The Dark Side of AI in Medicare Advantage: When Value-Based Payment Eclipses Value-Based Care

In the world of Medicare Advantage, a disturbing shift is occurring. What began as a noble push toward value-based care has increasingly morphed into an aggressive pursuit of value-based payment optimization, with artificial intelligence serving as the latest tool in this concerning transformation.

The scale of the issue is staggering. A Center for American Progress (CAP) analysis estimates MA plans are overpaid by 22 to 39 percent, translating to $83 billion to $127 billion in overpayments in 2024 alone. Physicians for a National Health Program (PNHP) estimates over $100 billion in annual overpayments, citing factors like attracting healthier enrollees, upcoding disease severity, and flawed price-setting procedures. And the Office of Inspector General (OIG) estimated $7.5 billion in overpayments in 2023 due to Health Risk Assessments (HRAs) that led to upcoding without additional care.

Meanwhile, recent legal actions highlight systemic issues. Cigna recently settled for $172 million over coding practices, UnitedHealthcare lost a significant court case regarding Medicare Advantage practices, and the Office of Inspector General has launched an investigation into Aetna’s practices.

From subtle to overt

More troubling is how openly this shift is now acknowledged in the industry. Some healthcare technology vendors, who once carefully couched their language around “finding unaddressed conditions,” now boldly advertise their ability to increase billing by 20% or more. This transformation reflects a fundamental deviation from the original intent of value-based care programs.

The mechanics of this shift are particularly concerning when examining how AI and technology are being deployed. Rather than using these tools to improve patient care, they’re increasingly being weaponized for revenue optimization. A recent Wall Street Journal investigation revealed how some organizations use AI-powered checklists to identify potential diagnoses that could increase reimbursement. In many cases, these “identified conditions” lack any corresponding treatment plans or clinical documentation.

Gaming the system

The practice extends beyond simple coding optimization. Consider the case of elderly patients who bruise easily — a normal aging process. Instead of acknowledging this natural occurrence, some practices are coding these instances as complex thromboembolytic conditions, despite the absence of any treatment plan or medical necessity. Similar patterns emerge with conditions like diabetic cataracts or depression, where minimal symptoms are leveraged for maximum reimbursement impact.

When diagnoses lack proper supporting documentation and patients are tagged with conditions they don’t actually have, it can represent real risks to patient care. In emergency situations, for example, falsely documented conditions could contraindicate potentially life-saving procedures, putting patients at unnecessary risk.

The financial implications ripple throughout the healthcare system. Hospitals, particularly in rural areas, are increasingly dropping Medicare Advantage plans due to unsustainable economics. While payers report robust profits, healthcare providers face shrinking margins, creating a dangerous imbalance in the system.

Technology at a crossroads

What’s particularly troubling is how AI and large language models (LLMs) are being positioned in this landscape. While these technologies have tremendous potential for improving patient care by identifying truly missed conditions and ensuring proper treatment, they’re increasingly being deployed primarily as revenue optimization tools. Some organizations are even offering revenue-sharing arrangements based on increased payments achieved through their AI-powered coding solutions.

A reckoning may be on the horizon. The current wave of investigations and legal actions suggests growing scrutiny. The prospect of whistleblower actions — incentivized by potential rewards of up to 10% of recovered funds — adds another layer of risk for organizations engaging in aggressive coding practices. Additionally, as patients become more aware that they’re being tagged with diagnoses they don’t have, which can affect their ability to obtain other types of insurance, public pressure may mount.

Back to basics

The solution lies in returning to the fundamental principles of value-based care. When healthcare organizations focus on genuine patient care under value-based programs, it naturally protects both the patient and the system’s financial sustainability. Conversely, when the focus shifts primarily to payment optimization, neither the patient nor the system benefits. Medicare’s financial stability is undermined, and patients face potential harm from incorrect diagnoses in their medical records.

The distinction between value-based care and value-based payment isn’t merely semantic — it represents a crucial fork in the road for healthcare delivery. As AI and other technologies become more prevalent in healthcare, we must ensure they’re deployed in service of improving patient care rather than merely optimizing revenue. The future of Medicare Advantage, and potentially the broader Medicare system, may depend on our ability to maintain this critical focus on actual patient care over payment optimization.

Photo: atibodyphoto, Getty Images

This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.