Why do physicians make so much?
According to this WaPo article, the average physician in the US earns $350K/yr. I didn’t click through to the actual data, but from the first table, I’m guessing that “average” means median, not mean. And physician income isn’t a Gaussian distribution—there’s a long right-hand tail for the specialties.
Why is this? It looks to me like supply-and-demand is a big factor. Despite the fact that allopathic and osteopathic medical schools have expanded class sizes in the last ten years, the supply of physicians hasn’t reflected that growth. This is because you can’t practice in the US without completing a residency, and until very recently, residency slots were not growing. Domestic and foreign-trained medical students compete for residency slots, so increasing domestic grads without commensurate increases in residency slots just means more competition, not more doctors.
Another factor mentioned is debt. Med students graduate with hundreds of thousands of dollars in loans. There’s also the opportunity cost of medical school, residency and for the higher compensated specialties, fellowships.
Obviously, not all physicians are raking in the big bucks. Primary care physicians and pediatricians are among the most poorly compensated (relatively speaking).
Not only is this not sustainable, but I suspect that it is already starting to change. Increasingly, tasks that were formerly performed by MDs are being done by physician assistants and nurses. Many routine surgeries are done partly or completely using robots. And in specialties like radiology, AI is proving to be more reliable than human docs, which will reduce demand for this specialty.
Why do physicians make so much?
Hate to bring this up, but
My two best friends from highschool went on to become physicians. (In those days it was a way to avoid the draft, and also to get into a high-paying profession. At least the last part is still true.) They will tell you that they have/had to pay a lot on education loans, and also a lot on malpractice insurance. But, anyway, they still make lots less than rock stars and pro athletes.
Can’t bring up pay for actors. because most actors just don’t make that much.
@Fred,
Yep, and less than Fortune 500 CEOs, too. But way more than the median household income in the US ($71,000/yr).
One of my best courses in highschool was biology, which still mystifies me because I really didn’t get biology. Probably not your experience. But microbiology probably doesn’t pay they way medicine does.
@Fred,
My PhD is in genetics. My wife’s PhD is in microbiology. We both were happy with our compensation. A prof of mine in grad school told me to get a faculty position at a medical school, because they pay more for the same job description. We both followed his advice. He was right.
Y’know, AI could end up doing a lot of damage to professions like medicine & law, not to mention creative writing. HVAC, fine carpentry, lawn maintenance may be ok though.
@Fred,
So far, AI has only damaged career opportunities for some lawyers and some physicians. I don’t count that as damage to the profession, only damage to certain career paths. Kinda like what cars did to buggy whip manufacturers or digital photography did to film.
What, in your view, are the top three downsides for AI in those fields?
A major problem with AI in medicine and law presently is that it takes as much time to check its accuracy as it would have taken to do the work “manually” in the first place. It’s tendency to make stuff up makes it actually dangerous in the context of medicine and law.
@ Jack, I think you’re confusing chatGPT with AI.
Here’s a summary of a recent study comparing AI-assisted breast cancer screening (intervention group) to double reading without AI (control group):
I read the article in The Lancet Oncology, and Drum doesn’t point out a concern about overdiagnosis, which is why oncologists will still be in the loop.
https://jabberwocking.com/in-new-study-ai-finds-20-more-breast-cancers-than-human-radiologists/?fbclid=IwAR3KtDZrXGzYKd_AFKP337RjabyJYp6OCOlgJM5WN6T_FsSJxoz0Fu-GdxM
Could be but the articles I’ve seen so far have looked at legal research, brief and motion writing for law. Evidence evaluation of large masses of data is another matter. I don’t have enough medical background (despite many years of handling injury and malpractice cases) to comment on AI applications there. I do recall studies demonstrating that requiring use of checklists, particularly in emergency rooms, prevented lots of errors
There is more of the same:
https://news.cgtn.com/news/2023-06-14/Macao-led-research-develops-AI-based-model-for-clinical-diagnostics-1kCTJCe2J68/index.html
June 14, 2023
Macao-led research develops new AI-based model for clinical diagnostics
A research team led by Macao University of Science and Technology (MUST) has developed a new AI-based model as a clinical diagnostic aid that processes multimodal input in a unified manner.
The model IRENE was designed to help make medical decisions by jointly learning holistic representations of medical images, unstructured chief complaint and structured clinical information, according to the team, which also included researchers from the West China Hospital of Sichuan University and the University of Hong Kong.
The study * was published in the Nature Biomedical Engineering…
* https://www.nature.com/articles/s41551-021-00704-1
@Jack,
Law and medicine are two very different fields. I don’t know of any field of law that prescribes medication, delivers babies, sets bones, ties sutures, performs surgery or images organs. My guess is that AI means very different things to the legal profession than what it does to the medical profession (other than malpractice law).
Joel:
Meet Jack Daniels a former law partner and founder of a firm in Chicago. I have known Jack for 20+ years. Helped myself and the family get through some rough times. He is a good guy to know just like you are. 🙂
@Joel,
Law doesn’t do any of those things. It does protect property, protect civil rights, keep innocent people out of prison, and help to keep the society from reverting to the law of the jungle; so I’d say it needs protection from technological distortion. Just sayin’/
Jack;
Meet retiring Professor Joel Eissenberg Department of Biochemistry and Molecular Biology Saint Louis University. Joel is helping me run Angry Bear since my friend Dan Crawford died of cancer recently. Not that this means you should not discuss things.
Verification of results, or the difficulty of doing so, would be a major problem.
Also said of quantum computing. Just because a solution is presented, there are probabilities associated with determining ‘veracity’, not certainty.
But based on experience with automation & computing, the career disruptions of AI & eventualy QC would likely be severe. In the end, there should be major benefits, or at least there could be. But there will job displacement for sure.
Here are my three.
1) The biggest is the matter of liability. Doctors, lawyers, engineers, and other certified professionals can be sued if they fail to do the work properly. Who gets sued if an AI tool makes a mistake? It could be the professional who relied on the tool or it could be the AI tool manufacturer. In either case, liability issues are going to slow adoption. (e.g. I remember reading a specification for a CPU chip that had an explicit section stating that the chip could only be used in medical applications with permission of the chip manufacturer.)
2) There’s also the matter of validation. That Swedish study sounds good, but it doesn’t have a lot of statistical power. By that measure, I’m not sure it’s even good enough to justify use of the system in Sweden, a relatively small country. It’s definitely not good enough to justify it in a larger nation like the UK or France. Radiology is a high risk profession, so radiologists tend to be conservative. They were cautious moving to digital imagery and wary of even simple image enhancing technologies like gamma adjustment. (e.g. I remember talking to a photogrammetry expert about the move to digital, and he was going to wait until someone did the necessary math and enough people validated it.)
3) There’s the matter of how well the tool works as part of best practice. There was a time when everyone was told get a mammogram, hey, get two, even you guys, get one. Then, it turned out that detecting a growth might have little or no impact on health or life span. So many tumors were so slow growing that the risks of intervention would be detrimental. It’s not clear that the AI system is solving an important problem.
@Kaleberg,
There are AI studies in clinical practice going back a decade. Time marches on and AI is improving. It probably won’t stop now.
It’s very clear to me and to the people who use AI in medicine that it is solving important problems. If it were not, people would not be investing huge amounts of money in it.
You are right that a risk of overuse of imaging is the diagnosis of “incidentalomas,” indolent or slow growing tumors. This has nothing to do with AI. This has to do with overuse of diagnostics to justify the purchase of imaging technology.
Radiology is a high risk profession, so radiologists tend to be conservative. They were cautious moving to digital imagery and wary of even simple image enhancing technologies like gamma adjustment. …
A New Working Paradigm for Radiologists…
Journal of the American College of Radiology – Feb 2022
Despite the fact that allopathic and osteopathic medical schools have expanded class sizes in the last ten years, the supply of physicians hasn’t reflected that growth. This is because you can’t practice in the US without completing a residency, and until very recently, residency slots were not growing….
[ I had no idea this was so. Thank you. ]
Where the international and domestic data can be located:
https://data.oecd.org/healthres/doctors.htm
Despite the fact that allopathic and osteopathic medical schools have expanded class sizes in the last ten years, the supply of physicians hasn’t reflected that growth….
[ The data are stark, with the US physician/population ratio far lower than any Western European country. I will work this up, but again I had no idea this was so. ]
https://data.oecd.org/healthres/doctors.htm
July, 2023
Practising physicians per 1,000 population, 2021
Austria 5.480
Norway 5.160
Germany 4.530
Spain 4.490
Switzerland 4.440
Denmark 4.380
Sweden 4.320
Czech Republic 4.260
Italy 4.250
Australia 4.020
Ireland 4.020
Slovak Republic 3.680
New Zealand 3.620
Finland 3.610
Poland 3.440
France 3.360
Israel 3.350
Slovenia 3.340
Hungary 3.300
Belgium 3.250
United Kingdom 3.180
Canada 2.800
United States 2.670
Japan 2.600
Korea 2.560
https://data.oecd.org/healthres/doctors.htm
July, 2023
Practising physicians per 1,000 population, 2021
Austria ( 5.48)
Norway ( 5.16)
Germany ( 4.53)
Spain ( 4.49)
Switzerland ( 4.44)
Denmark ( 4.38)
Sweden ( 4.32)
Czech Republic ( 4.26)
Italy ( 4.25)
Australia ( 4.02)
Ireland ( 4.02)
Slovak Republic ( 3.68)
New Zealand ( 3.62)
Finland ( 3.61)
Poland ( 3.44)
France ( 3.36)
Israel ( 3.35)
Slovenia ( 3.34)
Hungary ( 3.30)
Belgium ( 3.25)
United Kingdom ( 3.18)
Canada ( 2.80)
United States ( 2.67)
Japan ( 2.60)
Korea ( 2.56)
This highlighting of US doctor pay seems to be part of a propaganda campaign by the increasingly consolidated regional medical monopolies to increase their profits by cutting the cost of their high end labor. Most doctors are employees or contractors or even subcontractors now. They aren’t independent practitioners, and they aren’t the ones running the hospitals and clinics as they were until fairly recently. As far as making better numbers for the next quarter to fund the next share buyback or round of executive bonuses, doctors are just overpaid flunkies.
The problem is that a lot of people still have respect for the profession, so cutting them and their salaries down to size requires getting people to think that they are overpaid, lazy, whiny, irresponsible and so on. Getting people to resent the wages and benefits of union workers made it easier to smash the unions. Now it’s the doctors who need smashing, so we’re going to see a lot more articles like this one.
@Kaleberg,
So you think the WaPo (which was the source of my post on US doctor pay) is part of a propaganda campaign by the increasingly consolidated regional medical monopolies to increase their profits by cutting the cost of their high end labor? Do you have evidence for this, or is this just a conspiracy theory? Seems kinda far-fetched to me.
The original post was excellent.
Dean Baker, Beat the Press, would tell you that doctors have rigged the system to protect themselves from market competition. Demanding a residency at a US facility when other countries are demonstrably able to produce well-trained doctors is one example of rigging the system to “redistribute income upward.”
@Arne,
You may be right. Residency slots are funded by federal, state and private sources, but the federal government is the biggest source. If there were more federal dollars, there would be more slots. It wouldn’t surprise me if the AMA lobbies Congress to keep a tight leash on those funds, but I don’t really know.
The US has long accepted foreign-trained MDs who pass the boards, and many of these docs go on to underserved communities (see, e.g., Abraham Verghese, “My own country”). These docs compete for US residency slots. America could have lots more doctors if we really wanted them.
When I first started as a med school faculty, a student could fail the USMLE step one exam on the first try, pass on the second try and go on to graduate and match for a residency. Nowadays, if you fail on the first try, it’s nearly impossible to match for residency. These folks are stuck with hundreds of thousands of dollars in loans and can’t work as physicians.
Lagging behind almost all OECD countries in 2000 and in 2021:
https://data.oecd.org/healthres/doctors.htm
July, 2023
Practising physicians per 1,000 population, 2000-2021
United States
2000 ( 2.29)
2021 ( 2.67)