AI Algorithms Shape Health Insurance Decisions, Impacting Care

Health insurance companies have increasingly adopted artificial intelligence (AI) algorithms over the past decade to determine coverage for medical treatments. Unlike healthcare providers that utilize AI for diagnosis and patient care, insurers employ these systems primarily to make decisions regarding payment for services recommended by physicians. One prevalent application of this technology is in the prior authorization process, where doctors must obtain approval from insurers before proceeding with specific treatments.

AI algorithms assess whether a treatment is deemed “medically necessary,” influencing not only approval but also the duration of coverage, such as the number of days a patient can spend in the hospital after surgery. The implications of these decisions can be significant for patients facing serious medical conditions. If an insurer denies coverage for a recommended treatment, patients typically have three options: appeal the decision, which can be a lengthy and costly process; accept an alternative treatment; or pay for the recommended care out of pocket. The latter option is often financially unfeasible due to the high costs of healthcare services.

Concerns about the impact of these algorithms on patient health are rising. While insurers claim that AI can streamline decision-making and reduce unnecessary treatments, evidence suggests that the opposite may occur. There are instances where AI systems delay or deny essential care, prioritizing cost savings over patient well-being.

Patterns of Care Denial

Insurers input patients’ medical records and relevant data into their algorithms, which then compare this information against established medical standards to determine coverage eligibility. However, the lack of transparency around these algorithms raises questions about their operation and fairness. Insurers have not disclosed specific details about how decisions are made, leaving patients in the dark about the criteria used to evaluate their claims.

While employing AI for coverage decisions may reduce operational costs, it can also lead to detrimental outcomes. For instance, if a claim is denied and a patient opts to appeal, the process can extend for years. In urgent cases, such as terminal illnesses, the delay can lead to tragic financial savings for insurers if a patient does not survive the appeals process.

Research indicates that chronic illness patients, as well as marginalized groups including Black and Hispanic individuals and those identifying as LGBTQ+, are disproportionately affected by claim denials. Moreover, some studies suggest that prior authorization may inadvertently increase overall healthcare costs rather than decrease them.

Insurers argue that patients are not denied care outright, as they can always opt to pay for treatments themselves. This perspective overlooks the harsh reality that many patients cannot afford necessary care. The consequences of these decisions can lead to serious health ramifications.

Regulatory Landscape and Future Directions

Unlike medical algorithms used in clinical settings, AI tools employed by insurers remain largely unregulated. They do not undergo the same rigorous review as medical devices by the Food and Drug Administration (FDA), and insurers often classify their algorithms as trade secrets. This lack of oversight means there is insufficient public knowledge about the decision-making processes these algorithms employ, as well as no independent evaluation of their safety, fairness, or effectiveness.

Recently, the Centers for Medicare & Medicaid Services (CMS) announced that insurers in Medicare Advantage plans must consider individual patient needs rather than relying solely on generic criteria. Nevertheless, these regulations still permit insurers to establish their own decision-making standards without external validation of their algorithms.

Several states, including Colorado, Georgia, Florida, Maine, and Texas, have proposed legislation aimed at regulating the use of AI in insurance decisions, with California passing a law in 2024 requiring physician oversight of coverage algorithms. Despite these efforts, most state laws share significant limitations, allowing insurers to define “medical necessity” and determine how algorithms are applied without independent review.

Many health law experts advocate for the necessity of regulatory oversight for health insurance algorithms. In a forthcoming essay in the *Indiana Law Journal*, it is argued that the FDA is in a prime position to evaluate these algorithms prior to their implementation in coverage decisions. The agency already oversees various medical AI tools, ensuring their safety and effectiveness.

Although some contend that the FDA’s authority is restricted due to the definitions surrounding medical devices, there is potential for Congress to amend these definitions to encompass insurance algorithms. In the meantime, CMS and state governments could implement requirements for independent testing to assess safety, effectiveness, and fairness of these algorithms.

The movement toward regulating AI’s role in health insurance coverage decisions is gaining traction, yet it requires a stronger push. The health and well-being of countless patients depend on ensuring that these algorithms are used fairly and transparently.