|
 |
Prominent health insurers are increasingly using artificial intelligence (AI) to process medical claims and prior authorizations—insurer preapprovals required for prescribed care. Once intended to control costs and prevent overprescribing, prior authorization now frequently blocks coverage for care that doctors deem necessary.
UnitedHealthcare, Cigna, and Humana face lawsuits over their use of AI models such as nH Predict and PxDx, which can review coverage decisions in seconds. The companies argue that AI only assists, not decides. The lawsuit alleges these AI-assisted claims are error-prone: In some cases, up to 90 percent of AI-related denials are reversed on appeal to the insurer—if patients can manage the onerous process to appeal, which many do not.
These AI-driven denials are just the newest layer of bureaucracy between Americans and their health care. What few realize is that when this system fails, patients have almost no way to fight back. This is because a little-known 1974 law, the Employee Retirement Income Security Act (ERISA), shields most employer-sponsored plans from accountability.
As these AI tools spread across plan types, the risk of wrongful denials only amplifies.
ERISA Makes Wrongful Insurance Denials More Likely and More Burdensome on Patients
Originally passed to curb pension fraud and mismanagement, ERISA unexpectedly reshaped the health insurance industry by governing self-insured employer plans—where companies pay claims directly or through a third-party administrator, like an insurer. When ERISA was enacted in 1974, only about 6 percent of plans were self-insured. Today, roughly two-thirds of covered workers are on self-insured employer plans, meaning most Americans with employer health coverage now fall under ERISA’s limited protections.
Though ERISA was designed to protect workers, its structure—specifically its limitation on damages and preemption of state law—has created major unintended consequences for health policy, making it extraordinarily difficult to challenge insurance denials and reform prior authorization.
Take, for instance, Frank Wurzbacher, whose story was highlighted in a 1998 Senate hearing on the need for a Patients’ Bill of Rights. He didn’t realize how vulnerable he was when his employer switched insurance carriers as he fought prostate cancer. The new insurer denied coverage for the costly injections that kept his cancer in check, approving only full coverage for castration surgery—a permanent procedure designed to suppress testosterone. Unable to afford the injections, he underwent surgery, only to learn soon after that the insurer reversed its decision.
Because ERISA’s remedy and preemption provisions remain largely unchanged decades later, Frank couldn’t recover monetary damages if he brought a claim, and attorney’s fees would be left to a judge’s discretion. Even if he won, his only remedy would have been approval for the injections he no longer needed.
How ERISA Shields Insurers from Accountability
Frank’s story underscores how little protection nearly 100 million Americans have when navigating health insurance barriers, despite ERISA’s original aim to safeguard workers. The law not only harms less affluent employees but also weakens insurer accountability, allowing wrongful denials to persist.
This is because ERISA bars patients from recovering damages for pain and suffering or lost wages, for instance, allowing only the denied treatment and possible attorney’s fees as a remedy.
As a result, few patients have an incentive to sue. After all, lawyers have every reason to avoid cases that lack a monetary value, and poorer (and even middle-class) workers can’t usually risk bringing a lawsuit on their own. With few lawsuits and no real penalty for wrongful denials, insurers face little incentive to change, especially as AI promises faster, cheaper claim reviews at the patients’ expense.
In May 2025, I surveyed 2,569 US adults, asking whether they would sue over a coverage denial if they could receive the treatment but not money damages, and whether they would sue if they weren’t guaranteed to recover the costs of the lawsuit. Only 6 percent said they would be content to receive the health benefit only—the sole relief ERISA guarantees successful plaintiffs.
Rather, nearly all respondents indicated that, if bringing a case, they would want to recover damages, which ERISA precludes. Lower-income, less-educated, and less healthy respondents were far less inclined to sue under these conditions. In fact, respondents earning $50,000 or less annually were 13 points less likely than wealthier respondents to pursue cases without damages, and 11 points less likely if attorney’s fees were uncertain.
ERISA not only discourages patients from suing—shielding insurers from consequences for wrongful denials—but also disproportionately harms lower-income and marginalized workers, offering no real protection from legal costs.
ERISA Stops States from Improving Health Policy
States have begun addressing the causes of coverage denials—especially prior authorization—and are now moving to oversee insurers’ use of AI tools.
For example, California recently enacted Senate Bill 1120, requiring that any AI-based denial be reviewed by a physician in the appropriate specialty, joining dozens of other states that have adopted similar standards.
Sounds great on paper. But here’s the catch:
Because of how ERISA was written, these reforms don’t apply to most employer-sponsored health plans. Unlike other areas of health federalism—such as marijuana decriminalization, Medicaid expansion, or abortion—ERISA preempts any state law that “relate[s] to” self-insured health plans. As a result, state efforts to expand access or oversight leave the biggest coverage gaps untouched, even as insurers’ use of AI deepens workers’ vulnerability to denial of coverage.
Thus, any health insurance reforms meant to affect the majority of employer-sponsored insurance must come from the notoriously gridlocked Congress.
A 1974 Law That Washington Still Won’t Touch
In the late 1990s, lawmakers tried to fix ERISA through a Patients’ Bill of Rights, but the effort collapsed amid fears of “frivolous lawsuits,” concerns about rising premiums, and Bill Clinton’s impeachment. A decade later, congressional Democrats focused the Affordable Care Act on expanding coverage rather than strengthening protections for the already insured, leaving ERISA’s limits intact.
Insurers are hardly alone in using AI as a cheap substitute for labor-intensive procedures, but when these tools wrongly deny coverage, patients have little legal recourse. Without the ability to recover damages for pain and suffering, insurers face almost no serious penalty for error—errors that are likely to only increase given such tools’ imperfect track record. And because ERISA preempts state law, even the most ambitious state reforms can’t protect most workers.
As AI-driven prior authorization expands—even into traditional Medicare—Americans might finally confront a deeper question behind their coverage barriers: Why do they persist? The answer lies in a 1974 law shielding health plans, fair and unfair alike, from real accountability.
Fireside Stacks is a weekly newsletter from Roosevelt Forward about progressive politics, policy, and economics. If you enjoyed this installment, consider sharing it with your friends.
| ||||||||||
