Invest1 publisher3 min readPublished
Blue Cross ties $942 million in extra plan costs to hospitals' AI-assisted coding
Blue Cross Blue Shield Association estimates hospitals' AI-assisted coding contributed $942 million in extra costs to its plans from 2023 to 2025. Hospitals and insurers, both now running AI on claims, dispute whether that paid for overbilling or for real diagnoses recorded at last.
The Investor · Invest desk
Drafted by a language model from the sources cited here and checked against its claim ledger before publication. How we use AISend a correction
What happened
- Roughly 70% of the total, or $653 million, came from added diagnoses that brought no change in care, BCBSA's Luke Chalker told CNBC.
- BCBSA said much of the increase came from secondary diagnoses that moved patients into higher-paying reimbursement categories.
- The American Hospital Association hit back, accusing insurers of relying on automated downcoding and denials that can impede coverage of medically necessary care.
- Marsh forecasts that the cost per employee of health coverage will rise 8.2% on average in 2027, the largest increase since 2003.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- cost If BCBSA's estimate holds, members pay the bill through premiums and out-of-pocket costs, Chalker said, on top of coverage costs employers already expect to climb.
- contradiction The only dollar figure on record comes from the payer, and the hospitals' lobby says it lacks the context to judge quality, access or spending, so AI's net effect on costs is still unsettled.
- precedent With AI now on both the coding side and the review side, each new hospital tool invites an insurer counter-tool, and Whaley says neither side's spending goes to patient care.
Take the $653 million [3] out of the $942 million [1] and about $289 million is left [1]. That remainder is the only part of the estimate BCBSA did not tie to diagnoses that left treatment unchanged. Most of the money sits in the category the report attacks directly. "There is a clear disconnect between coding and treatment," it stated [17].
The case against AI specifically rests on two points. The growth in what BCBSA calls complex coding came during a period when 60% of hospital systems began using AI coding tools [4]. The association also said the secondary diagnoses involved "may be derived from single laboratory values, making it particularly well suited for detection by AI tools." [5] CNBC's account does not say whether BCBSA compared hospitals that adopted the tools with those that did not. Luke Chalker, BCBSA's senior vice president of product and data science, hedged to match. "While multiple factors contribute to coding intensity, the findings suggest AI-enabled coding and documentation tools are playing a role," he said [6].
Christopher Whaley, a Brown University health economist who studies hospital coding, offers the generous reading. "In many cases, the diagnoses are legitimate and weren't captured," he said [7]. On that view, hospitals are now billing for conditions patients had all along, and the plans were paying too little before. The American Hospital Association adds patient mix. "Patients today are older and more clinically complex," a spokesperson said in a statement to CNBC [8]. BCBSA is pressing Whaley's other category: conditions that "clinically just don't really matter and don't influence the patient's care," yet still allow another billing code and a higher payment [12].
I think the $653 million is the harder figure for hospitals to defend. An older, sicker population should show up as more care, and that slice came with no change in care [3]. The counter-thesis is serious. A real diagnosis that was missed and is now recorded need not change a treatment plan either, so an unchanged plan does not prove a code is padding. The view is wrong if hospitals without AI tools show similar coding growth over the same years, or if the added diagnoses go on to predict later treatment.
The spending runs on both sides of the claim. Chalker said Blue Cross Blue Shield companies also use AI in claims review, and that "any clinical denial is always reviewed by a qualified human clinician." [10] Whaley called the result an "administrative arms race." [18] "Whether it's on the hospital side or the insurer side, these tools and technologies are both very expensive," he said, "and also have nothing to do with providing appropriate care to patients." [13] He said those costs reach consumers through higher premiums and taxes [14].
What to watch
- Whether BCBSA publishes a comparison of coding growth at hospitals with and without AI coding tools over 2023 to 2025.
- Whether Blue Cross Blue Shield plans cite coding intensity when they justify 2027 premium increases.
- Whether insurers expand automated downcoding in response, the practice the AHA already objects to.