Hospitals say they’re losing the AI billing war. A new BCBS study suggests otherwise

 The Blue Cross Blue Shield Association has put out a second round of research estimating that hospitals’ growing use of AI-assisted billing tools has added roughly $942 million in excess costs to the BCBS system over two years, contributing new data to the contentious “AI bot war” that’s been playing out across the industry.

The analysis, covering the first quarter of 2023 through the fourth quarter of 2025, found that hospitals are increasingly billing inpatient stays as more medically complex than they were a few years ago, not because patients are having more surgeries and procedures, but because additional diagnoses are being attached to claims that push them into higher-reimbursement billing tiers. The share of inpatient cases billed as complex rose from roughly 37% in early 2023 to about 40% by late 2025. The association estimates that more than 55,000 excess complex cases were coded during that window, generating $653 million in reimbursements at an average of $11,800 per case.

“The disconnect between diagnoses and treatment suggests that AI is identifying more billable conditions, not sicker patients,” Luke Chalker, BCBSA’s senior vice president of product and data science, said in a news release.

In March, BCBSA published first-of-its-kind research focused on posthemorrhagic anemia in maternity cases, finding that diagnoses of the condition surged at a subset of hospitals without a corresponding rise in treatment. That analysis projected excess inpatient spending tied to AI-enabled coding at roughly $663 million nationally, with outpatient exposure of at least $1.67 billion. 

The latest study focused on major bowel procedures, a surgical category used primarily for colon cancer and diverticular disease. Within the examined period, the most complex claims in that group rose from 20.2% to 22.7%, while non-complex cases fell from 36.6% to 32.8%, accounting for $60.8 million in added costs.

The researchers focused on what they called clinical discordance, or the gap between what hospitals are coding and what they appear to be treating. The fastest-growing secondary diagnoses in the bowel procedure category were conditions like acidosis, low sodium and posthemorrhagic anemia, the kind of conditions that can show up on a routine lab report and get picked up automatically by billing software scanning for anything codable. Hospitals in the top quartile of complexity growth classified 76% of bowel procedures as complex, compared to 65% at other hospitals, even though those same hospitals were not delivering more care. Their ICU use, transfusion rates and length of stay were flat or even lower than their peers.

The anemia example was the sharpest, according to the study. Top-quartile hospitals diagnosed anemia at a rate 38% higher than peers, but their transfusion rate among those patients was lower, sitting at 16.9% compared to 19.3% at other facilities.

“If it was worth coding, there should have been something done,” Razia Hashmi, MD, BCBSA’s vice president of clinical affairs, said during a September media briefing on the research. She noted that when two similarly situated hospitals are treating similar patients and the care patterns are the same but the coding diverges, “something else is changing.”

Health system leaders, meanwhile, have been describing what they say is a widening technology gap in which payers have automated claims review, downcoding and denial generation at a scale providers cannot match. 

HCA Healthcare CFO Mike Marks said Sept. 15 that hospitals are “behind the payers” on AI-driven claims processing and that the administrative cost on both sides “is enormous.” Executives at Becker’s 11th Annual IT + Revenue Cycle Conference this month described the dynamic in even starker terms, with NYC Health + Hospitals’ Chief Revenue Officer, Marji Karlin, calling it a “rock ’em, sock ’em robot” fight in which “nobody’s going to win.”

On the other side, major insurers beyond the Blues plans have been building the public case for months. UnitedHealthcare CEO Tim Noel said in June that AI-driven revenue cycle tools are “a source and driver of trend,” though he predicted the pressure would eventually reach saturation. Aetna’s Chief Medical Officer, Ben Kornitzer, MD, said on a Becker’s podcast the following month that early data on AI’s impact in healthcare has been “largely inflationary,” with coding intensity rising but “no real strong evidence that people are getting different clinical outcomes.”

Dr. Kornitzer encouraged the industry to resist framing the situation as “an agentic bot war” between payers and providers and instead redirect the technology toward improving clinical care.

Mr. Chalker at BCBSA said during the media briefing that more research will be conducted on the issue, including in the outpatient space, because the trend “hasn’t stopped” and is not limited to any single surgical category. He noted that the roughly $942 million figure represents only what the association can tie to a divergence between coding and care, not total complexity growth.

“Coding has changed. That is a fact,” Mr. Chalker said. “And that has changed reimbursement, and that has changed and affected affordability.”

https://www.beckershospitalreview.com/payer/hospitals-say-theyre-losing-the-ai-billing-war-a-new-bcbs-study-suggests-otherwise/