Does AI increase healthcare costs? This Blue Cross report says yes

The Blue Cross Blue Shield Association (BCBSA) released a report on Sept. 24 claiming that hospitals are billing more conditions as “medically complex” than a few years ago. The proportion of cases that were billed as complex rose from “roughly 37% in early 2023 to about 40% by late 2025,” according to the study.

The report, which is based on an analysis of 55,000 excess complex cases, suggests that hospitals “are systematically adding secondary diagnoses that bump claims into higher-severity, higher-reimbursement diagnosis-related groups (DRGs) without a corresponding increase in clinical care.”

BCBSA blames AI, which hospitals are using more often to help input codes for patient care. This practice cost the Blues system—which has 118 million members across all 50 states, Washington, DC, and Puerto Rico—an extra $942 million in healthcare spending between 2023 and 2025. Much of the increase was driven by upcoding at teaching hospitals, the report found.

“The report raises a fair question, and nobody in the revenue cycle should defend a code that isn’t supported by the clinical record,” Missy Harbert, solution advisor at health tech Revecore, said in a statement. “Accuracy has to be the standard on both sides of the claim.”

The report though, she said, is based on claims data, not chart review.

“It can’t tell the difference between a diagnosis that was inappropriately added and one that was always present but previously missed,” she added.

Ben Teicher, a spokesperson for the American Hospital Association, agreed that it’s important to correctly assess a patient’s clinical status when delivering quality care.

But he told us that “patients today are older and more clinically complex…and new AI tools are helping providers appropriately capture their patients’ conditions to aid in care planning.”.

“Unfortunately, BCBSA’s analysis lacks the context needed to meaningfully assess how these tools impact healthcare quality, patient access, or spending,” he added.

But David Merritt, SVP of external affairs for BCBSA, said if patients were truly more medically complex, there would be “corresponding increases in treatment intensity and resource use.”

“The answer is not to slow innovation, but to establish the right guardrails: a flexible, risk-based approach to AI, appropriate human involvement, and accountability measures that ensure coding and reimbursement accurately reflect patient care,” he said in a statement.

This study adds to the ongoing debate over whether or not AI has legitimate ROI and whether it should be implemented in healthcare. Tools are being adapted quickly, with an estimated 75% of US health systems deploying at least one AI tool, according to data from research firm Eliciting Insights. Critics say AI may be saving time, but they also lack training, aren’t necessarily saving money, and may not be improving patient care.

If a hospital and an insurance company are both using AI to address prior authorization, for example, the incentives each system has are different or even in conflict, given payers and providers face different pressures, according to our past reporting.

“If we want an honest conversation about AI and cost, we have to look at both sides of the arms race, not just one,” Harbert said.