How AI Algorithms Decide Your Health Insurance Coverage—And Why Patients Are at Risk

The Impact of AI Algorithms on Health Insurance Decisions

Health insurance companies have increasingly adopted artificial intelligence (AI) algorithms over the past decade. Unlike medical providers who use AI to diagnose and treat patients, insurers apply these algorithms to determine whether to approve payment for treatments recommended by a patient’s doctor.

A common example is prior authorization, where doctors must get approval from insurers before delivering certain care. AI systems evaluate if the requested treatment is "medically necessary" and decide whether coverage should be granted. These algorithms also influence the extent of coverage, such as how many hospital days a patient can receive following surgery.

If an insurer denies coverage for a recommended treatment, patients typically face three options:

• Appeal the decision, which can be time-consuming, costly, and often requires expert support. In fact, only about 1 in 500 denials are appealed.
• Accept an alternative treatment that the insurer will cover.
• Pay out-of-pocket for the recommended care, which is often unrealistic due to high costs.

A Pattern of Withholding Care

Insurers presumably feed patient health records and other relevant data into AI algorithms to compare against medical standards for coverage decisions. However, these companies refuse to disclose how their algorithms work, making it impossible to evaluate their operation or fairness.

AI review saves insurers time and reduces the need for medical professionals to assess each case manually. The financial benefits extend beyond efficiency: if a valid claim is denied quickly and the patient appeals, the process can stretch on for years. For patients with serious or terminal conditions, insurers might financially benefit if the appeal delays care until the patient’s condition worsens or leads to death.

This raises concerns that algorithms could be used to withhold care for costly, long-term illnesses or disabilities. Research shows that patients with chronic illnesses face higher denial rates and suffer worse outcomes. Furthermore, claims denials disproportionately affect Black, Hispanic, and other nonwhite groups, as well as LGBTQ+ individuals.

Contrary to insurers' claims that patients can always pay for denied treatments themselves, the reality is that many cannot afford expensive care. These denials carry serious consequences for patient health.

Moving Toward Regulation

Unlike medical AI tools, insurance algorithms operate with little oversight. They do not undergo Food and Drug Administration (FDA) review, and insurers often claim their algorithms are proprietary trade secrets. This lack of transparency means there is no public information on how decisions are made or independent verification of their safety, fairness, or effectiveness.

Some regulatory progress is underway. The Centers for Medicare & Medicaid Services (CMS) recently mandated that Medicare Advantage insurers must consider individual patient needs, not just generic criteria. However, this still allows insurers to set their own standards and does not require independent testing before deployment. Moreover, these federal rules apply only to public health programs, leaving private insurers largely unregulated in this area.

Several states, including Colorado, Georgia, Florida, Maine, and Texas, have proposed or enacted laws to limit insurance AI use. California’s 2024 law requires physician supervision of AI-driven coverage decisions. Yet, most state laws still leave significant control to insurers in defining "medical necessity" and lack mandatory third-party algorithm review. State regulations also cannot fully address insurers operating across state lines or federal programs like Medicare.

A Role for the FDA

Experts argue that regulating insurance coverage algorithms is essential to protect patients. The FDA is well suited for this role, staffed with medical professionals capable of evaluating AI tools for safety and effectiveness. The agency already oversees many medical AI devices, and extending this oversight to insurance algorithms would create a consistent national standard instead of fragmented state rules.

One challenge is that current FDA regulations define medical devices as tools for diagnosing, treating, or preventing disease. Since insurance algorithms do not directly treat patients, Congress might need to update this definition to enable FDA oversight of these tools. Legislative changes could grant the FDA clear authority to regulate insurance AI algorithms.

Meanwhile, CMS and state governments could require independent testing of these algorithms for accuracy, fairness, and safety. This could encourage insurers to support a unified national standard rather than face inconsistent regulations across states.

The push for regulating AI in health insurance decisions has begun but requires stronger momentum. The stakes are high—patient health and access to necessary care depend on it.

To explore how AI is transforming the insurance industry and develop skills for managing AI tools effectively, visit Complete AI Training’s courses for insurance professionals.