Six Fronts, One War: How AI Turned Healthcare Billing Into an Arms Race

This report documents a healthcare billing conflict that most public coverage treats as a single dispute and that this investigation shows is a six front war, fought with the same class of artificial intelligence, deployed by different parties, pointed in different directions, against a population that can generally see only one front at a time. On the first front, hospitals deploy coding intensity artificial intelligence to identify billable clinical complexity a human coder might miss, a pattern documented directly in the Blue Cross Blue Shield Association's own published analyses and disputed by the American Hospital Association. On a second and third front, insurers, UnitedHealth through its nH Predict algorithm and Cigna through its PxDx system, deploy denial artificial intelligence to identify deniable technicalities a human claims adjuster might miss, now the subject of active federal litigation in the District of Minnesota and the Eastern District of California respectively. On a fourth front, some of those same insurers stand accused by the United States Department of Health and Human Services Office of Inspector General of using artificial intelligence assisted chart review to add diagnosis codes to their own Medicare Advantage risk adjustment submissions, inflating what the federal government pays them rather than what they pay a provider. On a fifth front, the federal government has adopted a structurally similar technology and payment model of its own: the Centers for Medicare and Medicaid Services' Wasteful and Inappropriate Service Reduction model, extending prior authorization review into Traditional Medicare for the first time. A sixth front is structural rather than algorithmic: a single corporate parent can now employ the physician documenting a clinical encounter, insure the patient the encounter concerns, and separately face federal scrutiny for how its own subsidiaries code government payments, a conflict of interest with no dedicated oversight body. Underneath all six fronts sits a legal architecture, established by the United States Supreme Court in Pilot Life Insurance Co. v. Dedeaux and Aetna Health Inc. v. Davila, that limits the financial consequences of a wrongful algorithmic denial for the majority of insured Americans to the value of the benefit itself. This report also introduces an original methodological contribution: an application of the selective labels problem, a documented concern in computer science literature on algorithmic decision making, to healthcare claims adjudication for the first time, showing how a denial model retrained on its own historical outcomes may be structurally unable to learn from its own errors. This report extends Koru Research Institute's prior work on healthcare's infrastructure dependencies in Report No. 008, Healthcare Runs on Compute Now: How AI, Data Centers, Cloud Infrastructure, and Energy Are Reshaping Healthcare Delivery, and its analysis of artificial intelligence's geopolitical footprint in the Global Risk Brief, The Bottleneck Beneath the AI Boom: How a Global Energy Average Hides Where Power, Water, and Chips Run Out. Marcus J. Frazier, Author | Koru Health, a Koru Capital Management Company | Published by the Koru Research Institute | Korucm.com ORCID: 0009-0005-5094-9040