We have a rather high overhead ratio in play. 1.1 million Physicians made a median salary of $239,200 in 2023, or around $263M.
U.S. health care spending grew 7.5 percent in 2023, reaching $4.9 trillion or $14,570 per person.
That's about 5% of healthcare spending going towards actual physicians. The logistic exigencies of dealing with insurance companies, seeking permission to treat, justifying, arguing, interacting with medical secretaries, is something on the order of an additional half of the typical doctor's time. They are strongly disincentivized from seeking out low probability explanations using tests which they need to seek permission to apply. They barely have the time to re-familiarize themselves with your chart.
We locked down the supply of physicians in 1997 via Congressional limits in how we fund medical residency slots. We further restrict physician time with problematically low rate setting in Medicare/Medicaid compensation, relative to market rates.
Everybody dies eventually. Everybody has healthcare needs.
A very large fraction of us will at some point suffer from some uncommon or rare condition, something that a physician doesn't see every day - a "zebra" rather than a "horse". There are enough of these conditions - tens of thousands now recognized - that the very low odds of having any particular problem adds up to high odds of suffering from at least one.
To recognize some of them, the doctor would need to synthesize an idea from half a dozen disparate pieces of information, two of which you've mentioned casually in a visit three years ago, two of which you've never told anyone, and one of which is in your chart, and one of which you mentioned casually to a specialist but who did not think it relevant enough to note down... And then they would need to do a differential against hundreds of megabytes of medical research. In five minutes allocated to your visit.
Doctors are trained with a certain degree of professional image to maintain, a sort of 'Wizard' status that must be kept to a tone of mysterious respect. Nobody wants uncertainty from a doctor, despite the fact that all of statistical reasoning and diagnostics is uncertainty, quantified. If they lose this image, it is felt, people start injecting bleach or suing them for malpractice or killing themselves over an innocuous granuloma that their doctor assures them is nothing. That is not a healthy environment to drill-down low evidence of probability outcomes.
Particularly if every test and treatment impoverishes you, and the impact of that poverty on your stress level and lifespan is dramatic.
While tripling the number of doctors would certainly help, AI and a great deal more population-scale diagnostics is just better suited to some of these limitations.
Is that what it will be used for? No. Initially, at least, it will actually be used to make everything worse, to automatically deny treatments, to argue with a doctor with the persistence of a chatbot.
And that's just for the stuff that's been medically recognized. Medical science has put comparatively little effort into researching ailments that cause chronic, non-contagious, non-fatal problems in your life, and less into resolving them. There is no agreement on how dandruff works, lower back pain was dismissed as psychogenic for decades, most headaches are undiagnosable, skincare is witchcraft.
A good doctor who sees you for a brief period every year is staring at your chart and furiously googling some of the things listed there to re-familiarize themselves for five or ten minutes before the visit, just to maintain an image of professional competency by not forgetting the words. Actually applying recent research... how?
>We locked down the supply of physicians in 1997 via Congressional limits
This part though is wrong. We temporarily limited the number of resident slots, but total number of residents has been growing for a while now. Not as fast as many people would like obviously.
Just looking at the numbers for recent years the total number of residents jumped from 134k in 2019 to almost 163k in 2024.
If you look at total number of physicians per 10k people, that number has been growing a steady rate since the 60s.
There are now 2x as many physicians per capita as there were in 1975.
U.S. health care spending grew 7.5 percent in 2023, reaching $4.9 trillion or $14,570 per person.
That's about 5% of healthcare spending going towards actual physicians. The logistic exigencies of dealing with insurance companies, seeking permission to treat, justifying, arguing, interacting with medical secretaries, is something on the order of an additional half of the typical doctor's time. They are strongly disincentivized from seeking out low probability explanations using tests which they need to seek permission to apply. They barely have the time to re-familiarize themselves with your chart.
We locked down the supply of physicians in 1997 via Congressional limits in how we fund medical residency slots. We further restrict physician time with problematically low rate setting in Medicare/Medicaid compensation, relative to market rates.
Everybody dies eventually. Everybody has healthcare needs.
A very large fraction of us will at some point suffer from some uncommon or rare condition, something that a physician doesn't see every day - a "zebra" rather than a "horse". There are enough of these conditions - tens of thousands now recognized - that the very low odds of having any particular problem adds up to high odds of suffering from at least one.
To recognize some of them, the doctor would need to synthesize an idea from half a dozen disparate pieces of information, two of which you've mentioned casually in a visit three years ago, two of which you've never told anyone, and one of which is in your chart, and one of which you mentioned casually to a specialist but who did not think it relevant enough to note down... And then they would need to do a differential against hundreds of megabytes of medical research. In five minutes allocated to your visit.
Doctors are trained with a certain degree of professional image to maintain, a sort of 'Wizard' status that must be kept to a tone of mysterious respect. Nobody wants uncertainty from a doctor, despite the fact that all of statistical reasoning and diagnostics is uncertainty, quantified. If they lose this image, it is felt, people start injecting bleach or suing them for malpractice or killing themselves over an innocuous granuloma that their doctor assures them is nothing. That is not a healthy environment to drill-down low evidence of probability outcomes.
Particularly if every test and treatment impoverishes you, and the impact of that poverty on your stress level and lifespan is dramatic.
While tripling the number of doctors would certainly help, AI and a great deal more population-scale diagnostics is just better suited to some of these limitations.
Is that what it will be used for? No. Initially, at least, it will actually be used to make everything worse, to automatically deny treatments, to argue with a doctor with the persistence of a chatbot.
And that's just for the stuff that's been medically recognized. Medical science has put comparatively little effort into researching ailments that cause chronic, non-contagious, non-fatal problems in your life, and less into resolving them. There is no agreement on how dandruff works, lower back pain was dismissed as psychogenic for decades, most headaches are undiagnosable, skincare is witchcraft.
A good doctor who sees you for a brief period every year is staring at your chart and furiously googling some of the things listed there to re-familiarize themselves for five or ten minutes before the visit, just to maintain an image of professional competency by not forgetting the words. Actually applying recent research... how?