Artificial intelligence is supposed to make healthcare more efficient, but it may also be making it more expensive. A new analysis from the Blue Cross Blue Shield Association suggests AI-powered hospital coding tools are helping identify more billable conditions, pushing patients into more expensive categories even when the care they receive does not appear to change.
The numbers are hard to ignore. BCBSA estimates that increased coding complexity added $942 million in healthcare spending between 2023 and 2025 for Blue Cross and Blue Shield companies. About 70 percent of that increase, more than $650 million, was tied to secondary diagnoses that moved patients into higher-severity and higher-reimbursement categories.
This is where AI enters the picture. More than 60 percent of hospital systems are now using AI-enabled technology capable of scanning electronic health records, laboratory results, and doctors’ notes. These systems can spot secondary diagnoses that might otherwise be missed and turn information buried in a patient’s record into additional billing codes.
On its own, finding a previously overlooked medical condition is not necessarily a bad thing. Better documentation could provide a more complete picture of a patient’s health, and hospitals should obviously receive appropriate payment when they provide more complicated care. The problem comes when the complexity of the bill increases without a corresponding change in treatment.
BCBSA says that is exactly what its claims data shows. The share of inpatient cases classified as medically complex increased from 37 percent at the beginning of 2023 to 40 percent by the end of 2025, according to reporting on the analysis. Yet researchers say they found no corresponding increase in care that would explain the additional complexity.
Researchers examined major bowel procedures as one example. At the highest level of complexity, claims increased from 10.2 percent to 22.7 percent, while the share of non-complex cases declined. Those changes accounted for nearly $61 million in additional claims costs in that category alone.
Anemia provides an even easier way to understand what may be happening. BCBSA found hospitals increasingly documenting anemia as a secondary diagnosis after major bowel surgery, which can make a patient appear medically more complex for billing purposes. If those patients were actually getting sicker, researchers expected to see treatment change as well.
That did not happen. BCBSA says anemia diagnoses increased without a corresponding increase in blood transfusions, raising the possibility that AI is simply getting better at finding conditions that increase reimbursement rather than identifying patients who require additional care.
This isn’t the first time BCBSA has spotted the pattern. Earlier this year, it examined tens of thousands of maternity admissions and found increases in acute posthemorrhagic anemia diagnoses without the expected increase in treatments such as blood transfusions. That research estimated roughly $663 million in inpatient spending and at least $1.67 billion in outpatient spending may be connected to more aggressive coding practices.
There is an important limitation here. BCBSA’s latest analysis does not prove that AI caused every dollar of the estimated $942 million increase. The organization says the changes coincide with growing adoption of AI coding technology and argues that the patterns suggest these tools are contributing to higher spending. Hospitals, meanwhile, have reason to use such software to capture legitimate diagnoses and make sure the care they provide is documented accurately.
There is also an obvious perspective to consider. BCBSA represents insurers that pay hospital claims, so it has a financial interest in preventing unnecessary increases in reimbursement. Its findings are based on de-identified claims data from BCBS companies, which collectively cover about one in three Americans, but this is still an analysis coming from one side of the healthcare payment system.
None of that makes the underlying question less troubling. AI is extraordinarily good at searching through enormous amounts of information and finding details humans might overlook. Apply that capability to medical billing, where a newly identified secondary diagnosis can increase reimbursement, and suddenly AI has a financial incentive attached to everything it discovers.
AI could absolutely reduce paperwork, improve documentation, and give doctors more time with patients. But if the technology also becomes exceptionally good at making the same hospital stay look more expensive on paper, patients may end up paying for AI efficiency through higher premiums and out-of-pocket costs.
That would be a strange version of healthcare innovation: smarter software, bigger bills, and essentially the same care.
Support independent tech journalism
NERDS.xyz is independently owned and operated. If you enjoy my coverage of Linux, AI, hardware, cybersecurity, and tech culture, consider supporting the site on Ko-fi.
Support NERDS.xyz
