When Algorithms Decide Who Gets Health Care

Key Takeaways

• AI-driven coverage algorithms lack FDA safety review, unlike clinical tools.
• Nearly 20% of insured Americans face denied physician-recommended claims.
• 82% of doctors report delays cause patients to abandon treatment.
• Medicare Advantage plans rely heavily on AI prior authorizations, increasing denials.
• CMS rules require clinician oversight and public disclosure, but leave many gaps.

Pulse Analysis

Artificial intelligence has transformed clinical decision‑making, but its migration into insurance coverage decisions has outpaced regulation. Unlike diagnostic tools that must clear FDA safety and efficacy hurdles, "coverage algorithms" are deployed by insurers as proprietary black boxes. This regulatory vacuum allows insurers to automate prior authorizations, fail‑first protocols, and length‑of‑stay limits without external validation, raising concerns that algorithmic bias or design flaws could silently deny necessary care.

The real‑world impact is stark. Nearly one in five insured Americans reports a denied claim for a physician‑recommended service, and a 2024 physician survey found 82% say delays cause patients to abandon treatment altogether. These postponements often trigger higher downstream utilization—more office visits, emergency department trips, and avoidable admissions. Medicare Advantage enrollees are especially vulnerable, as private plans increasingly rely on AI to enforce prior authorizations, while traditional Medicare remains largely exempt. The cumulative effect is a cycle where algorithmic denials inflate overall health‑care spending and erode patient trust.

Regulators have begun to respond. CMS’s 2023 rule for Medicare Advantage requires a qualified clinician to review algorithmic decisions and obliges insurers to disclose the criteria used. A 2024 rule adds faster electronic exchanges and clearer denial notices, yet both leave critical questions unanswered: the degree of human override, the transparency of model inputs, and the scope of services subject to prior authorization. Oliva argues that extending FDA medical‑device oversight—or amending the FD&C Act—to cover these gatekeeping tools would impose the rigorous safety testing already standard for clinical AI, protecting patients while providing clearer compliance pathways for insurers.