Artificial intelligence tools used to document and code hospital stays may be contributing to hundreds of millions of dollars in added health care costs, according to a new Blue Cross Blue Shield Association analysis that estimates changes in inpatient coding added $942 million in costs to Blue plans over two years.
The national association said September 24 that hospital claims increasingly classified patients as medically complex between 2023 and 2025, even though its claims data showed no corresponding increase in the intensity of treatment patients received.
BCBSA stopped short of proving that artificial intelligence caused each increase. Its analysis relies primarily on claims rather than complete patient charts and does not identify individual hospitals or AI vendors.
Still, the association says the timing and pattern raise concerns that increasingly common AI-enabled documentation and coding systems are finding additional diagnoses that move hospital stays into more expensive billing categories.
The American Hospital Association disputes that interpretation, arguing that hospitals are caring for older and sicker patients and that AI can help clinicians document legitimate medical complexity more accurately.
Complex Claims Increase
The share of inpatient cases classified as medically complex increased from roughly 37% at the beginning of 2023 to 40% by the end of 2025, according to the BCBSA analysis.
The association estimated that the change resulted in approximately $942 million in additional spending by Blue Cross and Blue Shield companies.
Roughly 70% (more than $650 million) was associated with secondary diagnoses that moved hospital stays into higher-paying diagnosis-related groups, according to the association.
More than 55,000 additional cases fell into those higher-complexity categories.
BCBSA said the concern is not simply that more diagnoses appeared in patients’ records, but that treatment patterns did not increase in tandem.
“If patients are truly sicker, we’d expect to see more treatment,” Luke Chalker, BCBSA senior vice president of product and data science, said in the association’s release.
Chalker pointed to anemia diagnoses as one example, saying hospitals recorded significantly more anemia without a corresponding increase in blood transfusions.
“The disconnect between diagnoses and treatment suggests that AI is identifying more billable conditions, not sicker patients,” he said.
Bowel Procedures Show Sharp Coding Shift
BCBSA examined major bowel procedures as one example of the trend.
Claims classified at the highest level of complexity increased from 10.2% to 22.7% between early 2023 and late 2025.
Hospitals in the top quartile for anemia diagnoses coded the condition in 13.7% of cases, compared with 9.9% at other hospitals. Among patients given that diagnosis, 16.9% received transfusions at the high-coding hospitals, compared with 19.3% elsewhere.
The association reported similar disconnects involving other measures of treatment intensity, including ICU utilization, reoperations and length of stay.
BCBSA says more than 60% of hospital systems now use AI-enabled technologies capable of examining clinical notes, laboratory results and other medical records for diagnoses that may affect coding.
Such systems can identify secondary medical conditions that human coders might previously have overlooked. Depending on the condition, an additional diagnosis can move an inpatient stay into a more highly reimbursed diagnosis-related group.
Hospitals Push Back
The American Hospital Association argues that higher coding intensity should not automatically be interpreted as improper “upcoding.”
The hospital group said patients receiving inpatient hospital care have become older and more medically complex while many less-complex procedures have shifted to outpatient settings.
An AHA and Vizient analysis found hospital case-mix index (a measure of patient complexity) increased roughly 5% between 2019 and 2024.
The AHA also said approximately 19% of hospital expense growth during that period reflected care for patients with greater medical complexity.
Hospitals have legal, ethical and contractual responsibilities to accurately document patients’ conditions, the group said, adding that human validation remains essential when AI is used in coding.
The organization has also accused commercial insurers of using automated systems in the opposite direction to “downcode” some claims and reduce reimbursement for care hospitals say was medically necessary.
The AHA cited a 2025 Medicare Payment Advisory Commission finding that coding practices among Medicare Advantage insurers contributed to roughly $40 billion in additional federal payments.
Study Has Limits
BCBSA acknowledges an important limitation: its analysis relies on insurance claims rather than complete clinical records.
That means the data cannot independently establish whether every additional diagnosis represented a legitimate condition that required no further treatment, a previously overlooked condition uncovered by better documentation, or inappropriate coding.
The analysis also does not identify specific Texas hospitals, individual health systems or particular AI vendors responsible for the changes.
Dr. Razia Hashmi, BCBSA vice president of clinical affairs, nevertheless said differences among otherwise similar hospitals deserve scrutiny.
“The question that is worth asking is [with] two similarly situated hospitals, treating similar patients, why would one hospital diverge?” Hashmi told reporters.
She acknowledged that some differences could reflect more accurate coding but said BCBSA believes technology-enabled upcoding is another possible explanation.
Texans Already Feeling Insurance Costs
The dispute comes as health insurance expenses and claim denials are already putting pressure on household budgets.
As The Dallas Express previously reported, Americans with employer-sponsored insurance paid an average of $120 per month for individual coverage and $571 per month for family coverage in 2025. Total family premiums, including employer and employee contributions, averaged $26,993 annually.
That DX report also found that insurers denied approximately one in five claims submitted to HealthCare.gov marketplace plans in 2024, including 19% of in-network claims and 37% of out-of-network claims.
Those figures do not mean every denial was improper. Claims can be rejected for reasons including excluded services, medical-necessity determinations, network issues, billing errors, missing information or failure to obtain required prior authorization.
Texas officials have also focused on insurance affordability.
The Dallas Express reported in July that Gov. Greg Abbott (R-Texas) proposed an Essential Benefits Health Insurance Plan that his office said could lower premiums by nearly 20%, or approximately $1,150 annually for a typical family, by allowing certain employer group plans to exclude some benefits required under state law while retaining federally required essential benefits.
AI On Both Sides Of The Bill
Blue Cross and Blue Shield of Texas, an independent BCBSA licensee, publicly acknowledges using AI in its own operations.
On its artificial intelligence information page, BCBSTX says AI can help process claims, automatically approve some requests and alert staff when something appears wrong.
The Texas insurer also says it does not use AI to replace physicians’ medical findings or deny care, stating that human judgment should continue to guide its use of the technology.
That distinction matters because some online summaries of the BCBSA study have portrayed the dispute as hospital AI generating larger claims while insurer AI automatically rejects those same claims. The September BCBSA analysis does not establish that scenario.
What it does document is a measurable rise in hospital coding complexity, an estimated $942 million in additional Blue plan spending and a dispute between insurers and hospitals over what that change represents.
BCBSA says the pattern suggests technology is increasingly identifying diagnoses that increase reimbursement without corresponding changes in treatment.
Hospitals say more complete documentation is capturing medical complexity that was already there.
As AI becomes increasingly embedded in both hospital and insurance operations, that disagreement is likely to become an increasingly important part of the broader fight over what Americans pay for health care.
