Medicare Advantage (MA) plans are learning what Medicare providers have long known about the flawed way in which Medicare uses statistical extrapolation in its audits. The Centers for Medicare & Medicaid Services (CMS) has indicated that, for the first time, it intends to apply statistical extrapolations to overpayments and other payment errors by MA plans, which were uncovered in recently released federal audits. CMS conducted 90 audits of MA plans examining billings for approximately 18,000 patients from 2011 through 2013 that resulted in a finding of about $12 million in net overpayments associated with the plans. CMS now intends to extrapolate the results of those audits across the entire membership of each plan – likely millions of patients – and recoup from the plans an estimated $650 million of total overpayments, which CMS has indicated it intends to recoup even though no action has been taken since the audits were conducted.
These circumstances and the issues that the MA plans are pointing out with Medicare’s extrapolation methodology may sound familiar to any Medicare provider that has been subjected to an extrapolated Medicare audit. CMS, its contractors, and other federal agencies, such as the Department of Health and Human Services (HHS) Office of Inspector General (OIG), have used statistical extrapolation against Medicare providers for decades, while strong opposition from the insurance industry has kept it from being applied to audits of MA plans. However, the recent growth of MA plans and the recent release of some CMS audit results have led CMS to begin to apply it to MA plans.
While statistical extrapolation can be a valid and useful auditing tool where the auditor carefully employs sound statistical principles and valid methodologies, Medicare extrapolations are often sloppy, imprecise, and difficult to challenge. As MA plans are now learning, Medicare extrapolations (especially those conducted by contractors) often include errors that skew the results, overestimate the alleged overpayment, or would render the extrapolation outright invalid in an academic or accounting setting, including: over-sampling high-value claims, double-counting claims, using a sample that does not represent the universe to which it is extrapolated, combining dozens of unrelated services or codes into a single sample frame or universe, ignoring information in favor of the provider, failure to use stratification or cluster sampling where the data set used demands it for precision, unacceptably low levels of precision, and demands for astronomical and business-shattering overpayments after having actually reviewed only a small number of claims. Moreover, HHS’s reviewers – who are charged with reviewing these audits for accuracy – will often bend over backwards to explain why an error-riddled extrapolation was appropriate, while the agency has built a fortress of regulations, caselaw, and manual provisions that make it very difficult to challenge even the most flawed extrapolations.