AI in U.S. Healthcare: Efficiency Tool or Driver of Rising Costs?

AI is moving deeper into an already expensive and administratively complex U.S. healthcare system, not just to help doctors diagnose and treat patients but to determine what hospitals bill, what insurance pays, and which claims get denied.

And for all the talk of AI as a productivity tool that will lead to efficiencies and serve as a deflationary force, some of the early evidence is raising questions about whether these tools will make an already costly system more expensive.

Blue Cross Blue Shield Association recently estimated that hospitals' use of AI-assisted medical coding contributed to close to $1 billion ($942 million) in additional costs for its health plans between 2023 and 2025. BCBSA said much of that increase came from secondary diagnoses that moved patients into higher-paying reimbursement categories, types of diagnoses that it said "may be derived from single laboratory values, making it particularly well suited for detection by AI tools."

The finding comes as hospitals, insurers, and other parts of the healthcare system increasingly use AI in billing, coding, and claims review, posing the question of whether reducing administrative burden can also intensify the financial incentives already built into the system.

According to Christopher Whaley, a health economist at Brown University who studies hospital coding, AI appears to be "accelerating, and in some sense making it easier to capture, the existing and underlying billing incentives that are in the system."

It is not that additional diagnoses identified through AI are necessarily inappropriate. "In many cases, the diagnoses are legitimate and weren't captured," he said.

But there are also conditions that "clinically just don't really matter and don't influence the patient's care," while still allowing another billing code to be applied and increasing payment, he said.

Blue Cross Blue Shield's AI findings

BCBSA wrote in its analysis that the growth in what it calls "complex coding" came during a period of time when 60% of hospital systems began using AI coding tools. "There is a clear disconnect between coding and treatment," its report stated.

Roughly 70%, or $653 million of the billing identified by the insurer, was tied to additional diagnoses that were not accompanied by a change in care, Luke Chalker, BCBSA's senior vice president of product and data science, told CNBC.

Chalker stopped short of attributing the entire increase to AI. "While multiple factors contribute to coding intensity, the findings suggest AI-enabled coding and documentation tools are playing a role," he said.

He added that consumers do have reason to be concerned. More complex coding can lead to higher reimbursement without more care. "Those costs can eventually show up in the form of higher premiums and out-of-pocket costs," he said.

According to the latest forecast from benefits consulting firm Marsh, the cost per employee for health coverage is expected to rise 8.2% on average in 2027, which would mark the highest increase since 2003.

The American Hospital Association pushed back on BCBSA's analysis. "Patients today are older and more clinically complex," and AI tools are helping providers "appropriately capture their patients' conditions to aid in care planning. The BCBSA's analysis lacks the context needed to meaningfully assess how these tools impact healthcare quality, patient access, or spending," an AHA spokesperson said in a statement to CNBC.

"It is particularly troubling to see insurers raising concerns about provider coding while continuing to rely on automated downcoding and denial practices that can impede coverage of medically necessary care, add burden on the workforce, and increase costs through administrative waste," the AHA spokesperson added.

Chalker said Blue Cross Blue Shield companies also use AI in claims review, but that "any clinical denial is always reviewed by a qualified human clinician."

An AI 'administrative arms race' in healthcare

As increasingly sophisticated AI tools are used on all sides of the healthcare system, it could create what Whaley called an "administrative arms race."

"Whether it's on the hospital side or the insurer side, these tools and technologies are both very expensive," he said, "and also have nothing to do with providing appropriate care to patients."

Whaley said those costs ultimately flow through the healthcare system to consumers, including through higher premiums and taxes.

Marisa Greenwald, a partner at Oliver Wyman's Health and Life Sciences practice, which advises hospitals and insurers on strategy, operations, and AI adoption in areas including revenue-cycle management, said AI is already producing benefits for health systems and doctors.

By reducing administrative work, physicians can spend more time seeing patients and less time finishing documentation after hours. Greenwald said health systems are using AI not only to improve efficiency but also to give doctors "some semblance of work-life balance back."

There is a financial benefit as well. Greenwald pointed out that with more time to see patients, physicians can increase patient volume, while more accurate coding can "catch additional acuity and diagnosis components," resulting in higher reimbursement.

But she said it remains difficult to know how much of the increase reflects genuinely better documentation versus other factors. "It's hard to disentangle how much of it is better accuracy. There's always going to be misuse and user error and overcoding."

Greenwald added that if providers use AI to improve coding and insurers respond with their own AI tools that lead to more denials, "the arms race is poised to exacerbate. The hope is going to be that on both sides of the equation, the players recognize that all we're doing is adding cost and burden into an already challenged and belabored system," she said.

Otherwise, healthcare could end up with "robots talking to robots and just fighting with each other," she added.

Hospital-insurer tensions are not new

Vanessa Moldovan, author of "The Healthcare Revenue Cycle AI Playbook" and head of RCM strategy at Magical, which develops AI-powered automation software for healthcare administration, said many of the tensions surrounding medical coding existed long before AI.

"The AI is new, but the rest of it is not new," she said. "You could line up 10 coders and have them all look at the same chart and they could code it differently," she added.

At the same time, providers operate under detailed insurer requirements governing which diagnoses and services will be reimbursed. AI can allow organizations to review far more records and documentation far more quickly. That can help providers get paid for care they actually delivered, Moldovan said.

But she drew a firm line when a diagnosis is not supported by the medical record. "If the patient didn't have those conditions, they didn't have those conditions," she said. That is one reason Moldovan opposes fully autonomous coding. "There should always be a human in the loop."

She said AI-generated coding should be audited much the way healthcare organizations have traditionally audited the work of human coders.

Chalker agreed, saying AI should "support decision-making, not replace human judgment."

The risks extend beyond whether a hospital receives a larger payment. Diagnoses can become part of a patient's medical record, and Moldovan said she worries about people placing too much confidence in AI-generated information.

"I think we're in danger of trusting it too much because we're like, 'Oh cool, it's AI, it must be smarter,'" she said.

But Moldovan also sees AI as a tool that can help providers respond to insurer denials. Revenue-cycle teams often lack the manpower to keep up with documentation requests and appeals, she said, explaining that AI can help them respond at greater scale.

"We can get more claims out. We can get more appeals out," she said.

That puts AI at multiple stages of the same billing dispute: helping providers document and code care, helping insurers scrutinize claims, and helping providers challenge denials.

But Brown's Whaley said the ultimate test should not simply be whether AI lowers healthcare spending.

If the technology increases spending while improving access or quality of care, he said, that could still be worthwhile.

"If AI tools are just used to kind of shuffle the cards a little bit more and make sure you come out on top and not do anything to patients," Whaley said, "then that's something that probably isn't worth investing resources in."