The murder of UnitedHealthcare CEO Brian Thompson ignited a nationwide discussion on the escalating healthcare costs and the prevalence of claim denials.

Today, AI is employed to deny millions of health insurance claims, while two-thirds of healthcare organizations intend to allocate more resources to AI in the coming three years.

In a media briefing on December 20, hosted by Ethnic Media Services, a panel of experts discussed the rise of AI usage in denying health insurance claims. 

Speakers

  • Dr. Katherine Hempstead, senior policy officer at the Robert Wood Johnson Foundation
  • Dr. Miranda Yaver, Assistant Professor of Health Policy and Management at the University of Pittsburgh and author of “Coverage Denied: How Health Insurers Drive Inequality in the United States”
  • California State Senator Josh Becker, author of SB 1120, the Physicians Make Decisions Act

The Healthcare Landscape and AI

Dr. Katherine Hempstead aptly described health insurance as the most dysfunctional, fragile, and poorly functioning example of the complex relationship between policyholders and insurers within the insurance industry.

She highlighted several factors that contribute to mistrust in this sector, contrasting it with the one-time nature of claims in life insurance. Unlike life insurance, health insurance involves ongoing interactions between insurers and policyholders, often during times of illness or when individuals are concerned about their future health. This prolonged contact creates a fundamental powerlessness for policyholders when providers assert the necessity of certain services, only to have insurers deny them. Additionally, there is significant fragmentation in coverage between states and different insurance plans.

A recent Gallup poll revealed that Americans’ satisfaction with the U.S. healthcare system has reached its lowest point since 2001. Only 44% of respondents rated the quality of healthcare as excellent or good, while 28% expressed positive sentiments about the coverage.

So, why are Americans growing increasingly dissatisfied with this system?

Ironically, the expansion of managed care and Medicare Advantage plans, which are considered positive developments, has brought to the forefront issues related to denied claims for essential services and life-changing medications, such as GLP drugs for diabetes and obesity.

Medicare Advantage (MA) is a Medicare-approved private health plan that caters to seniors and individuals with disabilities.

An October 2024 report released by the U.S. Senate revealed that the three largest MA companies—UnitedHealthcare, Humana, and CVS—collectively cover 60% of enrollees. These companies have significantly increased the number of algorithm-automated claim denials between 2019 and 2022, while simultaneously systematically limiting post-acute (PA) care, including home health services and long-term hospital care, to maximize profits.

UnitedHealthcare’s PA denial rate rose from 10.9% in 2020 to 22.7% in 2022 as the company accelerated automation. Humana’s PA denial rate was 16 times higher compared to its overall denials, while CVS’ PA denials remained unchanged despite a surge in PA requests by 57.5%.

Hampstead explained that when people appeal decisions, especially when the press is involved, the decision often changes, which can lead to cynicism among those who feel insurers will deny services until exposed or pressured. This drives inequity by disadvantaging individuals who don’t speak English as their first language, are lower-income, or less educated, making them less likely to dispute a decision.

At the core of the increasing claim denials lies the use of prior authorization, a process requiring insurance approval before a patient can receive a service or prescription.

Dr. Miranda Yaver acknowledged that tools like prior authorization and automation are intended to expedite delays, curb over-prescriptions, and impose limits on healthcare spending, which is high in the U.S. However, errors are acceptable in low-stake settings but become problematic in healthcare. A significant portion of healthcare spending, amounting to nearly $1 out of every $5 spent in the U.S., is allocated to administrative costs, representing 17% of the national GDP. In contrast, in 1960, healthcare accounted for only 5% of the national GDP, or $1 out of every $20 spent.

Furthermore, 30 cents on every medical dollar spent—approximately $750 billion annually—goes towards administrative costs.

Although U.S. healthcare spending has surged, so have out-of-pocket costs for all Americans, regardless of health insurance coverage, for healthcare not covered by a plan. These costs have risen dramatically from $115 in 1970 (adjusted for inflation in 2024, $677) to $1,425 per person in 2022, excluding monthly health insurance premiums.

For her upcoming book, “Coverage Denied: How Health Insurers Drive Inequality in the United States,” scheduled for publication in 2026, Yaver conducted a nationwide survey of 1,340 adults and found that 36% had experienced at least one claim denial, with 60% of those facing multiple denials.

“No matter who you are, you’re vulnerable to this, but the effects can cause the most inequities for people from marginalized groups less likely to realize they can appeal,” Yaver emphasized. “Even when people are successful in appealing these automated denials, we need to think about the equity costs … It’s time-consuming, physically and emotionally taxing, and can be expensive.”

Addressing AI Concerns

California State Senator Josh Becker expressed his dismay that the United States is essentially the only industrialized country operating in this manner. He authored the Physicians Make Decisions Act (SB 1120), which aims to limit the scope of AI by requiring doctors to make final decisions on patient treatment and oversee decisions made by AI systems, such as claim requests and prior authorization.

The bill was signed by Governor Newsom last September and will come into effect on January 5, 2025.

“We need the human element to ensure that health care decisions prioritize patient well-being over automated processes,” explained Becker. His bill, which faced opposition from insurance companies and support from physicians’ groups like the California Medical Association, gained less attention initially. However, it has since attracted interest from other states and the federal government, prompting hopes that it will inspire similar actions.

While AI has significant roles in disease detection and image analysis, Becker emphasized the importance of ensuring that only trained physicians can make appropriate healthcare decisions. He highlighted the risks associated with this approach, citing a 2023 lawsuit against Cigna Healthcare. The lawsuit alleged that Cigna used AI algorithms to reject over 300,000 pre-approved claims in two months, with an alleged 1.2 seconds spent on each rejection. Furthermore, 80% of customer-appealed claims were overturned.

Cigna covers or administers health plans for 18 million Americans, and the lawsuit claimed that a single medical director, Cheryl Dopke, rejected approximately 60,000 claims in a single month. In the same year, for-profit health insurance companies made a substantial profit of $70.7 billion, which Becker suggested may be partly due to practices like these denials. He emphasized that these practices contribute to cost containment.

Becker argued that denying claims every 1.2 seconds prevents healthcare providers from considering patients’ needs. He emphasized the need to redirect healthcare dollars towards providing care to those who truly require it.

 

All images provided by EMS.