Denying Coverage with AI: CMS’s New Medicare Model
The Trump Administration is bringing prior authorizations (PAs), the mandatory preapproval before insurance covers health care services, to Traditional Medicare. The government plans on using contracts with perverse incentives for companies that use artificial intelligence (AI) to carry out the work.
On June 27, the Centers for Medicare and Medicaid Services’ (CMS) Innovation Center announced a new public-private partnership model called the Wasteful and Inappropriate Service Reduction (WISeR) Model – lasting from 2026-2031. WISeR’s purpose is to reduce wasteful spending in Medicare. CMS cites the Medicare Payment Advisory Commission’s (MedPAC) finding that, in 2022, taxpayers financed $5.8 billion in low-value care, defined as “the provision of a service that has little or no clinical benefit or care in which the risk of harm from the service outweighs its potential benefit.”
The WISeR model will have CMS contract with “companies with expertise providing recommendations on medical necessity of coverage for payers using enhanced technology like AI.” While using AI, CMS will require that these companies have clinicians ultimately conduct reviews to validate medical determinations whether a product or service is wasteful. Prior authorizations currently exist for a relatively small number of services in Traditional medicare; however, WISeR will greatly expand these and work with companies with prior authorization experience in Medicare Advantage.
Yet, the introduction of prior authorizations – especially in this specific public-private partnership model – threatens to cause more harm than good. Firstly, PAs already exist in the privatized Medicare Advantage system along with standard private insurance, and we know they cause delays and denials of critical medical coverage. A 2024 American Medical Association survey of 1,000 physicians foundthat 93 percent reported prior authorizations delayed access to necessary care always (15 percent), often (42 percent), or sometimes (36 percent). Difficulties with the PA process can also lead to abandoning recommended care, with 82 percent of physicians reporting such abandonment always (2 percent), often (20 percent), or sometimes (60 percent).
In 2023, providers had to file around 50 million PA requests for patients in Medicare Advantage, with insurance companies denying around 3.2 million (6.4%). Yet, while patients and providers appealed denials only 11.7% of the time, they had an 81.7% success rate. Simultaneously, a Government Accountability Office (GAO) reportabout MA observed in a 2022 random sample of PA denials that 13% were improper because the care should have been covered under Medicare rules. If this held true for 2023, then – as only 9.5% of denials are successfully appealed – prior authorizations would have led to around 112,000 treatments getting improperly denied.
Prior authorizations also create an additional administrative burden on healthcare providers and will likely increase such costs. The AMA survey noted that physicians spend 13 extra hours per week completing PAs, and 40 percent of physicians have staff exclusively working on PAs. The US spends twice the amount on health care per person than other developed nations even though we enjoy worse health outcomes, and a significant contributor to these costs is higher administrative costs. We spend roughly twice as much on administration, and much of this comes from the billing and insurance complexities of the country’s predominantly for-profit, private insurance model.
Comparatively, Medicare has had significantly lower administrative costs, with 1.1 percent of spending going towards administration in 2023, compared to levels varying between roughly 12-18 percent over the last decade for private insurance. Introducing PAs in Traditional Medicare would not only increase the administrative burden on providers, but it will increase the costs directly borne by taxpayers through higher spending on administration.
Using AI To Deny Care
WISeR’s model also may incentivize corporate exploitation directly in Traditional Medicare through its contracting with outside companies. An October 2024 report by the US Senate Permanent Subcommittee on Investigations detailed how three of the largest insurance companies in MA – United Healthcare, Humana, and CVS – disproportionately denied PA requests to cover post-acute care for patients. Internal corporate documents from United Healthcare specifically touted the benefits of their Machine-Assisted Prior Authorization (MAP) technology, and CVS began employing an AI program to cover less post-acute care in 2021. These systems were meant to deny care, as one March 2022 CVS meeting documented that reducing PAs would make financial losses “too large to move forward.”
In WISeR, CMS will contract with companies that use AI, and it will reward those companies by having them “receive a percentage of the savings associated with averted wasteful, inappropriate care as a result of their reviews.” Similar to the Senate report’s findings regarding Medicare Advantage, the WISeR model incentivizes these third parties to deny as much care as possible to maximize profits. There is no evidence that these for-profit actors would prioritize ensuring access to health care and denying low-value care as narrowly as possible over simply maximizing their profits.
On top of perverse financial incentives, WISeR continues the Trump administration’s unsubstantiated faith in AI. As the Senate report details, reliance on AI and machine-models is meant to increase the volume of denials, and – as Medicaid and CHIP Payment and Access Commission (MACPAC) has also noted – companies can bias and manipulate the algorithms to overemphasize denying care.
AI has also not proven to be reliable, especially if CMS wants to allow private companies to use it to determine whether or not American seniors will get coverage for health care. Reporting of the FDA’s “Elsa” AI model suggests that it has difficultygetting facts correct, and the White House’s rollout of the MAHA (Make American Healthy Again) Report was mired by incorrect AI-generated references to studies, several of which did not exist.
Attempting to stop $5.8 billion in waste that goes to unnecessary and inappropriate care is a legitimate government interest. However – given (1) the introduction of PAs which have delayed and denied important services in private insurance, (2) CMS’s model of increasing administrative costs which already help make our health care system more expensive than other nations, (3) giving for-profit companies perverse incentives to deny care to maximize their own profits, and (4) relying on unproven AI models that can affect whether someone receives life-saving care – it is likely that the current approach will cost and hurt Americans more than it helps.
It is additionally essential to place the WISeR model in the overall health care policy context: MedPAC estimated that CMS will waste $84 billion in 2025 by overpaying the private insurance companies in Medicare Advantage – like United Healthcare, Humana, Aetna, and Elevance Health. Other estimates in recent years have pegged annual overpayments from around $80 billion to as high as $140 billion. Based on MedPAC’s 2025 estimate, the Committee for a Responsible Federal Budget projected$1.2 trillion in MA overpayments over the next decade.
Some Republican senators like Senator Bill Cassidy (R-LA) suggested including in the massive reconciliation legislation that Congress passed last week an attack on a key method insurance companies employ to steal taxpayer money: upcoding, meaning making patients appear sicker than they are to get more taxpayer money to cover them. Yet, the Administration and congressional Republicans ultimately did not push to include in the bill any provision dealing with this vast source of waste – which is more than 14 times larger than the $5.8 billion that PAs will allegedly address.
Tackling this waste threatens the large profits of powerful MA insurance companies, who spent $332 million on lobbying from 2020-2024. The MA industry’s lead lobbying group, the Better Medicare Alliance, specifically ramped up its lobbying spending in 2024 to counter growing critiques of overpayments. Former Representative Jim Greenwood (R-PA) – who helped establish the current MA program – remarked, “[t]he program’s original safeguards against excessive billing and cherry-picking enrollees have proven too weak in the face of powerful lobbying, limited oversight, and manipulative practices.”
Addressing wasteful and unnecessary spending in American health care is a legitimate focus for the government. Insurance companies have stolen hundreds of billions of dollars from American taxpayers in Medicare Advantage. Hospitals and other providers charge Americans significantly higher prices for the exact sameservices compared to other developed nations. Pharmaceutical companies research, develop, and market to doctors and patients alike products which do not provide added clinical benefits compared to existing, alternative treatments. Funnelling money to for-profit companies by rewarding them for denying health care coverage through using unproven, manipulable AI models is not the answer.
This first appeared on CERP.
The post Denying Coverage with AI: CMS’s New Medicare Model appeared first on CounterPunch.org.