By MICHAEL L. MILLENSON
Long before ChatGPT, whose question-answering choices still remain somewhat of a black box, there was an equally mysterious, question-answering black ball. I decided to ask them each of them how to solve the cost, quality and access issues labeled for more than half a century as “the healthcare crisis.”
The hard, plastic Magic 8 Ball was invented in 1946, two years before a landmark Supreme Court decision spurred a boom in employer-sponsored health insurance. It catapulted into kid-driven popularity in the 1970s, the same decade that rising healthcare costs propelled “healthcare crisis” into the public vocabulary.
The healthcare crisis is still with us, as is Magic 8 Ball, which, thanks to current owner Mattel, can now be consulted either in person (i.e., by holding and shaking it) or online. With a fiercely fought presidential election campaign underway, I decided that pitting the black box vs. the black ball to answer crucial health policy questions would likely provide just as much clarity as wading through weasel-worded white papers.
Both ChatGPT (Cost to OpenAI: $400,000 per day to operate) and Magic 8 Ball (One-time cost: $14.99) were up for the challenge, though they acknowledged it wouldn’t be easy.
“Can you help me solve the healthcare crisis?” I asked. “Signs point to yes,” Magic 8 ball replied, in its typically pithy, understated manner. ChatGPT, on the other hand, took my question as an invitation to show off its artificial intelligence.
“Addressing the healthcare crisis is a complex and multifaceted challenge that requires a holistic approach,” ChatGPT began. Then, as if a Washington think tank had been crossed with an academic policy conference, the Large Language Model offered a very large helping of language. There were 8 “key strategies,” each of which contained three bullet points, and each of which, I was advised, “involves detailed planning, resource allocation, and collaboration among various stakeholders, including government, healthcare providers, insurers, and the public.”
Then there was this diss when I asked about its competitor. “It’s a fun toy,” sneered the chatbot (if chatbots could sneer), “but it doesn’t provide reliable or informed answers.”
I decided to home in on specifics.
“Is a government-run single payer system the right answer?” I asked. “My sources say ‘no,’” Magic 8 ball told me. ChatGPT was more positive, with caveats.
“A government-run single-payer healthcare system is one potential solution to the healthcare crisis, and it comes with its own set of advantages and challenges,” the chatbot replied. It added, “Whether it is the ‘right’ answer depends on various factors” – and then, once more, went on to provide a long list of relevant ones.
I decided to inquire about an approach with bipartisan support. “Is value-based healthcare the best way to control costs?”
“It is decidedly so,” said the Magic 8 Ball immediately. But ChatGPT, usually lightning quick, waited perhaps 20 seconds before not only responding positively, but presenting an overview and specific suggestions. There were 5 advantages and 5 challenges, plus 3 examples of possible strategies (accountable care organizations, bundled payments and patient-centered medical homes), all tied together with 5 considerations for implementation.
“Ultimately, VBHC can be a key component of a broader strategy to reform healthcare systems and achieve sustainable cost control,” ChatGPT concluded.
That pattern continued as I probed about the need for more effective financial incentives to reward high-quality, cost-effective care, a central component of VBHC. “It is certain,” Magic 8 Ball quickly agreed. ChatGPT, meanwhile, again paused for a lengthy period (by its standards) before responding “thoughtfully” (by human standards).
“Yes,” it said, “effective financial incentives are crucial for promoting high-quality, cost-effective care. Properly designed incentives can align the interests of healthcare providers, payers and patients, leading to better health outcomes and more efficient use of resources.”
The chatbot then listed 5 types of financial incentives, 5 key elements of effective incentive programs and three specific examples incorporating them.
Continuing the financial incentives theme, I asked whether health savings accounts could help. Magic 8 Ball simply replied, “Yes,” while ChatGPT carefully pointed out that while HSAs “offer some benefits, they are not a comprehensive solution to the broader health care crisis.”
Like politicians, both ChatGPT and Magic 8 Ball sometimes hedged. “Are hospital mergers good or bad for patients?” I asked. “Ask again later,” said Magic 8 Ball. “Hospital mergers can have both positive and negative impacts on patients,” responded ChatGPT, before presenting a long list of why either might be the case.
“Is private equity buying doctors’ practices good or bad for patients?” I inquired. “Concentrate and ask again,” evaded Magic 8 Ball, followed by an incomprehensible, “Most likely.” ChatGPT allowed that this was “a complex issue, with potential benefits and drawbacks for patients,” before going on to the kind of pro and con balancing act any politician might admire.
I decided it was time to cut to the heart of the matter.
“Will health care costs ever be effectively controlled in America?” I demanded.
Magic 8 Ball tried to spare my feelings – “Better not to tell you now”– while ChatGPT, in its elliptical way, pointed me towards the unpleasant truth. While the challenge was not “insurmountable,” answered ChatGPT, it would require a “multi-faceted approach” involving “strong political will, stakeholder collaboration, and continuous evaluation and adjustment of strategies.”
In other words, “No.”
Michael Millenson is President of Health Quality Advisors and a long time THCB regular, he’s also a Forbes columnist where this piece first appeared.