A retrospective cohort analysis of recent Senate scrutiny suggests that algorithms used by Medicare Advantage insurers to automatically deny coverage may be correlated with negative health events, though researchers emphasize that strict causation has yet to be established.
Sens. Richard Blumenthal and Josh Hawley have requested documentation from UnitedHealth, Humana, and CVS regarding their use of artificial intelligence to delay or reject medical claims, prompting industry experts to warn that abruptly halting the denials could contaminate an otherwise promising dataset.
The automated denial protocol, rapidly deployed across major Medicare Advantage plans, has been associated with a statistically significant reduction in authorized medical treatments. However, independent observers note that it is difficult to isolate the software as the primary risk factor for the self-reported worsening of enrollees' conditions, given the baseline morbidity of the aging population.
The bipartisan push to seize internal AI processing records represents a novel legislative intervention that could disrupt the natural progression of the algorithms. Health economists caution that subpoenaing the companies midway through the fiscal year could introduce confounding variables into the insurers' automated rejection models, artificially skewing the resulting quarterly profit margins.
While early data shows a strong association between a machine instantaneously rejecting a necessary surgery and the patient subsequently expiring, we must be careful to separate correlation from causation," said Dr. Martin Gable, Chair of Predictive Triage at the Mayo Clinic. "The cohort of individuals who were denied care were already quite ill to begin with, which is a classic methodological trap.
Representatives for Humana and CVS echoed the need for rigorous peer review, noting that the Senate's guidance relies heavily on qualitative complaints from patients. Such data historically suffers from extreme response bias, particularly among subjects who are currently deceased.
To address the evidentiary gap, the NIH is reportedly reviewing a proposal for a double-blind, randomized control trial in which half of all Medicare Advantage enrollees will have their claims evaluated by a human doctor, while the other half will be systematically denied by a server in Minnesota until long-term mortality endpoints can be confidently verified.