Trajectory of Artificial Intelligence (AI) in Medicine

[DrMetal]

Advertisement - Members don't see this ad

This has to be the trajectory of AI in medicine, right? . . .

2024: AI software exists, but it is prohibited. You're not allowed to use it for notes, dictation, etc.

2026: lots of AI software exists, your hospital/clinic now has it's own. It's use is optional.

2028: use of AI software is now strongly encouraged (it's gotten that good)

2030: use of AI is now mandatory (it's gotten better than you). You must use it to generate your note, you can still modify it at your discretion.

2032: use of AI is now mandatory, no modifications allowed. (it's gotten so good, that you must use it, and you're not allowed to change it)

 

[VA Hopeful Dr]

If it gets to that point I don't even think we will be needed anymore

 

[dozitgetchahi]

It may eventually go that way, but I think the time gaps are going to be much longer. Like 5-10+ years for each.

Recall that there are still some dinosaurs out there on paper charts.

 

[DrMetal]

It may eventually go that way, but I think the time gaps are going to be much longer. Like 5-10+ years for each.

Recall that there are still some dinosaurs out there on paper charts.

Ya know with respect to most technological advances . . . I'd agree with you (takes a long time to implement and accept). It took a while for lab machines to replace human benchtop chemists (who used to manually titrate your CBC/CMPs). It took a while to accept robotic surgery.

But AI feels different. Its too good, too easy, and it aims to replace the most mundane part of our work (documentation, computer work).

Unfortunately it'll result in a reduction of RVUs. If a hospital HNP was previously say 4.5 RVUs, with AI now generating your note (and possibly placing orders), now same HNP is only 2.5 RVUs. Now you have to see more patients to get paid the same. Instead of rounding on 15 a day, you'll have to round on 30, or 45 patients.

 

[sloh]

Now you have to see more patients to get paid the same. Instead of rounding on 15 a day, you'll have to round on 30, or 45 patients.

And insurers who use AI to audit will say these claims are denied or downcode with some ridiculous stringent guidelines

 

[chessknt]

This has to be the trajectory of AI in medicine, right? . . .

2024: AI software exists, but it is prohibited. You're not allowed to use it for notes, dictation, etc.

2026: lots of AI software exists, your hospital/clinic now has it's own. It's use is optional.

2028: use of AI software is now strongly encouraged (it's gotten that good)

2030: use of AI is now mandatory (it's gotten better than you). You must use it to generate your note, you can still modify it at your discretion.

2032: use of AI is now mandatory, no modifications allowed. (it's gotten so good, that you must use it, and you're not allowed to change it)

Doubtful--planets would have to align. The tech would have to meet standards, the regulatory framework would have to allow it, patients would have to buy in to this, health systems would have to view this as a positive in their market among many other factors. If we approach a singularity where doctors are being replaced by AI this will happen in lower risk fields first and we will probably see a massive societal backlash/upheaval if that happens which will render this somewhat moot as it will be a black swan event.

Honestly with our crumbling infrastructure it seems exceedingly unlikely they can scale AI to a national level for any sort of high utilization service like healthcare. People have quickly caught on to the toxicity of data centers to their health and energy bills and we don't have the energy capacity to build this out either. Data centers are becoming less popular than the nuclear power plants we will inevitably need to power them....

 

[DrMetal]

The tech would have to meet standards, the regulatory framework would have to allow it, patients would have to buy in to this, health systems would have to view this as a positive in their market among many other factors.

But this is all kind of happening, right now.

All your points about Data Centers (their toxicity) are true . . . but no one seems to care all that much. It doesn't seem to stop anyone from using or implementing AI.

 

[chessknt]

But this is all kind of happening, right now.

All your points about Data Centers (their toxicity) are true . . . but no one seems to care all that much. It doesn't seem to stop anyone from using or implementing AI.

I dont see regulatory framework moving at all and patients buying in to it seems unlikely. We have yet to see the legal system test this either.

I admit faith in the legal/regulatory system to protect us from predatory tech bros is a losing prospect these days but the one thing that might be a major barrier is growing group of people that want nothing to do with this as it rapidly invades every aspect of our lives.

 

[dozitgetchahi]

I dont see regulatory framework moving at all and patients buying in to it seems unlikely. We have yet to see the legal system test this either.

I admit faith in the legal/regulatory system to protect us from predatory tech bros is a losing prospect these days but the one thing that might be a major barrier is growing group of people that want nothing to do with this as it rapidly invades every aspect of our lives.

Also, people are already suing hospitals over the use of ambient AI scribing systems, saying they weren’t properly consented for their use etc.

Lawsuit Alleges AI Platform Illegally Recorded Patient-Clinician Conversations

A lawsuit has been filed in the U.S. District Court for the Northern District of California against two healthcare organizations over their use of an A clinical AI tool used by health systems to ease the burden on clinicians by recording, processing, and transcribing patient-clinician...

www.hipaajournal.com

 

[DrMetal]

Also, people are already suing hospitals over the use of ambient AI scribing systems, saying they weren’t properly consented for their use etc.

Lawsuit Alleges AI Platform Illegally Recorded Patient-Clinician Conversations

A lawsuit has been filed in the U.S. District Court for the Northern District of California against two healthcare organizations over their use of an A clinical AI tool used by health systems to ease the burden on clinicians by recording, processing, and transcribing patient-clinician...

www.hipaajournal.com

True, but that's not gonna stop the trajectory of AI (you can solve the consent problem . . . "Consent, or you won't be seen")

 

[sloh]

True, but that's not gonna stop the trajectory of AI (you can solve the consent problem . . . "Consent, or you won't be seen")

Agreed. AI is an unstoppable force.

“What is genuinely new with frontier models is that the technology has finally outpaced the rate at which buyers can deploy it. That is the inversion. For most of enterprise software history, the constraint was “the tool cannot do what we need.” Now the constraint is “the tool can do far more than our org can absorb.” Look at any decent ambient scribe rollout. The model can transcribe and structure a clinical encounter at near-physician quality. The deployment friction is the EHR write-back, the template alignment with each specialty, the billing code mapping, the medico-legal documentation policy, the physician training, the QA loop, the audit trail, the malpractice insurance carrier sign-off. Each one of those is a project. None of them is solved by a better model.”

The OpenAI Anthropic AI Arms Race Pivoted From Models To Services & Deployment. And Healthcare Is The Stress Test.

How OpenAI’s PE-Backed Deployment Co And Anthropic’s Blackstone-Hellman-Goldman JV Are Fighting To Own The Services Layer

open.substack.com

 

[DrMetal]

Agreed. AI is an unstoppable force.

“What is genuinely new with frontier models is that the technology has finally outpaced the rate at which buyers can deploy it. That is the inversion. For most of enterprise software history, the constraint was “the tool cannot do what we need.” Now the constraint is “the tool can do far more than our org can absorb.” Look at any decent ambient scribe rollout. The model can transcribe and structure a clinical encounter at near-physician quality. The deployment friction is the EHR write-back, the template alignment with each specialty, the billing code mapping, the medico-legal documentation policy, the physician training, the QA loop, the audit trail, the malpractice insurance carrier sign-off. Each one of those is a project. None of them is solved by a better model.”

The OpenAI Anthropic AI Arms Race Pivoted From Models To Services & Deployment. And Healthcare Is The Stress Test.

How OpenAI’s PE-Backed Deployment Co And Anthropic’s Blackstone-Hellman-Goldman JV Are Fighting To Own The Services Layer

open.substack.com

I honestly think it will replace most of us cognitive physicians. Question is how fast will that happen? ( And the younger patient pool, < 50 yo, doesn't care about talking to a doctor anymore. They just want quick results and faster service)

Proceduralists could also be replaced with AI coupled with
advancements in robotics, but that'll likely take more time.

 

[chessknt]

I honestly think it will replace most of us cognitive physicians. Question is how fast will that happen? ( And the younger patient pool, < 50 yo, doesn't care about talking to a doctor anymore. They just want quick results and faster service)

Proceduralists could also be replaced with AI coupled with
advancements in robotics, but that'll likely take more time.

Again this will be seen in other lower risk fields first (think customer service, insurance, banking, retail, food service) and it won't be accepted at scale that the entire white collar class that forms the backbone of the USA economy and a majority of the voting block will suddenly cease to be employed. The level of economic/political disruption to further enrich the villain class will be the beginning of a black swan event unless the politicians hold it back. We have had iterations of this over the past century with the invention of the internet but never at the rapidity that is being seen now. Print media/phone operators etc all had many years of warnings coming down the lines as their employment naturally dwindled with fewer people training for these positions but terminating employment for half the country over a 5 year period will be a society-changing event that I dont think the people who run the country will ultimately allow.

 

[dozitgetchahi]

Again this will be seen in other lower risk fields first (think customer service, insurance, banking, retail, food service) and it won't be accepted at scale that the entire white collar class that forms the backbone of the USA economy and a majority of the voting block will suddenly cease to be employed. The level of economic/political disruption to further enrich the villain class will be the beginning of a black swan event unless the politicians hold it back. We have had iterations of this over the past century with the invention of the internet but never at the rapidity that is being seen now. Print media/phone operators etc all had many years of warnings coming down the lines as their employment naturally dwindled with fewer people training for these positions but terminating employment for half the country over a 5 year period will be a society-changing event that I dont think the people who run the country will ultimately allow.

I agree.

First, I don’t see the current AI tech actually working to flip society over that quickly. I used to work in semi rural Alabama. One thing that surprised me there was how much *hadn’t* changed socially and technologically over the previous 50 years or so. Some things were still really backwards there. Some Doctor’s offices were still on paper charts, etc etc, never mind social views and customs which frankly had more in common with pre- WW2 America than anything else. Some people didn’t have cell phones. So the thought that suddenly AI is going to “change everything” in these types of places when the last 50-100 years of technological innovation didn’t seems a bit rich. Not only that, but the current capabilities of generative AI seem to be a bit…overstated at best. The tech bros have a lot of money to make by boasting that their tech can do all sorts of things. Reality suggests otherwise.

Second, I don’t see this happening politically - at least not as fast as the AI tech brain trust is claiming it will.

 

[DrMetal]

I agree.

First, I don’t see the current AI tech actually working to flip society over that quickly. I used to work in semi rural Alabama. One thing that surprised me there was how much *hadn’t* changed socially and technologically over the previous 50 years or so. Some things were still really backwards there. Some Doctor’s offices were still on paper charts, etc etc, never mind social views and customs which frankly had more in common with pre- WW2 America than anything else. Some people didn’t have cell phones. So the thought that suddenly AI is going to “change everything” in these types of places when the last 50-100 years of technological innovation didn’t seems a bit rich. Not only that, but the current capabilities of generative AI seem to be a bit…overstated at best. The tech bros have a lot of money to make by boasting that their tech can do all sorts of things. Reality suggests otherwise.

Second, I don’t see this happening politically - at least not as fast as the AI tech brain trust is claiming it will.

You guys are more optimistic than me . . . but ok, we'll see how this plays out. I agree that---with respect to most previous instances---advances in technology are slow, and there's a significant barrier to adapt it.

But this feels different. AI has the ability to change, to improve, instantaneously, with very little human effort. It just seems too easy.

I guess the million dollar question is: at what point will AI affect your personal bottom line? (When might it cause a reduction in your RVUs, get paid less per encounter, per procedure, per shift . . . when will jobs become more scarce, making it harder for you to find your next one . . . etc). I give it 10 years.

 

[end stage fibro]

I do a lot of work in the regulatory space in my current role and some of that has to do with AI. For the foreseeable future, the government, the hospital, and lawyers will want someone that is putting their signature behind the AI. To me this will likely go from human-in-the-loop decisions to human-on-the-loop decisions(see 1 below). I think slowly over time as more and more payer contracts get tied to up side and down side risk (2), hospital systems are going to want to reduce non-value added variation in order to raise the floor of care. I think this can be done through protocolized and checkbox medicine with AI as oversight. Having physicians employed just to diagnose the one out of a hundred case is probably not that useful, to be honest, in an economic sense.especially in the inpt setting where care is expensive. I can see a future that is closer to an anesthesia model where one hospitalist will oversee several APPs that are playing hospitalist. Furthermore some specialties are already experiencing certain cuts for efficiency payments(3). I can easily see CMS saying that the average doctor can now see a hundred patients at a time and so a 99233 Will only be worth one RVU instead. that is, more patients + less time = less pay. buckle up

Furthermore I think it is a negative value proposition for physicians as well. My system already allows physicians to use the Epic AI patient summary feature. What does this mean in practice? I get tons of AI-slop discharge summaries that should be shortened down by 80%. They occasionally have hallucinations as well, though I have not uncovered anything major. Docs already don't read their own notes...you think theyre gonna read some AI DC slop?

Finally I do want to comment on what I think is true but may just be a personal bias. I am a huge believer in an adapt or die mentality. With upcoming issues like the Medicaid cliff, CMS trust fund insolvency, or God knows what else will come up, it is clear to me that physicians are going to have to have more skin in the game. We got into this quagmire by not having physician representation and instead allowing MBAs and other suits to make decisions for us. I think horse is out of the barn and now we are in the position of taking up all the risk but only a fraction of the spoils. On the hospitalist subreddit I see all the time "Don't take that job. Know your worth." Au contraire mon ami... you need to prove your worth. remaining old school and just ordering lasix 20 mg IV bId for a HF patient that is going to stay in the hospital for 5 days does not place you in a position of power.

1) "humans in the loop" refers to systems that require direct human involvement in making or approving decisions. "humans on the loop" approach positions humans as supervisors of automated systems. Instead of being directly involved in every decision, humans monitor processes and intervene only when necessary—typically when anomalies or edge cases arise.

2) this is happening more and more as we speak. I'm not sure how much the average hospitalist knows...Medicare Advantage is getting rolled into original Medicare risk contracts. Measuring period has already started and mid-2027 is when payment structures will go into effect. I've heard of similar risk structures being developed for Medicaid if that is even around in a few years. Inpatient only list is going away- that will be a hit to the bottom line too. Medicare is using AI too to eval claims and it sucks. rampant concerns regarding underpayment.

3) CMS Modernizes Payment Accuracy and Significantly Cuts Spending Waste | CMS

 

[DrMetal]

I do a lot of work on the regulatory space in my current role and some of that has to do with AI for the foreseeable future but the government, the hospital, and lawyers will want someone that is putting their signature behind the AI. To me this will likely go from human-in-the-loop decisions to human-on-the-loop decisions1. I think slowly over time as more and more payer contracts get tied to up side and down side risk 2, Hospital systems are going to want to Reduce non-value added variation in order to raise the floor of care. I think this can be done through protocolized and checkbox medicine with AI as oversight. Having physicians employed just to capture the one out of a hundred case is probably not that useful, to be honest, in an economic sense. I can see a future that is closer to an anesthesia model where one hospitalist will oversee several APPs that are playing hospitalist. Furthermore some specialties are already experiencing certain cuts for efficiency payments3. I can easily see CMS saying that the average doctor can now see a hundred patients at a time and so a 99233 Will only be worth one RVU instead. that is, more patients + less time = less pay. buckle up

Furthermore I think it is a negative value proposition for physicians as well. My system already allows physicians to use the Epic AI patient summary feature. What does this mean in practice? I get tons of AI-slob discharge summaries that should be shortened down by 80%. They occasionally have hallucinations as well, though I have not uncovered anything major. Docs already don't read their own notes...you think theyre gonna read some AI DC slop?

1) "humans in the loop" refers to systems that require direct human involvement in making or approving decisions. humans on the loop" approach positions humans as supervisors of automated systems. Instead of being directly involved in every decision, humans monitor processes and intervene only when necessary—typically when anomalies or edge cases arise.

2) this is happening more and more as we speak. I'm not sure how much the average hospitalist knows...Medicare Advantage is getting rolled into original Medicare risk contracts. Measuring period has already started and mid-2027 is when payment structures will go into effect. I've heard of similar risk structures being developed for Medicaid if that is even around in a few years. Inpatient only list is going away- that will be a hit to the bottom line too. Medicare is using AI too to eval claims and it sucks. rampant concerns regarding underpayment.

3) CMS Modernizes Payment Accuracy and Significantly Cuts Spending Waste

You hit the nail on the head. Question is, how fast?

 

[dozitgetchahi]

I do a lot of work in the regulatory space in my current role and some of that has to do with AI. For the foreseeable future, the government, the hospital, and lawyers will want someone that is putting their signature behind the AI. To me this will likely go from human-in-the-loop decisions to human-on-the-loop decisions(see 1 below). I think slowly over time as more and more payer contracts get tied to up side and down side risk (2), hospital systems are going to want to reduce non-value added variation in order to raise the floor of care. I think this can be done through protocolized and checkbox medicine with AI as oversight. Having physicians employed just to diagnose the one out of a hundred case is probably not that useful, to be honest, in an economic sense.especially in the inpt setting where care is expensive. I can see a future that is closer to an anesthesia model where one hospitalist will oversee several APPs that are playing hospitalist. Furthermore some specialties are already experiencing certain cuts for efficiency payments(3). I can easily see CMS saying that the average doctor can now see a hundred patients at a time and so a 99233 Will only be worth one RVU instead. that is, more patients + less time = less pay. buckle up

Furthermore I think it is a negative value proposition for physicians as well. My system already allows physicians to use the Epic AI patient summary feature. What does this mean in practice? I get tons of AI-slop discharge summaries that should be shortened down by 80%. They occasionally have hallucinations as well, though I have not uncovered anything major. Docs already don't read their own notes...you think theyre gonna read some AI DC slop?

Finally I do want to comment on what I think is true but may just be a personal bias. I am a huge believer in an adapt or die mentality. With upcoming issues like the Medicaid cliff, CMS trust fund insolvency, or God knows what else will come up, it is clear to me that physicians are going to have to have more skin in the game. We got into this quagmire by not having physician representation and instead allowing MBAs and other suits to make decisions for us. I think horse is out of the barn and now we are in the position of taking up all the risk but only a fraction of the spoils. On the hospitalist subreddit I see all the time "Don't take that job. Know your worth." Au contraire mon ami... you need to prove your worth. remaining old school and just ordering lasix 20 mg IV bId for a HF patient that is going to stay in the hospital for 5 days does not place you in a position of power.

1) "humans in the loop" refers to systems that require direct human involvement in making or approving decisions. "humans on the loop" approach positions humans as supervisors of automated systems. Instead of being directly involved in every decision, humans monitor processes and intervene only when necessary—typically when anomalies or edge cases arise.

2) this is happening more and more as we speak. I'm not sure how much the average hospitalist knows...Medicare Advantage is getting rolled into original Medicare risk contracts. Measuring period has already started and mid-2027 is when payment structures will go into effect. I've heard of similar risk structures being developed for Medicaid if that is even around in a few years. Inpatient only list is going away- that will be a hit to the bottom line too. Medicare is using AI too to eval claims and it sucks. rampant concerns regarding underpayment.

3) CMS Modernizes Payment Accuracy and Significantly Cuts Spending Waste | CMS

So…here are some questions and also basic spitballing ideas I have with regards to medical AI:

- How are we going to have AI increase patient throughput? The “thinking” part isn’t what slows me down as a physician, it’s, y’know, little “inconveniences” like actually talking to the pts and eliciting a proper history from them. The part nobody seems to be able to admit: AI isn’t actually going to be able to do this any faster than a doctor can. Unless the AI can literally read someone’s mind from a distance, it’s not going to be able to squeeze a history out from a pt *that* much faster than I can. You can’t make a pt talk any faster than they’re willing to do. (Especially when it has to deal with rambling, demented, circumstantial, personality disordered people who are anything but direct.)

- So are doctors going to be supervising fleets of AI chatbots…which are asking pts questions with digitized voices in exam rooms? Wait, what? We’re a ways off from AI being able to do this with any decent effectiveness. And for the AI to “increase throughput”, how long are these visits going to be? 5 minutes? Or is everything going to happen through a text based chatbot? How can humans be “on the loop” when the fundamental aspects of a doctors appointment are still well outside the grasp of a typical AI setup?

- (Never mind procedural specialties, where AI is even further away from being able to replace humans).

- Also, while there are pts who may be OK with AI visits…there are a fair few who won’t be thrilled with 5 minute visits with a computer that sounds like Max Headroom. I predict that there will be a new cash only market (sort of like what DPC is now) for those pts who want a “classic” physician experience. Some doctors will probably be able to make a nice niche for themselves.

- Of course CMS is trying to say they’re doing this with AI…CMS has been trying to en****tify hospital medicine and primary care for many years under the guise that “it saves money”. This brain dead reasoning has led to *****ic outcomes like allowing midlevels to take the wheel in primary care, which almost certainly wastes more money as said midlevels send out blizzards of nonsense consults and order boatloads of labs that they don’t even understand or know how to interpret…

Bottom line: these technological “transformations” rarely happen as fast as their overlords suspect or wish. They told us in the 1980s that we’d have paperless offices. This still hasn’t happened. Elon Musk said you’d be able to get in a Tesla and have it drive you from NYC to LA without you doing a damn thing…in 2018. That still hasn’t happened. I asked ChatGPT some rheumatology questions today and got really dumb answers. Lots of people out there don’t like AI and don’t want it touching everything they do and work with. These types of people certainly aren’t going to want to go to a doctors appointment where they talk to a screen while a physician sits with his thumb up his ass in the next room “supervising”. These people are going to want live visits with a physician, and many of them will be willing to pay for it.

As described above, the process is probably going to be much more gradual. But I suspect that even at 10 years from now, changes will have had less of an impact than many suspect. We’ll probably see articles about how the “promise of AI” hasn’t quite panned out as planned, it’s more complicated than everyone expected, surprise surprise. There is even evidence that AI companies are very overvalued, and that we may be sitting on a dot com style crash (except maybe worse) in the next couple of years.

 

[chessknt]

I do a lot of work in the regulatory space in my current role and some of that has to do with AI. For the foreseeable future, the government, the hospital, and lawyers will want someone that is putting their signature behind the AI. To me this will likely go from human-in-the-loop decisions to human-on-the-loop decisions(see 1 below). I think slowly over time as more and more payer contracts get tied to up side and down side risk (2), hospital systems are going to want to reduce non-value added variation in order to raise the floor of care. I think this can be done through protocolized and checkbox medicine with AI as oversight. Having physicians employed just to diagnose the one out of a hundred case is probably not that useful, to be honest, in an economic sense.especially in the inpt setting where care is expensive. I can see a future that is closer to an anesthesia model where one hospitalist will oversee several APPs that are playing hospitalist. Furthermore some specialties are already experiencing certain cuts for efficiency payments(3). I can easily see CMS saying that the average doctor can now see a hundred patients at a time and so a 99233 Will only be worth one RVU instead. that is, more patients + less time = less pay. buckle up

Furthermore I think it is a negative value proposition for physicians as well. My system already allows physicians to use the Epic AI patient summary feature. What does this mean in practice? I get tons of AI-slop discharge summaries that should be shortened down by 80%. They occasionally have hallucinations as well, though I have not uncovered anything major. Docs already don't read their own notes...you think theyre gonna read some AI DC slop?

Finally I do want to comment on what I think is true but may just be a personal bias. I am a huge believer in an adapt or die mentality. With upcoming issues like the Medicaid cliff, CMS trust fund insolvency, or God knows what else will come up, it is clear to me that physicians are going to have to have more skin in the game. We got into this quagmire by not having physician representation and instead allowing MBAs and other suits to make decisions for us. I think horse is out of the barn and now we are in the position of taking up all the risk but only a fraction of the spoils. On the hospitalist subreddit I see all the time "Don't take that job. Know your worth." Au contraire mon ami... you need to prove your worth. remaining old school and just ordering lasix 20 mg IV bId for a HF patient that is going to stay in the hospital for 5 days does not place you in a position of power.

1) "humans in the loop" refers to systems that require direct human involvement in making or approving decisions. "humans on the loop" approach positions humans as supervisors of automated systems. Instead of being directly involved in every decision, humans monitor processes and intervene only when necessary—typically when anomalies or edge cases arise.

2) this is happening more and more as we speak. I'm not sure how much the average hospitalist knows...Medicare Advantage is getting rolled into original Medicare risk contracts. Measuring period has already started and mid-2027 is when payment structures will go into effect. I've heard of similar risk structures being developed for Medicaid if that is even around in a few years. Inpatient only list is going away- that will be a hit to the bottom line too. Medicare is using AI too to eval claims and it sucks. rampant concerns regarding underpayment.

3) CMS Modernizes Payment Accuracy and Significantly Cuts Spending Waste | CMS

I dont think this is adapt or die--the proto-fascist tech weirdos behind the AI revolution who think that death doesn't exist for them should not be anywhere near healthcare and if the country can ever correct from its tailspin I have faith that it will be stopped. If tech takes over healthcare physician unemployment will be the least of our problems.

As for #2--the easiest solution to this that we have seen is to just get AI to fight the denials. AI fighting itself to waste even more resources and burn the Earth to the ground with pointless energy waste seems in perfect alignment with our values as a nation.

 

[dozitgetchahi]

I dont think this is adapt or die--the proto-fascist tech weirdos behind the AI revolution who think that death doesn't exist for them should not be anywhere near healthcare and if the country can ever correct from its tailspin I have faith that it will be stopped. If tech takes over healthcare physician unemployment will be the least of our problems.

As for #2--the easiest solution to this that we have seen is to just get AI to fight the denials. AI fighting itself to waste even more resources and burn the Earth to the ground with pointless energy waste seems in perfect alignment with our values as a nation.

Ever find it interesting that amidst global warming etc - where we should be conserving resources as much as possible - instead we’re building data centers to chug even more energy and water so as to fund the greed of the tech bros?

I agree, unleashing the AI racket against itself seems to be perfectly ironic.

And yes, letting the tech oligarchs take over the country (world?) is a fairly terrifying proposition. Because if AI eventually becomes as capable as they suggest with no real guardrails, the future is likely to look more like Terminator 2 [or insert your other favorite sci fi dystopia] than Star Trek…

 

[DrMetal]

And yes, letting the tech oligarchs take over the country (world?) is a fairly terrifying proposition. Because if AI eventually becomes as capable as they suggest with no real guardrails, the future is likely to look more like Terminator 2 [or insert your other favorite sci fi dystopia] than Star Trek…

It's pretty clear that mankind is on its downward trajectory. When a species is smart enough to create technology that can replace itself, then the countdown to extinction begins. And quite frankly, there's nothing wrong with that. All species must come to an end, including the human race. Sad that we only lasted 3-4 million years considering the dinosaurs lasted some 50 million years (only to be extinguished by a meteor).

Anyway, have a great Memorial Day weekend!

 

[sloh]

As for #2--the easiest solution to this that we have seen is to just get AI to fight the denials. AI fighting itself to waste even more resources and burn the Earth to the ground with pointless energy waste seems in perfect alignment with our values as a nation.

The "AI vs. AI" arms race in healthcare billing is inherently asymmetrical, and it fundamentally punishes the provider while rewarding the payer.

There’s a vast difference in the effort required to issue a denial versus the effort required to reverse one.

Fighting AI denials consumes massive amounts of uncompensated time. Even if you utilize a generative AI tool to draft a robust appeal letter, a licensed physician still has to review the generated text, ensure the clinical logic is sound, and assume the ultimate legal and professional liability for the submission.

 

[sloh]

@end stage fibro

With the IPO list shrinking, the burden of proof for an inpatient admission has shifted from the procedure itself to the physician's clinical documentation. Hospitalists are increasingly tasked with defending the medical necessity of an admission against strict Medicare Advantage auditing algorithms. This creates a demand for "Physician Advisor" roles, where hospitalists are embedded in pre-op risk clinics to build airtight, documented justifications for why a patient cannot be safely managed in an outpatient setting. 😎

 

[chessknt]

The "AI vs. AI" arms race in healthcare billing is inherently asymmetrical, and it fundamentally punishes the provider while rewarding the payer.

There’s a vast difference in the effort required to issue a denial versus the effort required to reverse one.

Fighting AI denials consumes massive amounts of uncompensated time. Even if you utilize a generative AI tool to draft a robust appeal letter, a licensed physician still has to review the generated text, ensure the clinical logic is sound, and assume the ultimate legal and professional liability for the submission.

It is also incumbent upon the insurance company to provide a similar response so it does waste their resources as well. Fighting down coding is a different story since that doesn't require clinical judgment but simply pointing out that billing rules were followed and that the denial is inappropriate.

You do have to wonder if this will help accelerate the collapse of the us healthcare system since it cannot possibly be worse in any other nation on earth.

 

[sloh]

It is also incumbent upon the insurance company to provide a similar response so it does waste their resources as well. Fighting down coding is a different story since that doesn't require clinical judgment but simply pointing out that billing rules were followed and that the denial is inappropriate.

You do have to wonder if this will help accelerate the collapse of the us healthcare system since it cannot possibly be worse in any other nation on earth.

The insurance companies, Medicare, and Medicaid simply contract these services out to vendors out now. And the “rules” are anything but clear.

“CMS-0057 does not require payers to disclose the underlying UM criteria vendor or the specific criteria version applied to a given denial. It does not require vendors like InterQual or MCG to make their criteria publicly inspectable. It does not address contingent-fee economics for claim editing. It does not address the vertical integration question of one entity owning the criteria, the edit engine, and the payer. The rule treats the payer as the decisionmaker and asks the payer to be faster and more transparent about its decisions, which is fine and useful, but leaves the rule-writing layer alone.”

The Hidden Rule Makers Behind Prior Auth and Claim Denials: How InterQual, MCG, HealthEdge, Zelis, Lyric, and Optum’s CES Write the Criteria Payers Use to Reject Care and Why That’s Where Real Reform

Thoughts on Healthcare Markets & Technology is a reader-supported publication.

open.substack.com

 

[end stage fibro]

sloh did you become a PA? lol. taking words out of my mouth. it's bleak out there, folks

i have responses i'll try to get back to everyone today or tomorrow

 

[sloh]

sloh did you become a PA? lol. taking words out of my mouth. it's bleak out there, folks

i have responses i'll try to get back to everyone today or tomorrow

With the way medicine is moving (unfavorably for us), I would like to become a PA!!!

 

[chessknt]

The insurance companies, Medicare, and Medicaid simply contract these services out to vendors out now. And the “rules” are anything but clear.

“CMS-0057 does not require payers to disclose the underlying UM criteria vendor or the specific criteria version applied to a given denial. It does not require vendors like InterQual or MCG to make their criteria publicly inspectable. It does not address contingent-fee economics for claim editing. It does not address the vertical integration question of one entity owning the criteria, the edit engine, and the payer. The rule treats the payer as the decisionmaker and asks the payer to be faster and more transparent about its decisions, which is fine and useful, but leaves the rule-writing layer alone.”

The Hidden Rule Makers Behind Prior Auth and Claim Denials: How InterQual, MCG, HealthEdge, Zelis, Lyric, and Optum’s CES Write the Criteria Payers Use to Reject Care and Why That’s Where Real Reform

Thoughts on Healthcare Markets & Technology is a reader-supported publication.

open.substack.com

Indeed that is part of a massive problem--the payers have no consequences for blanket denials. All of this of course is fixable but not on our current political trajectory where the tech Epstein class rules the country and that is going to have much bigger problems at scale than our employability.

I also don't really see what the adaptation to make is here. You use the tech and help make it smarter? You learn to become a programmer? You learn to trust the tech so much you can see 100 people a day and just not review the notes? How is another generation of physicians supposed to learn anything in that environment?

This isn't like using email or snail mail this has massive safety and legal implications and is being beta tested on all of us.

 

[sloh]

Indeed that is part of a massive problem--the payers have no consequences for blanket denials. All of this of course is fixable but not on our current political trajectory where the tech Epstein class rules the country and that is going to have much bigger problems at scale than our employability.

I also don't really see what the adaptation to make is here. You use the tech and help make it smarter? You learn to become a programmer? You learn to trust the tech so much you can see 100 people a day and just not review the notes? How is another eg eration of physicians supposed to learn anything in that environment?

This isn't like using email or snail mail this has massive safety and legal implications and is being beta tested on all of us.

The questions about how to adapt—becoming a programmer, trusting tech implicitly, or seeing 100 people a day— just illustrate the impossible position physicians will increasingly be put in. This environment forces a shift from practicing medicine to practicing documentation defense. The mental bandwidth required to preemptively satisfy an unseen "edit engine" takes away from patient care and creates an environment where the next generation of physicians is trained to treat the algorithm rather than the patient.

It is a profound structural problem when the entities defining the standard of care are insulated from the clinical consequences of those standards.

I don't really have any answers either. 😢

 

[sloh]

Coincidentally, just stumbled on a reddit thread about a hospitalist talking about his work environment today:

"I joined a job post residency about 4 years back. During that time in the job market the usual census was around 16-18 and more than 20 seemed kind of rare and undoable. Due to various concerns like increasing census and administrative burden I am exploring for the job opportunities and now it seems like census around 20 is a norm. Even in places with census around 16-17 there seems something tied to it like nights, supervising APPs. And many places have APPs tied to physicians these days. They advertise as no admissions and in reality there is some APP admitting patient under your lisence. Some places boast on round and go but they have census of 25-28 and how it is possible to round and go when you have more than 25 patients. On top of that corporate medicine has been increasing a lot of burden on health care workers and physicians with different metrics before 9, LOS, press ganey and so on. While talking with seasoned Hospitalist it looks like patients are getting more and more demanding. Might be there are some unicorn jobs but not sure what will it looks like further few years down the road. There are few good weeks but after finishing most of the weeks, it feels like coming out of war."

Response:

"Everyone will see higher volumes of patients, stagnant and declining pay (not keeping up with inflation plus getting paid less per patient due to higher volume). Admin will justify this by saying you can use AI to see patients faster while you shoulder more liability and more midlevel encroachment.

Also get ready to be constantly challenged by patients and family that their chatgpt says you are not doing your job right."

Response:

I am so glad I retired early to enjoy life in my thirties and got out in January!"

"5-6 years is not a huge time frame
It took the profession a decade to move from model of round & go ~> staying in house for 8+ hours

Next step with market saturation will possibly be standardization of rounding models with or without APPs

In another 5-6 years with increasing cuts we will see plateauing of salaries with increasing pockets of APP only programs
There probably will be a mandatory hospitalist fellowship for IM & APPs at that point.

Overall the profession will survive - the question how do we make it valuable"

 

[dozitgetchahi]

Coincidentally, just stumbled on a reddit thread about a hospitalist talking about his work environment today:


Response:



Response:

The essence of en****tification.

 

[sloh]

Whoah look what Open Evidence just did today:



"Until now, physicians using AI in clinic had to assemble the patient’s context themselves. Allergies, comorbidities, medications, prior procedures, copy-pasted in from the chart.Today we’re announcing a partnership with
@CedarsSinai

. OpenEvidence now works directly inside Epic, drawing on the patient’s full record and interpreting the medical literature through the lens of that specific patient.Cedars-Sinai is the first academic health system to deploy patient-aware clinical intelligence at enterprise scale. The clinician asks a complex question in natural language. The answer reflects both the best available evidence and the patient in front of them.Patient data is never stored after the clinical session or used for any other purpose."



"First, the gist of what was announced:→ Clinicians can ask questions in natural language and receive answers grounded in both research guidelines AND the specific patient in front of them (the holy grail of CDS AI!)→ Patient context includes prior procedures, comorbidities, medications, allergies, and longitudinal health data - not just a summary snapshot→ The system operates agentically: it interprets the clinical question, dynamically pulls relevant EHR data, evaluates current literature, and synthesizes a context-aware answer"

 

[dozitgetchahi]

With the way medicine is moving (unfavorably for us), I would like to become a PA!!!

Well, I’d like to become a PA too - but only if the amount of my remaining student loans are slashed to reflect my lower pay.

 

[dozitgetchahi]

Whoah look what Open Evidence just did today:



"Until now, physicians using AI in clinic had to assemble the patient’s context themselves. Allergies, comorbidities, medications, prior procedures, copy-pasted in from the chart.Today we’re announcing a partnership with
@CedarsSinai

. OpenEvidence now works directly inside Epic, drawing on the patient’s full record and interpreting the medical literature through the lens of that specific patient.Cedars-Sinai is the first academic health system to deploy patient-aware clinical intelligence at enterprise scale. The clinician asks a complex question in natural language. The answer reflects both the best available evidence and the patient in front of them.Patient data is never stored after the clinical session or used for any other purpose."



"First, the gist of what was announced:→ Clinicians can ask questions in natural language and receive answers grounded in both research guidelines AND the specific patient in front of them (the holy grail of CDS AI!)→ Patient context includes prior procedures, comorbidities, medications, allergies, and longitudinal health data - not just a summary snapshot→ The system operates agentically: it interprets the clinical question, dynamically pulls relevant EHR data, evaluates current literature, and synthesizes a context-aware answer"

This should be amusing. Every chart I’ve ever seen in Epic is laden with garbage information - drugs the pt hasn’t been taking in years, illnesses that either never happened or happened years ago and are resolved, etc etc. Just wait until AI starts “synthesizing” garbage in/garbage out answers…

 

[DrMetal]

Whoah look what Open Evidence just did today:



"Until now, physicians using AI in clinic had to assemble the patient’s context themselves. Allergies, comorbidities, medications, prior procedures, copy-pasted in from the chart.Today we’re announcing a partnership with
@CedarsSinai

. OpenEvidence now works directly inside Epic, drawing on the patient’s full record and interpreting the medical literature through the lens of that specific patient.Cedars-Sinai is the first academic health system to deploy patient-aware clinical intelligence at enterprise scale. The clinician asks a complex question in natural language. The answer reflects both the best available evidence and the patient in front of them.Patient data is never stored after the clinical session or used for any other purpose."



"First, the gist of what was announced:→ Clinicians can ask questions in natural language and receive answers grounded in both research guidelines AND the specific patient in front of them (the holy grail of CDS AI!)→ Patient context includes prior procedures, comorbidities, medications, allergies, and longitudinal health data - not just a summary snapshot→ The system operates agentically: it interprets the clinical question, dynamically pulls relevant EHR data, evaluates current literature, and synthesizes a context-aware answer"

It's happening and changing so quickly, right in front of our eyes. It took 10 years for most hospitals to adopt UptoDate, to put it in or near the EMR. Look at how fast they're adopting AI (in a fraction of that time).

I'm a pessimist by nature (shocker, I know) . . . but even from an objective lens, you have to see the writing on the wall here.

Ours is (of course) not the only profession seeing this change. My lawyer buddies are also very concerned!

 

[end stage fibro]

So…here are some questions and also basic spitballing ideas I have with regards to medical AI:

- How are we going to have AI increase patient throughput? The “thinking” part isn’t what slows me down as a physician, it’s, y’know, little “inconveniences” like actually talking to the pts and eliciting a proper history from them. The part nobody seems to be able to admit: AI isn’t actually going to be able to do this any faster than a doctor can. Unless the AI can literally read someone’s mind from a distance, it’s not going to be able to squeeze a history out from a pt *that* much faster than I can. You can’t make a pt talk any faster than they’re willing to do. (Especially when it has to deal with rambling, demented, circumstantial, personality disordered people who are anything but direct.)

- So are doctors going to be supervising fleets of AI chatbots…which are asking pts questions with digitized voices in exam rooms? Wait, what? We’re a ways off from AI being able to do this with any decent effectiveness. And for the AI to “increase throughput”, how long are these visits going to be? 5 minutes? Or is everything going to happen through a text based chatbot? How can humans be “on the loop” when the fundamental aspects of a doctors appointment are still well outside the grasp of a typical AI setup?

- (Never mind procedural specialties, where AI is even further away from being able to replace humans).

- Also, while there are pts who may be OK with AI visits…there are a fair few who won’t be thrilled with 5 minute visits with a computer that sounds like Max Headroom. I predict that there will be a new cash only market (sort of like what DPC is now) for those pts who want a “classic” physician experience. Some doctors will probably be able to make a nice niche for themselves.

- Of course CMS is trying to say they’re doing this with AI…CMS has been trying to en****tify hospital medicine and primary care for many years under the guise that “it saves money”. This brain dead reasoning has led to *****ic outcomes like allowing midlevels to take the wheel in primary care, which almost certainly wastes more money as said midlevels send out blizzards of nonsense consults and order boatloads of labs that they don’t even understand or know how to interpret…

Bottom line: these technological “transformations” rarely happen as fast as their overlords suspect or wish. They told us in the 1980s that we’d have paperless offices. This still hasn’t happened. Elon Musk said you’d be able to get in a Tesla and have it drive you from NYC to LA without you doing a damn thing…in 2018. That still hasn’t happened. I asked ChatGPT some rheumatology questions today and got really dumb answers. Lots of people out there don’t like AI and don’t want it touching everything they do and work with. These types of people certainly aren’t going to want to go to a doctors appointment where they talk to a screen while a physician sits with his thumb up his ass in the next room “supervising”. These people are going to want live visits with a physician, and many of them will be willing to pay for it.

As described above, the process is probably going to be much more gradual. But I suspect that even at 10 years from now, changes will have had less of an impact than many suspect. We’ll probably see articles about how the “promise of AI” hasn’t quite panned out as planned, it’s more complicated than everyone expected, surprise surprise. There is even evidence that AI companies are very overvalued, and that we may be sitting on a dot com style crash (except maybe worse) in the next couple of years.

you know, a friend of mine got onto telehealth. I was initially worried but the process has actually been pretty good. There is a form that is WAY more thorough than I would ask in clinic, I go over it, and OK the rx asynchronously. this is obviously a cherry picked patient population. rheum i think is a little bit of a black box in IM and this process might not work for you cus youre actually touching the patient. But even then I could 15 min appointment slots getting wittled down to 10 mins because "the patient filled out a form and you have AI now." you may counter, that is bad medicine. ok, so what? the horse has left the bad medicine barn years ago.

furthermore, i haven't been in a rheumatologist's office since med school. i imagine there are many any leather-bound books and it smells of rich mahogany. but i do know the inside of the hospital too damn well, and within its bowels the actual patient encounter is just an ad so that i can play medicine, and playing medicine can certainly be speedran with AI tool assistance . MDs in my 5+ hospital system are literally waiting in line for ambient scribes and can all already use AI DC summary features. we are in talks with a vendor for discharge med rec oversight (isn't that what an MD is for?). we used to have interdisciplinary rounds, but now we have an Epic dashboard that prepopulates an estimated discharge day based on the working DRG and the MD just rubber stamps it. ive siad before but the hospitalists job continues to get wittled down. how longuntil it is wittled down to nothing?


the rest of your post i agree with. CMS talks about all this quality garbage but its a thinly veiled guise to save money, and they make the hospital systems figure it out. LEAD, WISer, CPC initiaitve...all stupid af. but saves money

i agree with procedural specialties being safer for the time being. physical AI seems to be all the rage in china, here not so much. though nurse surgeons are a thing now i guess?

i agree that the process will be more gradual than in other fields, but when it comes to saving money i would not underestimate the medical industry, especially when the savings are shouldered by us

also i would love to get healthcare from this dude:

 

[end stage fibro]

I dont think this is adapt or die--the proto-fascist tech weirdos behind the AI revolution who think that death doesn't exist for them should not be anywhere near healthcare and if the country can ever correct from its tailspin I have faith that it will be stopped. If tech takes over healthcare physician unemployment will be the least of our problems.

As for #2--the easiest solution to this that we have seen is to just get AI to fight the denials. AI fighting itself to waste even more resources and burn the Earth to the ground with pointless energy waste seems in perfect alignment with our values as a nation.

sloh already answered #2. the fight is super asymmetrical.

i maintain that it is indeed adapt or die.for better or for worse, healthcare is getting more corporate by necessity. there is so much overhead just to allow for care delivery, that i dont think a health system operating on good feelings and good medicine can survice. good medicine the floor...you need to maintain margin in some capacity on top of that. and margin comes along with all pesky words docs dont like (even i dont!)....utilization, LOS, relative weight, risk adjustment. blah blah blah

i like tech stuff but i agree tech bros should not be involved in healthcare. i mean really the whole concept of for profit healthcare is nuts to me, yet here we are. we all know "non-profit" is a label only.

 

[end stage fibro]

@end stage fibro

With the IPO list shrinking, the burden of proof for an inpatient admission has shifted from the procedure itself to the physician's clinical documentation. Hospitalists are increasingly tasked with defending the medical necessity of an admission against strict Medicare Advantage auditing algorithms. This creates a demand for "Physician Advisor" roles, where hospitalists are embedded in pre-op risk clinics to build airtight, documented justifications for why a patient cannot be safely managed in an outpatient setting. 😎

dude dont get me started. i wander into the pain management forums sometimes and see them mentioning "oh just use autotext for the NCD language and youre set." idk if its different in the private setting, but the hospital gets crucified for exactly that. it didnt start this way but >50% of my job now is making sure that the words in a physicians document are actually representative of what happened. there is an incongruence an alarmingly high number of times.

i know this not just my system as well. as ive mentioend before i occasionally do consulting work for outside systems and see the same ****. academic centers are especially bad it. Its like some purity test to ignore all of reality and just practice ivory tower medicine

 

[end stage fibro]

The questions about how to adapt—becoming a programmer, trusting tech implicitly, or seeing 100 people a day— just illustrate the impossible position physicians will increasingly be put in. This environment forces a shift from practicing medicine to practicing documentation defense. The mental bandwidth required to preemptively satisfy an unseen "edit engine" takes away from patient care and creates an environment where the next generation of physicians is trained to treat the algorithm rather than the patient.

It is a profound structural problem when the entities defining the standard of care are insulated from the clinical consequences of those standards.

I don't really have any answers either. 😢

super well said. i actually learned a term for that (from AI, ha!) Pigouvian tax

editL i could have sworn it was in this thread but guess not, i am too lazy to look for where it was posted. one of my colleagues was on the IDSA comittee to write some guideline. a patient whipped out chatgpt and said look it says you need to treat my infection with zyvox. my colleage said listen i helped write the guideline on this, chatgpt is wrong. guess who the patient believed....

 

[Hollow Knight]

In response to the "Open-evidence chart summary" thing.....

I'm a baby doctor, don't know a world without EMRs. Every patient I've ever admitted to the hospital had a file with loads of inaccuracies (medications they don't take, allergies and conditions they don't have, etc) in the chart. If the chart itself is incorrect, why would I trust an AI summary of said chart when LLMS hallucinate on top of that? My professors taught us to do an old-school IM-style H and P for every admission.

The MBAs won't care, but I don't know why anyone thinks this is a good idea. Or maybe y'all's patient records are just generally more accurate than mine?

 

[dozitgetchahi]

In response to the "Open-evidence chart summary" thing.....

I'm a baby doctor, don't know a world without EMRs. Every patient I've ever admitted to the hospital had a file with loads of inaccuracies (medications they don't take, allergies and conditions they don't have, etc) in the chart. If the chart itself is incorrect, why would I trust an AI summary of said chart when LLMS hallucinate on top of that? My professors taught us to do an old-school IM-style H and P for every admission.

The MBAs won't care, but I don't know why anyone thinks this is a good idea. Or maybe y'all's patient records are just generally more accurate than mine?

Every EMR I’ve ever used is laden with nonsense just as you describe.

 

[sloh]

i agree with procedural specialties being safer for the time being. physical AI seems to be all the rage in china, here not so much. though nurse surgeons are a thing now i guess?

Not even sure the procedural specialties will stay ahead for very long.

VBC has been the slow-rolling reshuffle for two decades, but the next five years compress its impact dramatically because the mandatory bundled episode framework has finally been formalized. The TEAM model (Transforming Episode Accountability Model), which CMS finalized in late 2024 for January 2026 implementation, is mandatory for selected hospitals and covers five high-cost surgical episodes: lower extremity joint replacement, surgical hip and femur fracture treatment, spinal fusion, coronary artery bypass graft, and major bowel procedure. Each episode bundles thirty days of post-acute care into a single price, with hospitals at risk for cost overruns and rewarded for coming in under target.

The implications for the wealthiest specialties list are direct. Ortho surgeons performing TKA and THA in TEAM-mandatory hospitals will face downstream margin compression as hospitals push for shorter LOS, lower-cost implants, and post-acute care substitution (home health over SNF). The Comprehensive Care for Joint Replacement model (CJR) ran from 2016 to 2024 and demonstrated this pattern conclusively. Now extend that across all major joints, hips, fusions, CABGs, and bowel resections. Cardiac surg and spine surg, both of which sit comfortably in the top tier of the rankings, are about to learn what ortho learned during CJR.

In 5-10 years, specialties flat or compressed: ortho faces TEAM bundle compression. Average comp will probably plateau around current levels in nominal terms, which is real-dollar compression. Cardiothoracic and cardiac surg face TEAM-CABG and TAVR substitution and probably end up flat to mildly compressed. Spine surg faces TEAM fusion bundles and the ongoing migration to outpatient minimally invasive procedures that compress per-case revenue. Gen surg faces TEAM bowel episodes. General cardiology faces AI ECG compression and PFS pressure while interventional and EP cardiology rise; the specialty bifurcates more visibly than today. GI faces a near-term tailwind from USPSTF expansion of colonoscopy screening to age forty five plus, but Cologuard, Shield (Guardant), Freenome, and other non-invasive screening modalities will bend the colonoscopy volume curve eventually. Plus PFS pressure and PE multiple compression in a specialty that has been heavily rolled up. Modest compression long-term. General ophtho faces AI screening compression for diabetic and AMD work while cataract and retinal procedural work remains strong. Mixed. Derm faces general derm AI compression but cosmetic and Mohs strength. Average likely flat, top-quartile rising. by 2032, the gap between the top-paid specialty (probably plastic surg or ortho) and the top-quartile primary care PCP in a risk-bearing platform may compress to 1.3x or less. That is a meaningful narrowing from today’s roughly 2.2x ratio.

 

[DrMetal]

Every EMR I’ve ever used is laden with nonsense just as you describe.

In a sense, AI is the natural evolution/progression of the EMR. The EMR is the database (the data server), now AI can act on it and produce something (a note, a summary, etc). In time, AI will suggest order sets, then implement them. All we'll have to do is sign the chart, and even that requirement may become automated.

Every physician will have to round on (or see in clinic) 50 patients a day. Don't like it? Too much liability? Find something else to do . . .

 

[sloh]

In a sense, AI is the natural evolution/progression of the EMR. The EMR is the database (the data server), now AI can act on it and produce something (a note, a summary, etc). In time, AI will suggest order sets, then implement them. All we'll have to do is sign the chart, and even that requirement may become automated.

Every physician will have to round on (or see in clinic) 50 patients a day. Don't like it? Too much liability? Find something else to do . . .

If insurers successfully use AI-driven efficiencies to argue that "moderate MDM" now takes 30% less effort and slash the RVU values accordingly, the only way to maintain current revenue is to increase patient volume. It becomes a race to the bottom, turning physicians into high-volume chart editors rather than diagnosticians.


AI is an incredible tool for efficiency, but in a fee-for-service model, efficiency is rarely rewarded with higher pay; it's usually punished with lower rates and higher volume expectations.

And I guess to play devils advocate, one can say MDM is where the cognitive labor actually lives. If you are managing high-complexity patients, the patient's intrinsic complexity doesn't magically decrease just because an AI generated the summary. You still have to weigh the risk of morbidity, manage the prescription drugs, and decide on the treatment. Insurers can't easily downgrade the patient's disease burden, even if the charting was fast.

We’ll just see how this plays out but the trends aren’t promising. Make hay while the sun shines is my MO.

 

[chessknt]

Again I fail to see what part of adapt there is here. In this hellscape there is really no need for physicians to even exist outside of liability sponges so what exactly do we adapt to? Just accept that we are a liability sponges and become stupid from ceasing to have to actually do anything? Where do the next generation of physicians come from in this scenario?

 

[DrMetal]

Where do the next generation of physicians come from in this scenario?

They don't. This profession is ending. In 50 years (maybe sooner), the notion of a 'physician' will be obsolete, even comical (in the same way we laugh today about the notion of a town apothecary).

Liability can always be displaced (so I don't think even that can save us). The hospital (or clinic) as a sole entity can assume liability if no physicians are involved. Is that costly? Sure . . . but possibly cheaper than paying a staff of physicians.

 

[chessknt]

They don't. This profession is ending. In 50 years (maybe sooner), the notion of a 'physician' will be obsolete, even comical (in the same way we laugh today about the notion of a town apothecary).

Liability can always be displaced (so I don't think even that can save us). The hospital (or clinic) as a sole entity can assume liability if no physicians are involved. Is that costly? Sure . . . but possibly cheaper than paying a staff of physicians.

I don't think the tech God will be as acceptable as you do. Why not eliminate politicians and administration as well? Why not eliminate all professions outside of the blue collar trades?

 

[DrMetal]

I don't think the tech God will be as acceptable as you do. Why not eliminate politicians and administration as well? Why not eliminate all professions outside of the blue collar trades?

That's all on the table. To save it, you'll have to prove a human can do it better. Good luck with that . . .

The problem with medicine, is that, at its core, its a very 'dumb' profession. It's very algorithmic, very pattern oriented, very simple, to be honest (we're not solving complex differential equations).

And it's practitioners (doctors) are pretty 'dumb' too. We care about dumb things (excessive documentation, coding, rvus, etc). We're too busy worried about MOC, CME, etc. No one's paying attention (except for us, on this thread!)

Thus, medicine is ripe for takeover by AI. It's not even a fair fight.

 

[sloh]

Again I fail to see what part of adapt there is here. In this hellscape there is really no need for physicians to even exist outside of liability sponges so what exactly do we adapt to? Just accept that we are a liability sponges and become stupid from ceasing to have to actually do anything? Where do the next generation of physicians come from in this scenario?

There actually might be a diminishing need for the next generation of physicians, unfortunately.

Health systems are not buying ambient documentation software because they want to give nurses a better experience, even if that is how it is marketed. They are buying it because they are trying to close 150 to 200 basis point operating margin gaps and labor is the biggest lever they have.

The clinical decision support layer is where things get more speculative but also more interesting. Tools like OpenEvidence are already changing how clinicians access evidence at the point of care. If the AI layer can help a nurse practitioner work to the full scope of their license more confidently, you get leverage on physician labor costs. If it helps a specialist see more patients per day by reducing cognitive overhead on routine cases, you get capacity expansion without FTE growth. None of this is about replacing clinicians. It is about making existing clinicians more productive, which in a sector running at negative operating margins at many institutions, is the most urgent financial lever available.

Ambient clinical documentation tools, the Nuance DAX category, the Abridge category, can reduce the documentation burden on a physician or advanced practice provider by 50% or more per patient encounter. That is not theoretical. Those systems are in production at major health systems with published data behind them. A physician spending two hours per day on documentation who gets that back to one hour is not getting laid off. But a health system deploying that tool across 500 physicians is getting a capacity equivalent of 250 physician-hours per day without adding headcount. That is enormous.

 

[chessknt]

There actually might be a diminishing need for the next generation of physicians, unfortunately.

Health systems are not buying ambient documentation software because they want to give nurses a better experience, even if that is how it is marketed. They are buying it because they are trying to close 150 to 200 basis point operating margin gaps and labor is the biggest lever they have.

The clinical decision support layer is where things get more speculative but also more interesting. Tools like OpenEvidence are already changing how clinicians access evidence at the point of care. If the AI layer can help a nurse practitioner work to the full scope of their license more confidently, you get leverage on physician labor costs. If it helps a specialist see more patients per day by reducing cognitive overhead on routine cases, you get capacity expansion without FTE growth. None of this is about replacing clinicians. It is about making existing clinicians more productive, which in a sector running at negative operating margins at many institutions, is the most urgent financial lever available.

mbient clinical documentation tools, the Nuance DAX category, the Abridge category, can reduce the documentation burden on a physician or advanced practice provider by 50% or more per patient encounter. That is not theoretical. Those systems are in production at major health systems with published data behind them. A physician spending two hours per day on documentation who gets that back to one hour is not getting laid off. But a health system deploying that tool across 500 physicians is getting a capacity equivalent of 250 physician-hours per day without adding headcount. That is enormous.

So what is the adaptation we are supposed to make? Just retire and go get taken care of by an AI doctor? Or use the tools and increase productivity by 1000% but get paid the same?

I get that there are a lot of buzzwords like you used to make this seem like an incredible business maneuver (which we all know hospital admins excel at) but I genuinely believe this is destined for failure. This seems like a classic PE trojan that we have seen ad nauseum in the medical field--they start off with enormous promises and low prices to invade and become used everywhere, then drive the price in to the stratosphere and bleed the system to death.

I mean everything else is going to **** in the world so why not destroy healthcare at the same time but I don't really see why physicians are advocating for this here. The proto-fascist tech/PE billionaires that are so detached from reality that they view themselves as beings above humanity being in control of healthcare at an absolute level (ie no doctors to protect patients from business interests or advocate on their behalf) is not a world I want to live in.