Amid Hype, Experts Debate AI’s Impact on Health Insurance

Hype of artificial intelligence reached new heights in 2023, and the health care industry was no exception to that trend. Health insurance firms have been among the first organizations to deploy previous generations of big data tools at scale, and in turn, AI is set to proliferate among health insurers in the near future. How AI adoption will actually impact health insurance operations and finances is an open question.

Some insiders, like Moody’s Investors Service analysts, expect health insurers to reap dramatic operational and financial benefits by deploying AI. Other experts disagree, including a group of Stanford University researchers who published a Nov. 16 op-ed in JAMA arguing that due to the extreme complexity of medical administration in the U.S., AI using large language models (LLMs) “might exacerbate billing challenges for physicians” and could lead to additional challenges of filed claims by carriers.

Dean Ungar, vice president at Moody’s, says that AI is poised to revolutionize the health insurance industry. In an Oct. 24 research note, Ungar wrote that “with the recent advent of AI, health insurers will have another tool to create more efficiency and lower costs.”

AI is going to allow insurers to “increase the competitiveness of your products,” Ungar tells AIS Health, a division of MMIT. “You have more capacity to handle more volume. At the same time, the costs aren’t the same. [The AI] can do the work, so you don’t need more people.”

The Stanford researchers, on the other hand, point to the sheer complexity of the U.S. health care system as a reason why AI is unlikely to have the impact that evangelists say it will. They cite Bill Gates’ axiom that “the first rule of any technology used in a business is that automation applied to an efficient operation will magnify the efficiency. The second is that automation applied to an inefficient operation will magnify the inefficiency.”

“There are 57 billion negotiated prices in the U.S. market, or 94,335 for each service code,” the Stanford researchers write. “Seen in aggregate, it’s hard to justify this complexity and lack of standardization of the health insurance model in the U.S. This is not a question of how health care is financed, but a question of transaction processes underlying the market. Transaction costs in the U.S. are 10 times those in the Netherlands, which [also] has a private, multipayer health care system. LLMs are clearly an exciting technology, but the current market environment is far from optimized to enable this technology to provide a solution for practicing physicians. In fact, adding LLMs to this milieu might exacerbate billing challenges for physicians.”

UnitedHealth Says AI Already Generates Value

The overall impact of AI on health insurance will be difficult to measure in the near term. But firms including UnitedHealth Group claim that AI has already yielded significant benefits to their business — even though the diversified insurer is facing down a new lawsuit brought by Medicare Advantage members who allege that automated claims processing tools unfairly denied them coverage. It’s not the first time UnitedHealth has fought a legal battle over claims automation and big data software. Carriers including The Cigna Group have faced similar litigation.

Sandeep Dadlani, chief technology officer at UnitedHealth, said during the firm’s Nov. 29 investor day that AI is in place and delivering results — or is being implemented — in each of the company’s lines of business.

“We are at the cusp of a new era and generative AI is parting a groundswell of innovation across UnitedHealth Group,” Dadlani said. “We’ve identified three core focus areas for our AI strategy. First, administrative simplification. Here’s where we can make a big impact quickly. Taking friction out of the system in areas like claims processing or provider search and payment integrity. Second, data science platforms and businesses that underpin our growth pillars and capabilities. This is key for enabling value-based care. And finally, medical and clinical insights. Our goal is to free up the clinicians’ time, so they can focus on the care part of health care.”

In his presentation, Dadlani offered hypothetical examples of how UnitedHealth expects its existing AI tools to improve its operations. The firm’s chatbots and virtual assistants can use generative AI to analyze a given member’s medical data and claims history to better route care, he said: “Through advanced voice technology and generative AI, this initial interaction is now becoming intelligent. It can solve more problems than ever before,” and better prepares UnitedHealth employees for subsequent conversations with members.

In addition, Dadlani said, to find a “lower-cost option” for a given patient’s prescriptions, AI interlocutors “auto-generate suggested answers based on seamless integration with Optum Rx’s pricing tool.”

“There are dozens more generative AI use cases being rolled out,” Dadlani added. “We are using AI to solve our consumers’ most frustrating pain points, support our frontline workforce and enable clinicians to practice at the top of their license. Our collective vision is for a world where AI can aid in the prediction and ultimately the prevention of disease.”

Existing Big Data Use Offers AI Lessons

The current deployment of existing big-data tools offers an instructive example of how the AI arms race may play out in the health insurance industry.

While Ungar tells AIS Health that he’s certain that AI is “going to give you an opportunity to, first of all, get better outcomes,” he does say that the technology’s impact won’t be immediate.

In a Moody’s report, Ungar observed that “big data analytics…has been an increasingly important resource for health insurers” in recent years, and “while data analytics is making inroads in the sector, it still is in the early stages. The data not only has to be collected and analyzed; it has to be effectively transmitted to health providers who can access it easily and make it part of their normal workflow.”

AI, in other words, is just an acceleration of existing trends. Meanwhile, the Stanford researchers point out that big data powered by federally mandated price transparency disclosures has not yet slowed rampant cost growth in health care.

“Over the last year, the underlying business process [of health care] has come into clearer focus,” they write. “Data obtained under the Centers for Medicare & Medicaid Services transparency rules show that there are 317,987 health plans in the U.S. market, roughly one plan for every 1,000 people in the country. In the worst-case scenario, each plan can have its own set of covered benefits, its own documentation standards, its own payment processes, and payment rules.”

The Stanford researchers say that “rather than hope that LLMs are a panacea, the emergence of this technology calls for an extraordinary effort to address the underlying challenges of transactions in the U.S. market. Calls for administrative reforms in health care date back several decades but have never garnered the combined attention of the public and private sector. This emerging challenge of LLMs must be the catalyst to finally move this agenda forward.”

But Ungar says that sort of thinking is itself utopian.

“Everyone says we spend too much” on health care, he says. “The health insurance industry has to walk that fine line between not paying for things that aren’t really necessary and checking [spending] that may need to be checked. I’m sure you can find people who have counterexamples where they’re the victim. But there’s always going to be that tension…between providing the best care in the most user-friendly way” and “whenever you get a claim denied.”

“It’s always annoyed people,” Ungar says.

Contact Ungar via Michael Simon at Michael.Simon@moodys.com.

This article was reprinted from AIS Health’s weekly publication Health Plan Weekly.

0

0

0 Comments

0

0

• Spotlight on Market Access