Blue Cross Study Links AI Hospital Coding to $942 Million in Costs

The Blue Cross Blue Shield Association said on September 24 that AI-driven hospital coding tools added nearly $942 million in extra healthcare costs for its member companies between 2023 and 2025, in an analysis that hospital industry representatives dispute.⁸⁶¹

Where the costs came from

According to the BCBSA analysis, about $653 million of the increase, roughly 70 percent, was associated with hospitals billing more frequently for secondary conditions that pushed patient stays into higher-paying diagnosis-related groups.^5,s4,s61,s2

According to the association, over six in ten hospital systems have adopted AI-based coding software—tools that comb through patient records or pair with ambient scribes recording clinician conversations and generating notes—to surface diagnoses eligible for billing.^1,s8

Fierce Healthcare reported that roughly seven-tenths of the rise in coding intensity stemmed from over 55,000 extra cases where secondary diagnoses shifted claims into more severe billing tiers, and that complex cases billed to Blue plan enrollees climbed from 37 percent at the start of 2023 to 40 percent by the end of 2025.^10,s2

Diagnoses without matching treatment

The study found that, across the span from early 2023 through late 2025, diagnoses of partial intestinal blockages climbed 55 percent and diagnoses of excess bodily acid climbed 33 percent, yet treatment rates for bowel-disease surgeries did not rise in tandem.^4,s3

Anemia diagnoses also increased while blood transfusions among surgery patients stayed flat, according to the analysis, and Medical Daily reported that at hospitals most likely to diagnose anemia from blood loss, patients received transfusions less often than at other hospitals.¹²

Luke Chalker, senior vice president at BCBSA, said: "If patients are truly sicker, we'd expect to see more treatment." He added: "The disconnect between diagnoses and treatment suggests that AI is identifying more billable conditions, not sicker patients."¹

Dr. Razia Hashmi, the association's vice president of clinical affairs, said "more complex health cases did not correspond with higher rates of treatment for people with bowel disease surgeries," and separately said, "There may be an element of correct coding there, but the likelihood that this is technology-enabled upcoding is higher, in my view."⁶²

Industry response and prior estimates

Centene said health systems' reliance on AI software has driven overly assertive or improper reimbursement claims, and its chief executive, Sarah London, pointed to instances of overcoding, such as patients with simple fevers being labeled with sepsis.^1,s69

Luke Chalker described the relationship between insurers and providers not as a fair fight but as a lopsided rout in which insurers are consistently on the losing side.⁷

The American Hospital Association disputed the insurer analysis, arguing patients are genuinely sicker, and noted that hospital case-mix severity increased about 5 percent from 2019 through 2024.²³

A prior Blue Cross review of commercial claims conducted in March had calculated that around $663 million in inpatient costs and no less than $1.67 billion in outpatient costs nationally might trace back to AI-assisted coding, with part of that tied to anemia diagnoses following childbirth.³²

CMS Administrator Dr. Mehmet Oz described AI as a driver of rising billing costs, while the Medicare Payment Advisory Commission determined that upcoding was responsible for $40 billion worth of Medicare Advantage overpayments in 2025.⁹²

Caroline Pearson, who leads the Peterson Health Technology Institute as executive director, recounted a roundtable discussion in which investors, health plans, and providers all concurred that scribes are driving up coding intensity, citing participants who said: "The investors, the health plans, and the providers, in private, were like, 'OK, well, it's quite clear scribes are increasing coding intensity. One hundred percent.'"¹

Shiv Rao, founder of the AI startup Abridge, acknowledged risks of AI automation in healthcare while suggesting potential benefits, describing a possible "horrible dystopic future nobody wants to live in, with bots fighting bots, agents fighting agents."⁷

In a related note, UnitedHealth Group has indicated that AI might cut its costs by close to $1 billion in 2026, while HCA Healthcare anticipates roughly $400 million in AI-linked savings that same year, per the report, which also estimated total US healthcare AI expenditure at $1.4 billion in 2025.³