Is AI-driven coding fueling a rise in healthcare costs?

In March, the Blue Cross Blue Shield Association (BCBSA) and Blue Health Intelligence (BHI), the association’s data analytics partner, published research suggesting hospitals’ use of generative AI tools is linked to increased coding intensity and rising inpatient and outpatient costs. 

This research offers insight into the questions payers are asking about AI use among providers, but there is much more work to be done to understand what this technology means for healthcare costs as it is increasingly adopted by both insurers and providers.    

Billions more in spending? 

BHI analyzed commercial inpatient claims data between April 1, 2022, and March 31, 2025, with the intention of finding patterns in increasing coding intensity. Researchers wanted to find out if complex coding was increasing across the board or in a smaller group of hospitals. And they wanted to determine if hospitals with an increase in complex coding had discordant care; did the treatment change along with the severity of the conditions coded? The research homed in on acute posthemorrhagic anemia in maternity admissions. 

“What we found was that there was a pretty substantial shift — across the entire market but most noticeably at certain facilities…in the amount of CCs [complex conditions] and MCCs [major complex conditions],” Luke Chalker, chief product officer at BHI, said.  

Among 10% of hospitals with the most significant increase in complex DRGs, inpatient admissions with complex coding increased 59.8% by Q1 2025, according to the report. Among hospitals with the biggest jumps in postpartum anemia, treatment (transfusion) rates did not significantly increase. 

BCBSA indicates that the identified rise in complex coding coincides with the increased adoption of AI documentation and revenue cycle management tools. 

Coding intensity during the study period resulted in approximately $22 million more in spending for maternity admissions, according to the report. It suggests that “more aggressive, AI-enabled coding practices” could be tied to $663 million in inpatient spending and $1.67 billion in outpatient spending. 

Understanding how AI is impacting healthcare spending

While the BCBSA research suggests a connection between AI-driven billing and higher costs, it is not definitive. It does not identify what AI tools hospitals used, if any, during that period. And there are other issues that can impact coding intensity. 

“While the article had highlighted some of the DRG conditions specifically, it didn’t highlight the E&M codes that have had an ICD-10 changes that have occurred in the last several years that may be impacting the distribution of codes,” Jennifer Holloman, director of health IT policy at the American Hospital Association (AHA), told Digital Health Insights. 

 More research is necessary to understand how AI tools are being used and impacting healthcare costs, and not just on the provider side. Payers are also using AI tools. The National Association of Insurance Commissioners (NAIC) conducted a 2025 survey that found 84% of health insurers are using AI and machine learning in some way. They reported leveraging the technology for use cases such as prior authorization, fraud detection, utilization management, disease management program implementation and sales enhancement. 

The use of AI tools by providers and payers has become a so-called arms race. “The healthcare provider is trying to maximize its billings, its revenue, for the given services that it’s providing, and the health insurance company is trying to rationalize the costs that it experiences as an insurer,” said Amol Navathe, MD, PhD, a professor of medical ethics and health policy at the Perelman School of Medicine at the University of Pennsylvania and a senior fellow at the Penn Leonard Davis Institute of Health Economics. 

As payers and providers continue to use AI tools, where do researchers need to focus? Chalker sees the need for more work in the outpatient space and beyond the commercial population.   

“We have to apply all of this work on a government population because they’re the largest payer in our system,” said Chalker. “We all need to understand: What is the added cost to CMS?” 

The big question, according to Navathe, is: “Are these AI services actually reducing administrative costs in the system, or are they paradoxically increasing administrative costs in our system?” 

The need for new policy and oversight 

Research that elucidates the use of AI in healthcare and its impact on costs is essential for policymakers and other stakeholders tasked with developing oversight. 

“If we don’t have the data informing policymakers and industry, then we just aren’t going to have a place to even start,” Navathe said. 

Policy enactment moves at a pace much slower than AI innovation and adoption. Rather than pushing for ambitious changes immediately, Navathe argues that data collection is the place to start. 

“Let’s require collection of this data and then ongoing reporting, just like…we do for prescription drugs and certain types of devices. If we can get that post-marketing surveillance data, then that forms the foundation,” he said. “If we are able to grant some flexibilities to the government to be able to collect this data, then many stakeholders in the ecosystem can start to use that data.” 

Access to that data can help policymakers understand the ways in which AI is impacting spending without improving care and how to design policy that actually captures the benefits of AI, according to Navathe. 

Transparency is a common theme for stakeholders seeking to understand and manage AI in healthcare. In its study, BCBSA calls for: “Greater transparency and clearer expectations tying coded severity to objective clinical indicators and treatment are needed to ensure payment reflects the true complexity of delivered care.” 

In comments sent to HHS, the AHA has called for more “algorithmic transparency” so clinicians have insight into why requests are denied, citing the 2025 NAIC survey, which reports 77% of insurers do not disclose AI use to providers or physicians. 

In its recommendations to HHS the AHA is also “really encouraging agencies to synchronize policy frameworks with existing frameworks like HIPAA, ensuring that there are consistent standards across the stakeholders in the healthcare ecosystem from developers and third parties to providers as covered entities,” said Holloman. 

While more transparency and alignment will be necessary, Navathe is already seeing ways in which current payment policy is not fit for the increasing use of AI. 

“The way that we tend to price healthcare delivery or healthcare services is focused on the input cost. In other words, how much time does it take a doctor? How much skill does it take a doctor? What are the costs of supplies that are required to deliver a service?” he explained. “It’s not really going to work in the context of AI and software-based services because the whole notion of…the cost of delivering a service, it can be very, very low, whereas the cost of developing technology could be very, very high.” 

In an article of Health Affairs, Navathe proposes aligning AI reimbursement policy not with inputs but outcomes. 

 The debate around AI and healthcare spending is just getting started. As legislators and government agencies wade into the debate with the aim of creating policy and oversight to capture the benefits of AI, they need to address cost and value for all healthcare stakeholders.   

“Patients are the ones that we definitely don’t want to forget because they’re the ones who are paying the insurance premiums, and they’re the ones who oftentimes are paying significant out-of-pocket costs,” Navathe said. 

Carrie Pallardy, a Chicago-based freelance writer and editor, began her career covering healthcare more than a decade ago. Her work has taken into many different industries, but covering healthcare delivery remains a constant focus. She can be reached at [email protected] or on LinkedIn.