In their November 5 leader, “Patients are acting more like consumers. Good,” The Economist says that more and more Americans are taking their health matters into their own hands: asking AI to answer their health questions, bypassing traditional healthcare systems in favor of buying medications (especially GLP-1 obesity meds) via websites (including some run by the drug makers).
They're definitely right about this activity: OpenAI says more than 300 million people ask ChatGPT about health every week. Morgan Stanley estimates self-ordered tests are a $4 billion a year market in the US, with that number having doubled since 2021. And Eli Lilly says about half of its new obesity prescriptions in the US are filled directly, via its website, for cash — bypassing traditional prescribers and insurers.
Where The Economist stumbles is in their explanation for all this.
The Economist's assumptions
In what is becoming a pattern, The Economist misses the disruption in health, and doubles down on assumptions.
For example, The Economist says that people are using these tools because they can't get or afford a visit:
“Yet for all that money [spent on healthcare in the US], about one adult in ten could not get an appointment when they needed one last year, and more than a quarter put off or skipped care because of the cost. No wonder patients are taking matters into their own hands.”
But later in the article the authors say that “Nearly all consumer health care is paid for out of pocket, so for now it mostly serves the well-off. ”
So if I follow this, we're saying that
- people use ChatGPT for their health questions because of the high cost of healthcare
- also: most of the people using ChatGPT for their health questions are well-off
How, I wonder, does it make sense to posit that it is the well-off who are most sensitive to healthcare costs, and least able to get appointments?
Well, it doesn't.
To make matters worse, The Economist ignores the fact that 96% of ChatGPT users pay nothing at all — hardly the definition of a luxury good — and GLP-1 pills start at $149, a price that many of my working-class patients pay, recognizing that it's less expensive to them than taking an unpaid day off from their hourly-wage job to visit the clinic.
A different assumption
Like The Economist, I have an assumption, a different one. Mine is that AI, along with much of the preceding information technology, is a technology that disrupts knowledge monopolies. Like that of the healthcare system.
The Economist gives a nod to this in the article:
Most profound of all is the erosion of the original source of market power: a doctor’s hold on medical knowledge. In 1963 Kenneth Arrow, an economist, said that health care could not work like an ordinary market partly because the information a doctor possesses is “necessarily very much greater than that of the patient, or at least so it is believed by both parties”. Technology has since narrowed that gap.
But after noting that technology is narrowing the gap between the knowledge of the doctor and the knowledge available, via technology, to the average consumer, The Economist fails to apply that to the phenomenon they are describing.
The gap between a consumer's knowledge and the doctor's is exactly why we choose to go to the doctor, and why that consumer chooses to pay increasing amounts of time and money to do so. But if technology has narrowed that gap, and continues to do so, we would expect that consumer to be correspondingly less willing to pay up.
Applied to health, that means people choose these tools because they're cheap and fast and convenient and good enough. This, of course, is a big part of Christensen's famous theory of disruptive innovation: disruptive products are worse in substantial ways than the products they are disrupting, but better in other important ways, usually price or convenience.
Do many people think that asking health questions of Claude is as good as asking them of a doctor with 20 years of practice experience? I suspect not, but do they think it's good enough? Increasingly they do.
Settling the question
The Economist is probably correct that scarcity and cost explain part of this consumer health phenomenon. But the authors fail to acknowledge that it is an assumption, not a proven fact. And they miss the opportunity to then make note of what data we would need to gain a better understanding of what's happening, data that we just don't have yet. Data that would show which of our assumptions plays a greater role.
For example:
- Why people use these tools. Ask people who use AI for health questions, or buy GLP-1s and lab tests direct, why they did. Then break down the answers by insurance status and whether they have a regular doctor.
- Who uses them. Income and insurance data on ChatGPT health users, direct GLP-1 buyers and people who order their own labs, compared to those who don't
- What happens next. Whether people who ask AI first skip the visit, delay it or book it anyway, and how they do afterwards.
It’s probably some mix of scarcity, cost, and good enough. But each calls for a different response. More appointments fix scarcity. Lower prices fix cost. Neither brings back the patients who left because the consumer alternative is good enough.
Getting the mix wrong is expensive. If policymakers take The Economist’s word that consumer health care mostly serves the well-off, we can expect Congress to start proposing programs to put these tools within reach of the less well-off — who are likely already using them.
And if health system execs believe that this behavior stems from a lack of available appointments, they may spend money increasing slots, only to find that “good enough” accounts for more than they or The Economist realize.