New research from the Blue Cross Blue Shield Association is offering what the organization describes as the first publicly available data linking hospitals’ use of AI to rising healthcare costs, adding a concrete data point to a debate that has largely played out in insurers’ earnings calls and anecdotal claims.
The research, conducted by BCBSA’s data analytics arm, Blue Health Intelligence, looked at commercial inpatient claims from tens of thousands of maternity admissions nationwide, covering a three-year window ending in March 2025.
The study focused on acute posthemorrhagic anemia, a condition that typically warrants clinical intervention, including in some cases blood transfusions. The researchers found that diagnoses of the condition surged at a subset of hospitals, but the increase in coding was not matched by a corresponding rise in treatment. Among the top 10% of hospitals by growth in case complexity, the proportion of maternity patients with that diagnosis climbed from roughly 4% in mid-2022 to more than 12% by early 2025. The increase was concentrated at large health systems.
The cost impact, according to BCBSA, has been severe for insurers and employers. Increased coding of acute posthemorrhagic anemia added $22 million to maternity-related hospital costs in one year across the facilities studied. Projected nationally, the researchers put potential excess inpatient spending tied to AI-enabled coding at roughly $663 million, with outpatient exposure of at least $1.67 billion.
“Within the participating BCBS plan’s commercial inpatient population in this analysis alone, per member costs increased by 9% between 2023 and 2024,” the study said. “Coding intensity is estimated to contribute about 20% of this increase.”
Luke Chalker, chief product officer at BHI, told Becker’s that the Blues are uniquely positioned to surface these patterns.
“We have this extraordinarily rich data set that is effectively complete for 60+ million Americans across every geography,” he told Becker’s. “Leveraging that has given us the ability to be able to tap into these fingerprints faster than most.”
The new research arrives against a backdrop of widening tension between payers and providers over AI-assisted billing and documentation.
Since mid-2025, major insurers including UnitedHealth Group, Elevance Health, Cigna and Centene have been flagging what they describe as “aggressive provider coding” in their quarterly earnings calls.
Some revenue cycle executives and physicians have even described the dynamic as an “arms race,” with both sides deploying AI tools to manage and improve their margins.
Hospital leaders have pushed back, arguing that higher-acuity coding reflects more specific documentation and a sicker patient population, not coding manipulation.
“The idea that hospitals are coding aggressively to drive up costs is misleading,” Robert Boos, former vice president and chief revenue officer at Lynchburg, Va.-based Centra Health, told Becker’s in October. “Coding is governed by strict federal and industry standards, and health systems invest heavily in compliance, training and auditing to ensure accuracy.”
The BCBS study also examined hospitals that have publicly announced adoption of AI coding tools. At one such hospital, overall case complexity climbed 6.7% following that disclosure, while other hospitals in the same state saw complexity grow less than 1% over the same timeframe.
“When coding changes and it drives a reimbursement change, both a payer is paying more and a member is paying more, from an affordability perspective, through out of pocket maximums and through their deductibles,” Mr. Chalker said. “Our perspective is we want to create transparency beyond the rates and beyond the fact that our population is getting older. Let’s break down the trend contribution to this affordability issue.”
