Medicare Extrapolation: How Statistical Sampling Turns a Small Audit Sample Into a Large Overpayment

Imagine a Medicare contractor reviews 40 of your claims and finds that 15 were not supported, for a total of about $6,000 in denied payments. A few weeks later, your Medicare Administrative Contractor (MAC) sends a demand letter for more than $500,000. How did $6,000 become half a million dollars? The answer is extrapolation.

Extrapolation is the process of using a statistically valid random sample to estimate the total overpayment across a larger group of claims, called the universe. Instead of reviewing every claim, the contractor reviews a sample, calculates the overpayment in the sample, and projects that result to all claims in the universe. Because the projection can be much larger than the sample findings, extrapolation is one of the most financially significant tools in Medicare program integrity.

This article explains what Medicare extrapolation is, the legal basis for it, when contractors may use it, how the sampling process works step by step, how the overpayment estimate is calculated, what documentation contractors must keep and provide, and how providers can reduce their risk. A companion article explains how to challenge a statistical sample and extrapolation. This article is educational and does not replace legal or statistical advice.

What Is Extrapolation?

In Medicare program integrity, extrapolation means estimating an overpayment for a full population of claims based on the results of reviewing a statistically valid sample. CMS's Medicare Program Integrity Manual, Chapter 8, Section 8.4, provides detailed instructions for contractors on using statistical sampling for overpayment estimation.

The basic idea is simple:

1. Define the group of claims under review (the universe and sampling frame).

2. Randomly select a sample from that group.

3. Review each sampled claim to determine whether it was overpaid, correctly paid, or underpaid.

4. Use statistical formulas to estimate the total overpayment across the full group.

When the sample is properly selected and reviewed, the estimate is considered a valid basis for an overpayment determination, even though most claims in the universe were never individually reviewed.

Why Medicare Uses Extrapolation

Reviewing every claim for a provider with thousands of claims would be costly and slow for both the government and the provider. Statistical sampling allows contractors to estimate overpayments efficiently. CMS has long treated statistical sampling as a legitimate method for estimating overpayments, and CMS Ruling 86-1 sets out the agency's policy and history on the use of sampling and extrapolation in Medicare overpayment determinations. Medicare Appeals Council decisions have relied on that Ruling and have noted that CMS Rulings are binding on Administrative Law Judges (ALJs) and the Council.

The Legal Basis and Limits on Extrapolation

Section 1893(f)(3) of the Social Security Act, added by the Medicare Prescription Drug, Improvement, and Modernization Act of 2003 (MMA), limits when Medicare contractors may use extrapolation. Under that provision, a Medicare contractor may not use extrapolation to determine overpayment amounts to be recovered by recoupment, offset, or otherwise unless the Secretary determines that:

• There is a sustained or high level of payment error, or

• Documented educational intervention has failed to correct the payment error.

The statute also provides that there is no administrative or judicial review of determinations of sustained or high levels of payment error under this provision. CMS's Program Integrity Manual repeats this point, and federal courts and the Medicare Appeals Council have held that providers generally cannot challenge the sustained or high level of payment error determination itself.

How Contractors Show a Sustained or High Level of Payment Error

CMS's Program Integrity Manual explains that a sustained or high level of payment error may be determined through a variety of means, such as:

• High error rates compared with similar providers.

• The provider's history of similar non-compliance.

• CMS approval in connection with a payment suspension.

• Information from law enforcement investigations.

• Allegations of wrongdoing by current or former employees.

• Audits or evaluations by the HHS Office of Inspector General (OIG).

If none of these circumstances applies, the manual instructs the contractor to consult with CMS before creating a sample or requesting records for extrapolation. In practice, the error rate found in a probe or in the sample itself is often used to support the determination.

Which Contractors Use Extrapolation?

CMS's Program Integrity Manual applies its statistical sampling instructions to several types of contractors, including:

• Unified Program Integrity Contractors (UPICs).

• Recovery Audit Contractors (RACs).

• The Supplemental Medical Review Contractor (SMRC).

• Medicare Administrative Contractors (MACs), including DME MACs.

The manual notes that RACs and the SMRC need approval from their CMS contracting officer's representative at each step. Regardless of which contractor performs the review, only MACs issue demand letters and recoup overpayments.

The HHS Office of Inspector General also uses statistical sampling and estimation in its audits, and OIG makes its RAT-STATS statistical software available to the public. OIG audit recommendations are then acted on by CMS and its contractors.

The Eight Steps of Medicare Statistical Sampling

CMS's Program Integrity Manual lists the major steps contractors follow when conducting statistical sampling for overpayment estimation:

1. Identify the provider or supplier.

2. Identify the period to be reviewed.

3. Define the universe and the sampling unit, and construct the sampling frame.

4. Assess the distribution of the paid amounts in the sample frame to decide on the sample design.

5. Assess the sample size and the corresponding confidence interval.

6. Design the sampling plan and select the sample from the sampling frame.

7. Examine each sampling unit and determine whether there was an overpayment or underpayment.

8. Estimate the overpayment and follow the notification and collection procedures.

The manual requires each step to be documented in enough detail that the sample could be replicated. Below is a plain-English explanation of each key concept.

The Review Period

The review period is the span of dates of service or claim payment dates included in the review. The manual explains that the period is chosen with a statistician, considering factors such as how long the pattern of error is believed to have existed, claim volume, coverage policies, prior prepayment review, the dollar value relative to the cost of sampling, and time limits on reopening claims. For cost-reported claims, the period must fall within one cost-reporting year, and periods longer than one year may require separate samples.

The Universe

The universe is the full group of claims or services that meet the review criteria for the period, such as all paid claims for a specific service by the provider during the review period. For Part B claims, the manual explains that the universe consists of all fully and partially paid claims for the selected period and services.

The Sampling Unit

The sampling unit is the item that is selected and reviewed. CMS's manual explains that a sampling unit can be a claim line, a claim, or a cluster such as a beneficiary or a treatment day. The choice of sampling unit affects how the sample is drawn and how the estimate is calculated.

The Sampling Frame

The sampling frame is the list of all possible sampling units from which the sample is drawn. Ideally, the frame matches the universe. The manual requires the case file to include everything needed to recreate the frame and the sample.

Sample Design

Contractors may use different designs, such as simple random sampling or stratified random sampling. In stratified sampling, the frame is divided into groups, called strata, often based on the paid amount, and a sample is drawn from each stratum. Stratification can improve precision when paid amounts vary widely.

Random Selection

The sample must be selected using a probability method in which every sampling unit has a known, non-zero chance of selection. The manual requires contractors to document the source of random numbers, the software or table used, any starting point, and any seed value so the selection can be replicated. It identifies reputable tools such as SAS, SPSS, and OIG's RAT-STATS.

Sample Size

The manual explains that sample size affects precision but is not the only factor. It states that it is neither possible nor desirable to specify a minimum sample size that applies to all situations, and that a small sample does not by itself make an estimate invalid if probability sampling and estimation procedures were properly followed.

Review of Sample Claims

Each sampled claim is reviewed under the Medicare coverage, coding, and payment rules in effect when the services were furnished. The reviewer determines whether each claim was paid correctly, overpaid, or underpaid, and documents the findings.

How the Overpayment Is Calculated

The Sample Overpayment

First, the contractor adds up the overpayments found in the sampled claims. Claims for which the provider did not submit records are generally counted as overpayments. Underpayments are counted as negative overpayments, which reduce the total.

The Point Estimate

Next, the contractor uses statistical formulas to estimate the total overpayment in the universe. The simplest version, for a simple random sample, multiplies the average overpayment per sampled claim by the number of claims in the frame. This is called the point estimate.

For example, suppose:

• The sampling frame contains 2,000 claims.

• The contractor reviews a random sample of 50 claims.

• The total overpayment in the sample is $5,000, or an average of $100 per sampled claim.

The point estimate would be $100 multiplied by 2,000 claims, or $200,000.

The Lower Limit of the Confidence Interval

Because any sample involves some uncertainty, the point estimate may be higher or lower than the true overpayment. To account for this, CMS's manual generally directs contractors to demand the lower limit of a one-sided 90 percent confidence interval rather than the point estimate. This approach gives the provider the benefit of the statistical uncertainty, because it is designed so that the actual overpayment is very likely to be at least that amount.

In the example above, if the sample results were highly variable, the lower limit might be $150,000 instead of the $200,000 point estimate. If the results were very consistent, the lower limit would be closer to the point estimate.

The manual allows contractors to demand the point estimate in some situations where precision is high and there are sound reasons, and it provides that if the lower bound exceeds the total payment amount in the frame, the demand is reduced accordingly.

Why a Small Sample Produces a Large Demand

The size of the demand depends on three things:

• The number of claims in the universe.

• The average overpayment per sampled claim.

• The variability of the sample results, which affects the confidence interval.

A provider with a large volume of claims and a moderate error rate can face a large extrapolated demand even if the sample overpayment was small. That is why extrapolated overpayments often surprise providers.

What Contractors Must Document and Disclose

CMS's manual requires contractors to maintain complete documentation of the sampling methodology, including:

• The universe and sampling frame, including an electronic copy of the frame.

• The review period.

• The definition of the sampling unit.

• Claim identifiers and dates of service.

• The random numbers used and how they were generated.

• Reviewer worksheets and overpayment or underpayment worksheets.

The documentation must be sufficient to allow replication and validation of the sample by an administrative or judicial body.

The manual also requires that the demand letter and final results explain the sampling methodology, including the universe, frame, sampling unit, selection procedure, strata and sample sizes, period, estimation method, sampling error, and the overpayment amount for each sampled claim. The manual further requires a qualified statistician or person with equivalent expertise to review and approve the methodology, and the statistical expert must review the results before the findings or demand letter is issued.

Notification and the Review Process

CMS's manual describes notification and review procedures for contractors conducting statistical sampling reviews:

• Contractors generally give advance written notice of the review, which includes the reason for the review, the time period, the list of claims, the review location, appeal rights, how sampling will be used to estimate the overpayment, and possible recovery methods.

• Providers generally have 30 calendar days to submit or make records available for this type of review, and claims for which records are not provided may be denied.

• Reviews may occur at the provider's location or at the contractor's office.

• Entrance and exit meetings are encouraged but not required. An exit meeting may include a discussion of preliminary findings.

• The contractor should generally complete the review within 90 calendar days, absent extenuating circumstances.

Providers should always follow the specific deadlines stated in the contractor's request.

What Happens After the Demand Letter

After the review, the MAC issues an overpayment demand letter. Key points include:

• The demand letter explains the overpayment amount, the basis for it, the sampling methodology, appeal rights, and repayment options.

• Interest generally begins to accrue if the overpayment is not repaid within the stated timeframe.

• Providers can appeal through the Medicare fee-for-service appeals process, beginning with a redetermination.

• Under Section 1893(f)(2) of the Social Security Act and 42 CFR 405.379, filing a redetermination within 30 days of the demand letter, and a reconsideration within 60 days of the redetermination decision, can limit recoupment while those appeals are pending.

• Providers that cannot repay immediately may request an extended repayment schedule.

When Appeals Change the Sample Results

If claims in the sample are overturned on appeal, the extrapolated overpayment generally must be recalculated based on the revised sample results. CMS's manual addresses changes to the overpayment estimate resulting from appeals. This is one reason why appealing individual sample claims can significantly reduce the extrapolated amount: each overturned claim reduces the average overpayment that is projected to the universe.

How Extrapolation Connects to Other Enforcement Risks

Extrapolation often occurs in reviews with broader compliance implications:

• UPIC investigations may use extrapolation alongside referrals to law enforcement or payment suspensions.

• OIG audits commonly use statistical sampling and estimation and may recommend that providers identify and return similar overpayments outside the audit period.

• The 60-day overpayment rule at 42 CFR 401.305 may require providers to evaluate whether similar overpayments exist beyond the reviewed period. Providers may use statistical sampling in their own internal investigations to estimate overpayments.

• Repeated errors after education, such as after unsuccessful Targeted Probe and Educate rounds, may support a determination that documented educational intervention failed, which is one of the statutory bases for extrapolation.

How Providers Can Reduce Extrapolation Risk

Monitor Error Rates

Because a sustained or high level of payment error supports extrapolation, keeping error rates low is the best protection. Conduct regular internal audits of high-risk services and correct problems promptly.

Take Education Seriously

If your MAC provides education through TPE or other programs, implement changes and document them. Failure to correct errors after documented education is a statutory basis for extrapolation.

Respond Completely to Record Requests

Claims for which records are not provided are generally counted as overpayments in the sample, which inflates the extrapolated amount. Track every request and submit complete records on time.

Use Data Tools

Review PEPPER reports, Comparative Billing Reports, and your own data to identify areas where your billing differs from peers.

Prepare Early When Sampling Is Announced

When a contractor notifies you of a statistical sampling review, involve experienced counsel and a qualified statistician early. Review the request carefully, preserve records, and organize documentation for every sampled claim.

Keep Strong Documentation

Clear, complete, signed, and timely documentation is the foundation for defending each sampled claim. Every claim overturned on appeal reduces the extrapolated overpayment.

Common Misconceptions About Extrapolation

• "The contractor can only collect what it found in the sample." Not true. If extrapolation is properly authorized and conducted, the contractor can project the sample results to the universe.

• "The sample is too small to be valid." CMS's manual states that a small sample does not by itself make an estimate invalid if probability sampling and estimation were properly done, and the Medicare Appeals Council has rejected challenges based only on sample size.

• "I can challenge the finding of a high error rate." The statute bars administrative and judicial review of the sustained or high level of payment error determination.

• "If the contractor did not follow the manual exactly, the extrapolation is void." CMS's manual states that failure to follow its sampling instructions does not by itself invalidate the sample or projection. A challenge generally must show actual statistical invalidity or an error that affects the overpayment amount.

• "Appealing individual claims does not matter." It matters a great deal. Overturning sample claims can significantly reduce the extrapolated overpayment.

Frequently Asked Questions

What is Medicare extrapolation?

It is the use of a statistically valid random sample of claims to estimate the total overpayment across a larger universe of claims.

When can a Medicare contractor use extrapolation?

Under Section 1893(f)(3) of the Social Security Act, only when there is a sustained or high level of payment error or when documented educational intervention has failed to correct the error.

Why is the demand usually less than the point estimate?

CMS's manual generally directs contractors to demand the lower limit of a one-sided 90 percent confidence interval, which accounts for sampling uncertainty in the provider's favor.

Is there a minimum sample size?

CMS's manual does not set a minimum sample size and states that a small sample does not by itself make an estimate invalid.

Can extrapolated overpayments be appealed?

Yes. Providers can appeal both the individual sample claim determinations and, with some limits, the validity of the statistical sampling and extrapolation methodology.

Conclusion

Extrapolation is how a small audit sample can turn into a large Medicare overpayment demand. When a contractor determines that there is a sustained or high level of payment error, or that education has failed, it can review a random sample of claims and project the results to the full universe, usually demanding the lower limit of a 90 percent confidence interval. Understanding how sampling works, what contractors must document, and how appeals can change the results helps providers respond effectively. The best protection is prevention: low error rates, strong documentation, complete responses to record requests, and early involvement of experienced professionals when a statistical sampling review begins.

How HealthBridge Can Help

HealthBridge is a healthcare management consulting firm that helps providers prepare for and respond to Medicare reviews that involve statistical sampling. We help organize documentation for sampled claims, conduct internal audits to reduce error rates, coordinate with statistical experts and legal counsel, evaluate demand letters, and support appeals. If you are facing a sampling review or an extrapolated overpayment, please get in touch with us. We will be happy to help you with this topic.

References

• Centers for Medicare & Medicaid Services. Medicare Program Integrity Manual, Chapter 8, Section 8.4, Use of Statistical Sampling for Overpayment Estimation. https://www.cms.gov/regulations-and-guidance/guidance/manuals/downloads/pim83c08.pdf

• U.S. Social Security Administration. Social Security Act, Section 1893, Medicare Integrity Program. https://www.ssa.gov/OP_Home/ssact/title18/1893.htm

• Office of the Law Revision Counsel, U.S. House of Representatives. 42 U.S.C. 1395ddd, Medicare Integrity Program. https://uscode.house.gov/view.xhtml?req=granuleid:USC-prelim-title42-section1395ddd&num=0&edition=prelim

• Centers for Medicare & Medicaid Services. CMS Ruling 86-1. https://www.cms.gov/regulations-and-guidance/guidance/rulings/downloads/cmsr8601.pdf

• U.S. Department of Health and Human Services, Departmental Appeals Board. Medicare Appeals Council Decision, Lakeside Foot Clinic (October 15, 2009). https://www.hhs.gov/sites/default/files/static/dab/decisions/council-decisions/lakeside_footclinic.pdf

• U.S. Department of Health and Human Services, Office of Inspector General. RAT-STATS Statistical Software. https://oig.hhs.gov/compliance/rat-stats/

• Electronic Code of Federal Regulations. 42 CFR 405.379, Limitation on Recoupment of Provider and Supplier Overpayments. https://www.ecfr.gov/current/title-42/section-405.379

• Electronic Code of Federal Regulations. 42 CFR 405.1032, Issues Before an ALJ or Attorney Adjudicator. https://www.ecfr.gov/current/title-42/section-405.1032

• Electronic Code of Federal Regulations. 42 CFR 405.1064, ALJ Decisions Involving Statistical Samples. https://www.ecfr.gov/current/title-42/section-405.1064

• Electronic Code of Federal Regulations. 42 CFR 401.305, Requirements for Reporting and Returning of Overpayments. https://www.ecfr.gov/current/title-42/chapter-IV/subchapter-A/part-401/subpart-D/section-401.305

• Centers for Medicare & Medicaid Services. Original Medicare (Fee-for-Service) Appeals. https://www.cms.gov/medicare/appeals-grievances/fee-for-service

• Centers for Medicare & Medicaid Services. Medicare Financial Management Manual, Chapter 3, Overpayments. https://www.cms.gov/regulations-and-guidance/guidance/manuals/downloads/fin106c03.pdf