AI Is Adding Nearly $1bn to Blue Cross Health Costs – Study

Artificial intelligence tools are contributing to a sharp increase in the severity of medical conditions documented in insurance claims, adding nearly $1 billion to costs for Blue Cross Blue Shield insurers over a two-year period, according to a study released Thursday by the Blue Cross Blue Shield Association.

The findings highlight an emerging tension in healthcare's adoption of AI. Technologies such as clinical documentation tools and ambient scribes are designed to help physicians capture more information from patient records and conversations, potentially improving documentation and reducing administrative work. But insurers are increasingly questioning whether the additional diagnoses being documented are translating into more treatment or simply generating higher reimbursement.

Between 2024 and 2025, providers more frequently billed for secondary conditions, meaning conditions separate from the primary illness being treated. That increase added $653 million in costs for Blue Cross Blue Shield companies, according to the study.

Overall, more intensive care classifications contributed $942 million in additional costs over the two-year period compared with 2023.

The findings are based on inpatient billing data from hospitals and other medical facilities. The Blue Cross Blue Shield Association represents 31 independent health insurers covering more than 100 million people, according to a representative.

BCBSA said healthcare providers were using AI technology to identify secondary conditions by scanning existing patient records and deploying ambient scribes that listen to conversations between doctors and patients and automatically generate medical notes.

The issue centers on how hospitals and insurers classify the complexity of a patient's condition. When additional or coexisting conditions are documented, an inpatient case can be categorized as more complex, potentially resulting in higher payments from insurers.

The increase in documented conditions does not necessarily mean AI is causing hospitals to provide unnecessary treatment. AI systems may identify legitimate medical conditions that previously went undocumented. But BCBSA's analysis raises questions about whether the increase in diagnoses is accompanied by corresponding changes in patient care.

"If patients are truly sicker, we'd expect to see more treatment," said Luke Chalker, senior vice president of product and data science at BCBSA.

The association found notable increases in certain secondary conditions among patients undergoing major bowel surgeries. Between the first quarter of 2023 and the fourth quarter of 2025, diagnoses of partial intestinal blockages increased 55%, while diagnoses involving an excess of acid in the body increased 33%, according to the study.

Yet BCBSA said the rise in documented complexity did not correspond with higher treatment rates among patients undergoing bowel-disease surgeries. That disconnect matters for insurers because reimbursement systems often rely on the severity and complexity of a patient's condition. If AI-enabled documentation increases the number of conditions recorded without a corresponding increase in treatment, insurers could face higher claims costs without evidence that patients required substantially more medical intervention.

Dr. Razia Hashmi, BCBSA's vice president of clinical affairs, pointed to anemia as one example. Anemia can require treatment such as blood transfusions when a patient's red blood cell count is insufficient. Hashmi said the study did not find an increase in transfusions corresponding with the greater number of anemia diagnoses.

"The disconnect between diagnoses and treatment suggests that AI is identifying more billable conditions, not sicker patients," Chalker said.

That conclusion has significant implications for the economics of healthcare AI. The technology was initially promoted largely as a way to reduce administrative burdens, allowing doctors to spend less time documenting consultations and more time with patients. Ambient AI scribes, for example, can listen to clinical conversations and prepare draft notes that physicians review.

But the same systems can also extract information that may affect how a patient's case is coded. As these tools become embedded in hospital billing and documentation workflows, the financial consequences can extend beyond administrative efficiency.

For insurers, the concern is that better documentation could become a mechanism for systematically increasing reimbursement.

Health insurers, including Centene, have already raised concerns that AI adoption by health systems has contributed to aggressive or inappropriate reimbursement claims. The BCBSA findings add data to a broader debate over how artificial intelligence is changing the relationship between clinical documentation, coding, and payments.

The gap between improved documentation and upcoding is likely to gain more interest. A patient may genuinely have several conditions that were previously missed or poorly documented, in which case AI could simply be making the medical record more complete. But if newly identified conditions have little effect on treatment or patient outcomes, insurers may question whether the additional diagnoses should command higher payments.

There is also a broader question about who ultimately bears the cost of AI-enabled changes in medical coding. Higher reimbursements paid by insurers can feed into overall healthcare spending, potentially affecting premiums and costs for employers and consumers.

Hospitals and doctors may argue that they should be compensated appropriately when technology allows them to identify conditions that were previously overlooked. The BCBSA study does not establish that the diagnoses were inaccurate or fraudulent. Its central finding is that the rise in documented complexity has outpaced changes in treatment in some areas.

The finding is expected to matter more as regulators, insurers and healthcare providers develop rules around AI-generated clinical documentation.

The healthcare industry is rapidly moving toward systems that can read medical records, listen to consultations, identify diagnoses, and automate portions of clinical administration. The financial impact could therefore extend well beyond the cost of buying AI software.