The Impact of AI Algorithms on Health Insurance Coverage
Over the last ten years, health insurance companies have increasingly relied on artificial intelligence (AI) algorithms to manage coverage decisions. Unlike healthcare providers who use AI to diagnose and treat patients, insurers use these algorithms to determine whether to approve payment for recommended treatments and services.
A common example is prior authorization. This process requires doctors to get approval from insurers before delivering certain care. AI algorithms help insurers quickly decide if the requested treatment is “medically necessary” and thus eligible for coverage. These systems also influence decisions such as the allowable length of hospital stays post-surgery.
If an insurer denies payment for a recommended treatment, patients typically face three choices:
- Appeal the decision, which can be time-consuming, costly, and complex—only 1 in 500 denials are appealed.
- Accept an alternative treatment covered by the insurer.
- Pay out-of-pocket for the recommended care, often an unrealistic option due to high costs.
A Pattern of Withholding Care
In theory, insurers feed patient records and relevant data into AI systems that compare this information against medical standards to decide coverage. However, insurers do not disclose how these algorithms operate, making it difficult to assess their fairness or accuracy.
Using AI to review claims saves insurers time and reduces the need for medical reviewers. But it may also serve financial interests by denying valid claims swiftly. If a patient appeals, the process can drag on for years. For seriously ill patients, insurers might benefit financially if the delay means the case never resolves due to the patient’s death.
This raises concerns that algorithms could be used to withhold care for costly, chronic, or terminal conditions.
Research shows patients with chronic illnesses are more frequently denied coverage, suffering as a result. Additionally, Black, Hispanic, and other nonwhite ethnic groups, as well as LGBTQ+ individuals, face higher denial rates. Some studies suggest prior authorization may even increase healthcare costs.
Insurers argue patients can pay for treatments themselves, but this overlooks the severe health risks when care is unaffordable.
Moving Toward Regulation
Unlike clinical AI tools, insurance AI algorithms face little regulation. They do not require FDA approval, and insurers often claim their algorithms are trade secrets. This lack of transparency means no independent verification exists to confirm their safety, fairness, or effectiveness.
There is some progress. The Centers for Medicare & Medicaid Services (CMS) now requires Medicare Advantage insurers to base decisions on individual patient needs rather than generic rules. However, insurers still control how they define “medical necessity” and must not prove their systems’ effectiveness before use. Also, these federal rules don’t cover private insurers outside federal programs.
A few states—including Colorado, Georgia, Florida, Maine, Texas, and California—have introduced laws to regulate insurance AI. For instance, California mandates licensed physicians supervise these algorithms. Still, most state laws allow insurers wide latitude in decision-making and lack requirements for third-party algorithm testing. Moreover, state laws cannot regulate Medicare or insurers operating across state lines.
A Role for the FDA
Many health policy experts argue that regulating insurance AI tools is essential to protect patients. The FDA is well positioned to oversee these algorithms since it already evaluates many medical AI products for safety and effectiveness. FDA regulation could provide a consistent national standard instead of fragmented state rules.
One obstacle is the current FDA definition of a medical device, which covers tools used to diagnose or treat disease. Insurance algorithms don’t fit this category, so congressional action may be needed to expand FDA authority.
Meanwhile, CMS and state regulators could require independent testing of insurance algorithms to ensure they are safe, accurate, and fair. This might also encourage insurers to support a unified regulatory framework.
The effort to regulate how AI influences health insurance coverage is underway but requires stronger momentum. The stakes are high—patient access to necessary care depends on it.
For healthcare professionals interested in the intersection of AI and healthcare policy, exploring current AI training courses can provide valuable insights into how AI tools are reshaping the industry.
Your membership also unlocks:
