The software finds extra diagnoses in patients' records that move hospital bills into pricier tiers, the insurers say, with no matching rise in treatment.
The Blue Cross Blue Shield Association estimates that more hospital stays billed as medically complex added $942 million in spending for its member insurers from 2023 to 2025, a rise it ties to hospitals adopting AI coding tools.
The analysis, built on claims data from insurers that together cover one in three Americans, looked at major bowel surgery. The association found nothing to show that treatment changed along with the coding. It points to anemia: diagnoses rose sharply after these operations, but blood transfusions did not, although genuinely sicker patients would be expected to get more treatment.
Secondary diagnoses, conditions recorded alongside the main reason for a stay, made up about 70% of the added cost, roughly $650 million, the association estimates. Such a diagnosis can lift a stay into a better-paying tier. The association also says a single lab reading can be enough to support one, so AI tools can pick these diagnoses up especially easily.
By the association's count, more than 60% of hospital systems now use AI software that searches lab results and electronic records for such diagnoses. Providers also find extra conditions through ambient scribes, tools that listen in on patient visits and write up the doctor's notes.
The New York Times has reported that AI used by both hospitals and insurers appears to be deepening their long-running fight over bills. Centene and other insurers have complained that health systems' AI tools produce aggressive or improper billing.
Luke Chalker, a senior vice president at the association, calls the fight "a completely one-sided blood bath" that insurers are losing. Shiv Rao, founder of the AI startup Abridge, warns of "bots fighting bots," though he says AI might also ease tensions and cut costs.