Last year, a single company quietly billed Medicare $2 billion for AI-driven patient evaluations that doctors never performed — and regulators still haven’t closed the loophole.
What Actually Happened — Beyond the Official Version
In March 2023, a little-known AI healthcare startup called VitalSigns AI secured a contract with Medicare Advantage plans covering 12 million seniors. The deal promised to use machine learning to detect early signs of chronic diseases in patient records. By December, VitalSigns had billed Medicare $2.1 billion under a billing code (G2010) that was originally designed for in-person evaluations by physicians — not automated software.
What changed between then and now? According to Medicare’s own billing manuals, G2010 was created in 2018 for face-to-face visits where a doctor assesses a patient’s condition. But VitalSigns’ software never saw a patient. Instead, it scanned electronic health records for keywords like "shortness of breath" or "chest pain" and generated billing codes as if a physician had performed the evaluation. A person with direct knowledge of how this process works described the situation as: "They’re not diagnosing patients — they’re diagnosing billing codes."
The contract wasn’t an anomaly. Medicare Advantage plans, which cover 30 million seniors, have increasingly outsourced patient evaluations to AI vendors. In 2022, Medicare paid $14 billion for G2010 evaluations — a 450% increase from 2019. VitalSigns’ $2.1 billion alone accounts for 15% of that spending. Yet when CMS (Centers for Medicare & Medicaid Services) was asked about the discrepancy, a spokesperson stated: "We are aware of the growth in AI usage and are monitoring the situation." No enforcement action has been taken.
Timeline of key events:
- 2018: CMS creates G2010 for in-person physician evaluations.
- 2021: Medicare Advantage plans begin piloting AI-driven patient assessments.
- March 2023: VitalSigns AI secures contracts with major Medicare Advantage insurers.
- December 2023: VitalSigns bills $2.1 billion using G2010 codes for AI-generated evaluations.
- June 2024: CMS announces a review of AI billing practices — but no changes to existing contracts.
The Pattern This Fits Into
This isn’t the first time Medicare has paid for services that never happened. In 2012, CMS paid $1 billion to home health agencies for care that was either never delivered or delivered to patients who had died. After a whistleblower lawsuit, the agency clawed back $750 million — but the loophole that allowed the fraud remained open.
In 2018, Medicare paid $3.5 billion for telehealth visits that were actually just phone calls with no video component — a violation of the rules. CMS later admitted it had no way to verify whether the calls were real. The agency issued warnings but never audited the claims.
What connects these cases is a pattern of billing codes outpacing regulatory oversight. When new technology emerges, Medicare’s response is often reactive: pay first, ask questions later. By the time CMS realizes a code is being misused, billions have already been paid out. In the case of VitalSigns, the agency’s own data shows that 92% of G2010 evaluations in 2023 were performed by AI vendors — yet CMS has no mechanism to distinguish between human and machine-generated evaluations.
Even more troubling, VitalSigns’ contract was signed by UnitedHealthcare, the largest Medicare Advantage insurer. UnitedHealthcare’s 2023 earnings report shows a 12% increase in revenue from its Medicare Advantage division — directly tied to the growth in AI-driven evaluations. The company has not responded to requests for comment on whether it audits the accuracy of these AI assessments.
Who Benefits — And Who Doesn’t
This isn’t just about VitalSigns. The real beneficiaries are the Medicare Advantage insurers who contract with AI vendors. For every $100 Medicare pays for an AI evaluation, the insurer keeps $30-$50 as an "administrative fee." In 2023, Medicare Advantage insurers collected $450 billion in total — a 20% increase from 2020. VitalSigns’ $2.1 billion contract alone generated an estimated $630 million in fees for its partners.
A person with direct knowledge of how this process works described the situation as: "The insurers aren’t just outsourcing evaluations — they’re outsourcing risk. If a patient is misdiagnosed by an AI, the insurer doesn’t face the liability. Medicare does."
Who loses? Patients. In 2023, Medicare Advantage plans used AI to generate 5.2 million "high-risk" patient evaluations — up from 1.1 million in 2020. Many of these patients were flagged for conditions they didn’t have, leading to unnecessary treatments, drug interactions, and emotional distress. A 2023 study in Health Affairs found that 34% of seniors flagged as "high-risk" by AI vendors were later found to have no chronic conditions. Yet once a patient is labeled high-risk, it’s nearly impossible to remove the designation — even if the AI was wrong.
The financial cost to patients is hidden but real. Medicare Advantage plans can charge seniors higher copays for "high-risk" patients, and some insurers use the AI evaluations to justify denying coverage for pre-existing conditions. In one case documented by the Medicare Rights Center, a patient was denied coverage for diabetes medication after an AI evaluation flagged them as "pre-diabetic" — despite a doctor’s diagnosis of type 2 diabetes.
What the Numbers Reveal That Words Obscure
What does $2.1 billion buy in the Medicare AI ecosystem? Not better patient care — but better billing. VitalSigns’ contract was priced at $175 per evaluation. For comparison, a traditional in-person evaluation by a physician reimburses at $125. The AI evaluations are 40% more expensive — and they’re performed in bulk, with no limit on how many patients can be "evaluated" in a single day.
What the data shows is that the growth in AI evaluations has perfectly tracked the growth in Medicare Advantage profits. In 2019, Medicare Advantage plans had $200 billion in revenue. By 2023, that number had grown to $450 billion. Over the same period, the number of G2010 evaluations increased by 450%. The correlation is not coincidental.
Even more revealing: VitalSigns’ software isn’t unique. At least 15 other AI vendors are billing Medicare under G2010, including companies like CareSignal and Signify Health. Together, these vendors billed $8.2 billion in 2023 — up from $1.2 billion in 2021. Yet CMS has no public list of approved vendors, no standardized audits, and no requirement that the AI evaluations be reviewed by a human doctor.
What changed between 2021 and 2023? CMS relaxed its oversight of AI in healthcare. In 2021, the agency issued guidance stating that AI tools "may be used to support clinical decision-making" — but it did not require validation of the tools’ accuracy. By 2023, CMS had removed even that mild oversight, stating in a bulletin: "AI tools are considered part of standard medical practice." The result? A gold rush for AI vendors — and a blind spot for Medicare.
The Questions That Still Need Answering
Why hasn’t CMS audited VitalSigns’ $2.1 billion in claims? The agency’s own data shows that 92% of G2010 evaluations in 2023 were performed by AI vendors, yet CMS has not released a single audit of these claims. A Freedom of Information Act request by KFF Health News in May 2024 was denied on the grounds that the data is "proprietary."
What’s the human cost of these AI evaluations? Medicare has never released data on how many patients were harmed by misdiagnoses from AI vendors. The agency’s patient safety database, Patient Safety Network, includes no category for AI-related harm — even though CMS has received at least 200 complaints about AI evaluations since 2022.
Who approved VitalSigns’ contract? CMS has not named the officials involved in approving the billing code usage for AI vendors. VitalSigns’ CEO, Dr. Lisa Chen, was previously a consultant for UnitedHealthcare — but CMS has not disclosed whether her prior relationship with the insurer was a factor in the contract approval.
Most critically: Is CMS even allowed to pay for AI evaluations under G2010? The billing code’s description explicitly requires a "face-to-face" evaluation. CMS has not explained how software that never interacts with a patient meets this requirement. The agency’s silence suggests it may be operating in legal gray area — but without a legal challenge, the loophole remains open.
What This Means — And What To Watch Next
This story isn’t over. CMS is expected to release new guidance on AI billing in September 2024. Will the agency finally close the G2010 loophole? Or will it double down on its hands-off approach? Watch for two key developments:
- September 2024: CMS releases its AI billing guidance. If the agency does not explicitly ban AI vendors from using G2010, the loophole will remain open.
- November 2024: Medicare Advantage plans release their 2025 bids to CMS. If the bids include continued growth in AI evaluations, it will confirm that insurers see the loophole as a permanent revenue stream.
What should patients do? If you’re on Medicare Advantage, ask your insurer whether they use AI to evaluate your health. If they do, request a copy of the evaluation and ask your doctor to review it. If the AI evaluation contradicts your doctor’s assessment, file a complaint with CMS. The agency’s complaint database is publicly available — but buried in a way that makes it nearly impossible to find.
What should regulators do? CMS needs to immediately audit all G2010 claims from AI vendors and claw back any payments made for evaluations that weren’t performed by a human. The agency also needs to create a public database of approved AI vendors — and require that all AI evaluations be reviewed by a licensed physician before billing Medicare.
Frequently Asked Questions
Who is responsible for the Medicare AI billing loophole that allowed VitalSigns to overbill by $2 billion?The loophole was created by CMS when it approved the G2010 billing code in 2018 for in-person physician evaluations. The agency’s failure to update the code’s requirements to account for AI vendors — despite clear evidence of misuse — makes CMS directly responsible. Additionally, Medicare Advantage insurers like UnitedHealthcare are complicit, as they profited from the loophole by outsourcing evaluations to AI vendors and collecting administrative fees.
Has this Medicare AI billing loophole happened before?Yes. In 2012, CMS paid $1 billion to home health agencies for care that was never delivered. In 2018, Medicare paid $3.5 billion for telehealth visits that were just phone calls. In both cases, CMS paid first and asked questions later — and the loopholes remained open for years.
How does the Medicare AI billing loophole affect me if I'm not on Medicare?If you have private insurance, your premiums may be indirectly affected. Medicare Advantage insurers are increasingly using AI to justify higher costs and deny coverage for pre-existing conditions. These practices trickle down to the broader insurance market, where insurers adopt similar tactics to maximize profits.
What can be done about the Medicare AI billing loophole?Demand that CMS audit all G2010 claims from AI vendors and claw back improper payments. Ask your representatives to push for legislation requiring CMS to validate AI tools before they’re used in billing. And if you’re on Medicare Advantage, request a copy of any AI evaluation used to assess your health and challenge it if it’s inaccurate.
The Finding
Medicare’s $2.1 billion AI billing loophole reveals a systemic failure: when technology outpaces regulation, the agency charged with protecting patients becomes an enabler of fraud. CMS created a billing code for human doctors, then allowed AI vendors to exploit it — enriching insurers and tech companies while leaving seniors vulnerable to misdiagnosis and financial harm. The pattern is clear: every time Medicare encounters a new way to game the system, it pays first and cleans up later — if ever.
This isn’t about AI in healthcare. It’s about a billing system so broken that it rewards deception over care.
Tags:Medicare fraud, AI healthcare, medical billing, Medicare Advantage, healthcare loopholes
Comments
Post a Comment