AI Coding Tools Added $942 Million to Hospital Bills Without Sicker Patients, Blue Cross Analysis Finds

If you've opened a hospital bill in the last few years and felt a jolt of sticker shock, there may be a reason beyond inflation or your deductible resetting. A new claims analysis from the Blue Cross Blue Shield Association suggests that artificial intelligence tools now widely used by hospitals to code patient charges are quietly inflating the severity of bills, even when the actual care delivered hasn't changed at all.

Think of medical coding as translation work. A nurse or doctor writes down what happened during a hospital stay in plain clinical language, and a coder converts that into the standardized billing codes insurers use to decide how much to pay. For decades, that translation was done mostly by hand, case by case, by trained human coders. Now AI software increasingly does the first pass, scanning charts and suggesting codes, including secondary diagnoses that can bump a claim into a higher-paying category.

BCBSA's data shows that shift adding up fast. The share of inpatient cases its member plans classify as medically complex climbed from 37% at the start of 2023 to 40% by the end of 2025. That three-point jump might sound small. It isn't. About 70% of that increase in "coding intensity" traces back to more than 55,000 additional cases where a secondary diagnosis pushed a claim into a higher-severity, higher-reimbursement category known as a diagnosis-related group, or DRG.

Translate that into dollars, and the number gets harder to ignore: $942 million in additional costs shouldered by BCBSA's member plans over two years. Of that, $653 million came specifically from those secondary-diagnosis upgrades, averaging roughly $11,000 per excess complex case.

What the Data Actually Shows

The most striking part of BCBSA's findings isn't the price tag. It's what didn't change alongside it.

"Critically, what we found is underneath all of that data was no change in corresponding care for a more complex patient," Luke Chalker, BCBSA's senior vice president of product and data science, told reporters during a briefing on the findings. Coding got more intense, in other words, but the treatment patients actually received stayed roughly the same.

To test that claim, BCBSA researchers looked at hospitals in the top quartile for billing complex DRG cases and compared their actual treatment intensity to peers. That meant tracking things like ICU admissions, blood transfusions, reoperations and how long patients stayed. In 65% of completed DRG cases, these high-complexity-coding hospitals showed similar or even lower treatment intensity than hospitals that coded less aggressively.

One specific example makes the pattern concrete. Within major bowel procedures, claims coded at the highest complexity level rose from 10.2% to 22.7% of cases, while straightforward, non-complex claims fell from 36.6% to 32.8%. That single shift accounted for nearly $61 million of the added costs BCBSA identified.

The report also flags posthemorrhagic anemia, a diagnosis researchers describe as a common "bump code," meaning one that's frequently added to push a claim into a higher-paying category. Hospitals in the top quartile for diagnosing this condition did so in 13.7% of cases, compared to 9.9% elsewhere. Yet those same hospitals had lower transfusion rates among diagnosed patients, 16.9% versus 19.3%. If more patients were genuinely developing this complication, you'd expect more transfusions to treat it, not fewer.

"The consistent inverse relationship between diagnosis-based complexity and both aggregate resource utilization and diagnosis-specific procedural intervention is the strongest indicator that coding escalation reflects documentation practice changes rather than actual patient acuity shifts," the white paper states, in language that essentially says the numbers point to billing behavior, not sicker patients.

Razia Hashmi, BCBSA's vice president of clinical affairs, put it more plainly. "The question that is worth asking is, with two similarly situated hospitals treating similar patients, why would one hospital diverge?" she said. "There may be an element of correct coding there, but the likelihood that this is technology-enabled upcoding is higher, in my view."

BCBSA is careful to note limits here. The analysis relies on claims data rather than the underlying clinical charts, which would offer a more direct window into whether patients really were sicker. Chalker acknowledged that gap but said internal work with Blue plans that do have access to clinical documentation has "re-emphasized and demonstrated this effect" in similar terms.

Hospitals, for their part, describe their embrace of AI coding tools differently: as a defense mechanism. Payers have grown increasingly aggressive about denying or delaying claims, and hospital leaders have said automated coding helps them recover reimbursement they're rightfully owed while easing administrative strain on overworked staff. It's a framing Chalker pushed back on directly, arguing that a rise in cost without a matching rise in value delivered to patients isn't a correction. It's a distortion.

"We should be reimbursing for care delivered, that is more critical than anything else because that's supposed to be the design of how it all works," Chalker said. "Because of this mechanism, we've started to deviate from that overall."

Why It Matters

This isn't an abstract accounting dispute buried in a white paper. Higher coding intensity without more actual care translates, eventually, into higher premiums, higher out-of-pocket costs and higher taxes to fund public insurance programs. BCBSA's leaders were explicit that they see this as unnecessary spending passed along to employers, taxpayers and everyday enrollees, not a legitimate correction of past underpayment.

It also lands amid broader cost pressure across the system. Health plans have already flagged AI-driven coding as one factor behind projected 2027 premium increases of roughly 9%, alongside rising drug prices. And it sets up what BCBSA itself describes as an AI arms race: payers deploying algorithms to deny claims faster, hospitals deploying algorithms to code them more aggressively, with patients and premium payers caught in the middle of a fight neither side is fully winning. BCBSA says this analysis, focused on inpatient hospital stays, is just the first installment. More scrutiny of outpatient care and other DRG categories is coming, meaning this debate over what AI is actually optimizing for in healthcare billing is far from settled.