How AI Algorithms Decide Your Health Insurance Coverage—and Why It Matters

The Impact of AI Algorithms on Health Insurance Coverage

Over the last ten years, health insurance companies have increasingly adopted artificial intelligence algorithms to make coverage decisions. Unlike AI in hospitals or clinics that assists with diagnosing or treating patients, insurers use AI to determine whether to approve payment for treatments recommended by physicians.

A common example is prior authorization, where a doctor must get the insurer’s approval before providing care. Many insurers rely on algorithms to decide if the requested care is “medically necessary” and therefore eligible for coverage. These systems also influence decisions on the extent of care—for instance, the number of hospital days covered after surgery.

If an insurer denies coverage for a treatment your doctor recommends, you typically face three choices:

• Appeal the denial, which can be time-consuming, costly, and require expert assistance. In fact, only 1 in 500 denials are appealed.
• Agree to an alternative treatment that the insurer will cover.
• Pay out of pocket, which is often unrealistic due to high healthcare costs.

Patterns of Withholding Care

Insurers presumably input patient health records and relevant data into AI algorithms and compare that against current medical standards to decide on claims. However, insurers typically keep these algorithms confidential, making it impossible to know exactly how coverage decisions are made.

Using AI to review claims saves insurers time and resources by reducing the need for medical professional review. But beyond that, if an AI system denies a valid claim and the patient appeals, the appeal can drag on for years. If a patient is seriously ill, insurers might financially benefit if the process delays resolution until after the patient’s death.

Insurers argue that patients can pay for treatments themselves if coverage is denied. This raises concern that algorithms could be used to withhold costly care for long-term or terminal conditions such as chronic illnesses or disabilities. Research supports this, showing patients with chronic conditions face higher denial rates and subsequent harm.

There are also disparities: Black, Hispanic, and other nonwhite patients, as well as individuals identifying as LGBTQ+, are more likely to experience denials. Some studies suggest prior authorization may actually increase overall healthcare costs rather than reduce them.

The argument that patients can always pay for denied treatments ignores reality. Many cannot afford the care they need, and these decisions have serious health consequences.

Regulatory Landscape and Challenges

Unlike AI tools used in medical diagnosis or treatment, insurance coverage algorithms face little regulation. They are not subject to Food and Drug Administration (FDA) review, and insurers claim their algorithms are proprietary trade secrets. This lack of transparency means no external evaluation ensures these tools are safe, fair, or effective.

Some progress is happening. The Centers for Medicare & Medicaid Services (CMS) recently required Medicare Advantage plans to consider individual patient needs in coverage decisions rather than rely solely on generic criteria. Yet, insurers can still set their own standards, and there’s no requirement for independent testing before deployment.

Federal rules only apply to public health programs like Medicare and Medicaid. Private insurers outside these programs remain unregulated by these standards.

Several states—such as Colorado, Georgia, Florida, Maine, and Texas—have proposed or passed laws to regulate insurance AI. For example, a 2024 California law mandates licensed physician oversight of coverage algorithms. However, many state laws still leave too much discretion to insurers regarding definitions of “medical necessity” and algorithm use. They also lack requirements for third-party algorithm review.

Even strong state laws have limitations since states can't regulate Medicare or insurers operating across multiple states.

The Case for FDA Involvement

Given the gap between insurer practices and patient needs, many experts argue that regulating health insurance algorithms is urgent. The FDA is well-positioned to evaluate these algorithms because it already oversees many medical AI tools for safety and effectiveness.

FDA oversight would create a unified national standard instead of a patchwork of state rules. However, current FDA authority only covers medical devices "intended for use in diagnosis, cure, mitigation, treatment, or prevention of disease." Since insurance algorithms don't diagnose or treat, Congress might need to amend definitions to grant the FDA regulatory power over these tools.

Meanwhile, CMS and states could require independent testing for safety, accuracy, and fairness. This might encourage insurers to support a single national standard rather than face inconsistent regulations.

Insurance coverage decisions directly impact patient health. Strengthening oversight of AI algorithms in this space is essential to ensure fair access to care.

For those interested in AI applications and regulations in healthcare, exploring specialized training can provide valuable insights and skills. Check out Complete AI Training's courses by job role for options tailored to healthcare and insurance professionals.