Retrieving Billions in Overpayments by CMS

Amid swirling accusations that Medicare Advantage Organizations (MAOs) are overbilling the U.S. government and calls for better oversight, the Centers for Medicare & Medicaid Services announced in early February that it would investigate overbilling by those plans. They expect to recoup 4.7 billion dollars through this program.

This article focuses on the relatively young technologies that enable CMS to uncover overbillings, whether they be errors or fraud. The article is based on an interview with Kel Pults, chief clinical officer and vice president of government programs for MediQuant. A future article will explain how Medicare Advantage plans are trying to improve data collection and reporting, and how AI helps.

Challenges of Investigating Overpayments

Undeserved payments are needles lurking in the haystack of 135 million Americans enrolled in Medicare, Medicaid, and the Children’s Health Insurance Program (CHIP). But the needles pile up fast. Improper payments for Medicare alone were estimated to be $43 billion for a single year.

Conventional fee-for-service payments, which reimburse a doctor for each visit or other intervention, are hard enough to investigate. It’s even harder to determine payments for more open-ended fee-for-value plans, which include Medicare Advantage.

Medicare Advantage programs are reimbursed on the basis on how much they are expected to spend on a patient. A person who has smoked for 60 years and is diagnosed with congestive heart failure and diabetes is expected to cost the plan more money than a professional dancer who has maintained an active life style. The difference between these situations drive differences in per-patient CMS reimbursements, which are called “risk-adjusted payments.”

But how sick is a patient, really? There are lots of ways to make patients seem sicker than they really are by exaggerating parts of their histories.

Currently, Pults told me, CMS is focusing on overbillings during the height of the COVID-19 pandemic. Although the recent CMS rule doesn’t mention COVID-19 explicitly, the crisis around that pandemic provided a new loophole for Medicare Advantage plans to recoup overpayments.

Think back to the height of pandemic, especially pre-vaccine. In 2020 and 2021, doctors, nurses, and other staff risked their lives hourly to treat patients who had COVID-19, often without knowing who had it. To provide incentives for both the front-line staff and the clinical institutions employing them, CMS offered bonuses, or add-on payments, for treating patients who had the disease.

In short, for a patient who died of COVID-19 the provider received a higher reimbursement than for a patient who died of cancer, even if the money expended by the clinician and payer on the two victims was comparable.

The add-on payment created an unfortunate incentive to mark the primary cause of death COVID-19 when it might not have been. If a patient was admitted to a hospital for COPD or cancer and was subsequently found to have COVID-19, they shouldn’t trigger an add-on payment because the “primary cause of death” was COPD or cancer. It’s possible that providers and plans reclassified the patients as COVID-19 victims in order to get the add-on payments. This was particularly ironic if the patient developed COVID-19 as a result of entering the hospital for a different condition.

How can such overpayments be uncovered? One has to intensively examine the diagnosis and decisions leading up to the original hospital admission. It’s unfeasible to review every hospital admission for 2020 and 2021, so CMS will need advanced computing tools to choose suspicious deaths.

Preserving Data Integrity

For 24 years, MediQuant has offered cloud-based active archiving to the healthcare industry. The archive is a layer above generic services such as Amazon’s S3, providing an interface to retrieve and search old clinical data. Figure 1, for instance, shows a screen offering various kinds of information on a patient. In an active archive, discrete data is collected and stored in a database table with a high level of detail. This type of data is both measurable and reportable – which can help healthcare providers meet regulatory requirements and to provide support during an audit.

A New Release of Information Request is a form listing types of available reports, broken into clinical categories.
New Release of Information Request

Figure 1: New Release of Information Request

The task of determining what really caused a patient death involves searching several histories, each of which could be made of up different sections stored in different documents. Each relevant item, such as a vital sign, might be stored separately. For instance, one red flag calling for a review is if COVID-19 is listed as the primary cause of death and another was listed as secondary.

Legacy record systems add extra fragmentation to the records. Pults said that one of their clients had 250+ legacy systems, which is not uncommon.

MediQuant can use machine learning to find the relationships among all these scattered facts, look for particular diagnosis codes, and check certain parameters to narrow down what to review out of millions of records.

About the author

Andy Oram

Andy is a writer and editor in the computer field. His editorial projects have ranged from a legal guide covering intellectual property to a graphic novel about teenage hackers. A correspondent for Healthcare IT Today, Andy also writes often on policy issues related to the Internet and on trends affecting technical innovation and its effects on society. Print publications where his work has appeared include The Economist, Communications of the ACM, Copyright World, the Journal of Information Technology & Politics, Vanguardia Dossier, and Internet Law and Business. Conferences where he has presented talks include O'Reilly's Open Source Convention, FISL (Brazil), FOSDEM (Brussels), DebConf, and LibrePlanet. Andy participates in the Association for Computing Machinery's policy organization, named USTPC, and is on the editorial board of the Linux Professional Institute.

   

Categories