Hospitals’ increasing use of artificial intelligence for patient coding has increased the frequency of inpatient stays classified as medically complex—boosting the bills sent to payers despite no apparent changes in the care being delivered, according to a new claims analysis from the Blue Cross Blue Shield Association.
The association found that the share of medically complex cases billed to its Blue plan members has risen from 37% at the beginning of 2023 to 40% by the end of 2025.
About 70% of the coding intensity increase stems from more than 55,000 more cases from the 2023 baseline in which secondary diagnoses pushed a claim into a higher-severity, higher-reimbursement diagnosis-related group (DRG).
The increase in coding intensity since baseline translates to an estimated $942 million of additional costs shouldered by BCBSA’s member plans over two years, of which $653 million stemmed from secondary diagnoses ($11,000 per excess complex case).
“Critically, what we found is underneath all of that data [was] no change in corresponding care for a more complex patient,” Luke Chalker, senior vice president of product and data science at BCBSA and one of the analyses’ authors, told reporters during a briefing on the findings. “We now see that coding has materially changed. We see that. Non-Blues see that—they talk about it sometimes in earnings reports and things like that. But we find no evidence of a corresponding change in care, and that’s because the reimbursement mechanisms that exist allow this.”
Chalker and other BCBSA leads who spoke to reporters described the increases—particularly those with no accompanying changes in care delivery—as unnecessary healthcare spending for plans that then leads to higher premiums and out-of-pocket costs for enrollees, employers and taxpayers.
The $942 million estimate “is purely for [where] we believe there is no change in care delivered,” Chalker said, and excludes anything that led to a hospital documenting that it delivered additional care.
“That’s the stuff … that’s a no-brainer from my perspective,” he said. “That’s the stuff that hospitals should bill for, and that’s the stuff we should pay for.”
The coding intensity changes since 2023, BCBSA said, stem from “systemic” adoption of AI revenue cycle management tools, which hospitals and health systems have said are helping them more accurately and efficiently code claims amid increased scrutiny from payers. BCBSA, in an accompanying release, referenced a June survey in which more than 63% of healthcare organizations reported using AI in their revenue cycle workflows.
Thursday’s analysis included a snapshot of one specific DRG where the increased coding complexity is apparent. Within major bowel procedures, for instance, claims at the highest level of complexity rose from 10.2% to 22.7% while non-complex cases dipped from 36.6% to 32.8%, together accounting for nearly $61 million of the analysis’ incremental claims costs.
And to affirm its thesis of “clinical discordance” between increased diagnoses and delivered treatments, BCBSA looked at the top quartile of hospitals with complex DRG cases. These hospitals had similar or lower treatment intensity (ICU utilization, transfusion, reoperation and median length of stay) than their peers, with 65% of completed DRG cases, per the white paper.
Additionally, BCBSA’s report highlighted posthemorrhagic anemia (“a common bump code”), where the top quartile of hospitals for anemia diagnosis (13.7% versus 9.9% for the remainder) had lower rates of transfusion among those who were diagnosed (16.9% versus 19.3%).
“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,” BCBSA’s white paper reads.
“The question that is worth asking is [with] two similarly situated hospitals, treating similar patients, why would one hospital diverge?” Razia Hashmi, BCBSA’s vice president of clinical affairs, told reporters. “There may be an element of correct coding there, but the likelihood that this is technology-enabled upcoding is higher, in my view.”
BCBSA acknowledged that the analysis is limited due to its reliance on claims rather than clinical documentation, which would be a more direct measure of whether patients were substantially sicker than peers without the increased coding complexity and thus accurately billed.
Chalker, on this point, said he believed that a similar analysis with clinical data on hand would make a stronger case for BCBSA’s position. “And at least in the work that we've done with Blue Plans and who do have more access [to charts] because of provider relationships with that clinical data, they've been able to kind of re-emphasize and demonstrate this effect,” he added.
Thursday’s white paper specifically focuses on hospital inpatient services, with Chalker adding that BCBSA plans to continue releasing analyses related to other areas like outpatient care or other noteworthy DRGs.
As financial headwinds have closed in on every aspect of the healthcare system, payers and providers, respectively seeking to limit expenses and boost revenue, have found themselves in an AI arms race with regard to claims submissions and approvals.
Hospital leaders have said their increased adoption of the technology is in part an answer to payers’ “frustrating” willingness to deny or stall reimbursements, as well as a means to reduce administrative burden and spending.
Chalker pushed back on the framing of AI tools securing long-deserved payments for hospitals, saying that a demonstrable rise in expenses with no corresponding increase in value for downstream stakeholders is a change for the worse.
“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. And because of this mechanism, we’ve started to deviate from that overall," he said.