overpayments must be returned, Burris says.
Industry experts didn’t want to speculate on what percentage of providers might be receiving overpayments and might fall into the category of having large or systemic billing problems. However, all types of providers have at least some overpayments — many of which occur without bad intent.
Statistical sampling is a methodology to apply the rate of billing errors found in a sample of claims to a similar total population of claims. It is frequently used by the HHS Office of Inspector General (OIG), the Department of Justice, zone program integrity contractors (ZPICs), recovery audit contractors and other auditors to calculate providers’ overpayments.
The final rule specifically allows providers to use extrapolation proactively to quantify the amount to be returned, using “sound and accepted principles” that include randomly selecting claims from the applicable population of similar claims and extrapolating only within the time frame covered by the population from which the sample was drawn.
Statistical sampling is very accurate if performed correctly, experts contend.
How to apply statistical sampling
To determine whether your practice should use statistical sampling, first conduct a probe sample or audit. Do so after you receive credible information that there may have been an overpayment.
You’ll probably need to tailor the audit, say to particular years or a particular identified coding issue, Burris says.
If the probe audit reveals a larger, more systemic problem, extrapolation may be appropriate. Several decisions need to be made, in large part depending on the circumstances.
Most overpayments merely involve billing errors — say by a new coding employee — and will be returned to your Medicare administrative contactor (MAC). In these cases, you typically can (but don’t have to) conduct your own extrapolation following CMS’ program integrity manual section 8.4, which outlines what CMS is looking for in extrapolation and using a statistical sampling tool.
One of the most popular statistical sampling tools for this purpose is called RAT-STATS, which was created by the OIG and is available free on its website along with an instruction guide.
“It’s pretty easy to use and you don’t need to be a statistician,” says Gary Keilty with FTI consulting in Washington, D.C. RAT-STATS even allows users to utilize different “confidence” levels (the degree of certainty that the sample correctly depicts the total population) and “precision” levels (the estimated range of accuracy) percentages, he adds.
CMS accepts RAT-STATS as a statistical sampling tool for extrapolation calculation purposes, although its use is not required and a MAC may not have a preferred method.
Once your calculation is complete, go to the refund form from the MAC and submit the overpayment, Burris says.
“Most of the time the MAC will simply say ‘Thank you,’ and that’s the end of it,” he says.
If you’ve uncovered more than a simple error, the fraud and abuse or Stark laws(*) may be implicated. If the government is already involved, you’ll need more manpower and will want to engage an experienced attorney to handle the issue, including hiring an expert statistician, determining which government entity is involved to return the overpayment to (and what else that may involve), what the statistical sampling will entail and how best to approach the government.
Often the provider representative and the government will agree to negotiate smaller claims populations or timeframes subject to extrapolation, look at the sample together or use a statistician to provide a report, and then the parties come to a settlement based on those discussions, Burris says.
Using an attorney also will keep your investigation protected under attorney-client privilege, Keilty says.
It’s important to identify which government agency is involved, since different agencies prefer different methods of statistical sampling. For example, the OIG, which handles repayments under its self-disclosure protocol for issues involving the antikickback statute, prefers extrapolation using a 90% confidence level and a 25% precision level, Keilty says.
Consider 6 extrapolation tips
1. Don’t automatically assume that a discovered error requires extrapolation. A small error involving one or a few claims does not rise to a systemic issue that lends itself to extrapolation, Burris says.
2. If you’re handling your own extrapolation, make sure it’s a method the government accepts. This includes utilizing RAT-STATS as a statistical sampling tool and utilizing preferred confidence and precision levels. If you hire an outside statistician, ask what method he/she uses and whether the government finds that method acceptable. The government may be more likely to challenge your sampling — and overpayment amount — if it’s less accepting of the method being used.
3. Make sure you choose the right type of sample and sample size. For example, the sample needs to be randomly generated. RAT-STATS and even Microsoft Excel contain random generators, Keilty says.
4. Be careful about what you’re including in the extrapolation. For example, extrapolation assumes the total population it’s being applied to is very similar. If it isn’t, you may be including too many claims and triggering a higher overpayment than you should be repaying.
5. Note that if you use extrapolation, you don’t need to make two repayments — one for the errors uncovered during the probe audit and another covering the extrapolated amount. You can include them in just one payment, Burris says.
6. Include documentation explaining how the extrapolation was conducted and what method was used. The final 60-day overpayment rule requires this information be included when reporting and returning the overpayment, Burris says.
(*) Stark Law: 42 U.S.C. § 1395nn – Limitation on certain physician referrals.
Original Title: “Save time and money — extrapolate to calculate overpayments”