Will being a doctor be boring with AI?

[voxveritatisetlucis]

I feel like one of the main reasons that I went into medicine was the ability to solve problems, have a cerebral career etc. Now everybody talks about the future in which AI algorithms are integrated into Epic and automatically come up with the assessment and plan. I’m not even worried about not having a job, but to me this would make medicine extremely boring. I wish it were the 1980s/1990s when doctors would round on their own patients. Is surgery going to be the only intellectually stimulating specialty in the future (aside from academics who can get grants/do research).

---

Members don't see this ad.

 

[Hollow Knight]

No

 

[Tangerine123]

I wish it were the 1980s/1990s when doctors would round on their own patients.

As if we don't do that nowadays?

I don't see such "catastrophic" scenario happening anytime soon. At least not in my field (Psychiatry). Would need input from other fields, but I doubt it's that big of an issue (except for maybe rads and path)

 

[Mr.Smile12]

The EMR changed medicine over a decade ago, and there is much commentary how more distant the doctor-patient relationship has also become. Mid-level encroachment and private equity certainly don't make the health care field boring. It's still a significant piece of the economy and AI is just another disruptive force. Especially in managing health systems.

But challenges still remain with rare diseases or highly complex health management issues. No single gene diseases have been cured yet as far as I know.

 

[dgdg]

We were one of the first with an EMR in our town about 20 years ago, never had a single paper chart. Certainly some wonderful capabilities came along - eRxs, e-lab ordering, messaging, finding charts, offsite access, drug interactions and allergy alerts. I love all those things. Why did it stop there? Maybe if we could afford EPIC, we might have some of that, but we are not some large corporation.

20 years later, we are still there with only fundamental EHR capabilities (EPIC giants aside) as our health care system continues to serve our people less and less. Now, with electronic data I spend a great deal of time with AI prior authorizations and rejections for necessary care. The greatest updates to our EHR has been chasing the MIPS reporting which seems to change just enough every year. I spend inordinate time entering data for quality of care benchmarks that have not been proven to increase quality of care. In fact, our national quality of care benchmarks have declined overall during this time. It remains expensive and out of reach for many.

I see AI in medical news articles more of a marketing hype by private equity entrepreneurs seeking to generate more revenue without regard to solving our actual problems. I want to know how much a drug will cost my patient before I send the prescription. I want easy access decision making support tools for standard screenings and common acute and chronic diagnoses. I want diagnostic tools based on reading my soap note data (custom subjective, exam, labs, and advanced imaging. So many things have not happened over 20 years.

 

[GoSpursGo]

No.

 

[Dral]

We were one of the first with an EMR in our town about 20 years ago, never had a single paper chart. Certainly some wonderful capabilities came along - eRxs, e-lab ordering, messaging, finding charts, offsite access, drug interactions and allergy alerts. I love all those things. Why did it stop there? Maybe if we could afford EPIC, we might have some of that, but we are not some large corporation.

20 years later, we are still there with only fundamental EHR capabilities (EPIC giants aside) as our health care system continues to serve our people less and less. Now, with electronic data I spend a great deal of time with AI prior authorizations and rejections for necessary care. The greatest updates to our EHR has been chasing the MIPS reporting which seems to change just enough every year. I spend inordinate time entering data for quality of care benchmarks that have not been proven to increase quality of care. In fact, our national quality of care benchmarks have declined overall during this time. It remains expensive and out of reach for many.

I see AI in medical news articles more of a marketing hype by private equity entrepreneurs seeking to generate more revenue without regard to solving our actual problems. I want to know how much a drug will cost my patient before I send the prescription. I want easy access decision making support tools for standard screenings and common acute and chronic diagnoses. I want diagnostic tools based on reading my soap note data (custom subjective, exam, labs, and advanced imaging. So many things have not happened over 20 years.

Agree. These concepts are why it's important for us as physicians to take the reigns here and direct how AI helps us, both to make our day to day easier in terms of giving better more effective care to patients and to help decide how AI should fit into healthcare overall. Furthermore, if we take the reigns, we can direct AI as a tool to help us instead of others who might be more apt to develop AI to take over more of our jobs/responsibilities (with likely making more money in mind over other things). Hopefully we can help develop AI to make our jobs better, help patient care, and not replace us vs if we let others do it, overall bottom line in terms of money will likely be deemed more important, with not great results for us (and all the things you mentioned that go with that - first sentence of your last paragraph).

A good precedent for this is medical specialties who take over and determine their own AUC guidelines instead of letting government, insurance, or other larger bodies determine it/them.

 

[Goro]

I feel like one of the main reasons that I went into medicine was the ability to solve problems, have a cerebral career etc. Now everybody talks about the future in which AI algorithms are integrated into Epic and automatically come up with the assessment and plan. I’m not even worried about not having a job, but to me this would make medicine extremely boring. I wish it were the 1980s/1990s when doctors would round on their own patients. Is surgery going to be the only intellectually stimulating specialty in the future (aside from academics who can get grants/do research).

Seeing the quality of AI right now, definitely no.

It will be useful in Heme/Onc, but the computer ain't the one having to deal with a live, scared patient.

 

[Sidus1011]

"There are people in every time and every land who want to stop history in its tracks. They fear the future, mistrust the present, and invoke the security of a comfortable past which, in fact, never existed."

 

[VA Hopeful Dr]

I feel like one of the main reasons that I went into medicine was the ability to solve problems, have a cerebral career etc. Now everybody talks about the future in which AI algorithms are integrated into Epic and automatically come up with the assessment and plan. I’m not even worried about not having a job, but to me this would make medicine extremely boring. I wish it were the 1980s/1990s when doctors would round on their own patients. Is surgery going to be the only intellectually stimulating specialty in the future (aside from academics who can get grants/do research).

Sweet Jesus no.

First, care has improved with hospitalists compared to traditional PCPs who did clinic and inpatient.

Second, lifestyle back then was just awful. I already make less than almost every other specialist out there, now you want me to also work the same/worse hours than many of them? Hard pass.

Third, PCPs are about the only ones who rounded a lot back then who don't know. Cardiology, anything surgical, GI, all still round.

 

[SurfingDoctor]

SurfingDoc: “AI, follow the CPG and write my note whilst I sip on this White Russian”

AI: “You got it boss”

 

[ACSurgeon]

I feel like one of the main reasons that I went into medicine was the ability to solve problems, have a cerebral career etc. Now everybody talks about the future in which AI algorithms are integrated into Epic and automatically come up with the assessment and plan. I’m not even worried about not having a job, but to me this would make medicine extremely boring. I wish it were the 1980s/1990s when doctors would round on their own patients. Is surgery going to be the only intellectually stimulating specialty in the future (aside from academics who can get grants/do research).

Past present and future, yes

 

[guytakingboards]

I use AI semi-regularly in my job as a radiologist and it's niche at best.

 

[SunBakedTrash]

AI will not replace pathologists, only reduce tedious stuff like counting mitoses or screening acellular pap smears. The biggest threat to pathology (or any medical specialty) is crappy leadership.

 

[operaman]

Oh no not at all. You also have to think how AI gets rolled out: models will get developed and then sold to private practices and hospital systems. Hospitals won’t buy it unless they can see some sort of financial benefit. So if an AI scribe means their docs can see an extra hour of patients each day, then that adds up fast. If an AI can handle an Epic inbox and save a staff FTE form doing it, the that’s cost effective too.

I have a hard time seeing how an AI replacing the thinking and counseling aspects of doctoring can be financially viable in the current system. Remember: docs are seen as revenue generators. An AI replacing a doc will never actually be able to bill for that service which means it doesn’t make the hospital any money.

 

[ChordaEpiphany]

I have a hard time seeing how an AI replacing the thinking and counseling aspects of doctoring can be financially viable in the current system. Remember: docs are seen as revenue generators. An AI replacing a doc will never actually be able to bill for that service which means it doesn’t make the hospital any money.

We will almost certainly have AI-aided diagnosis and management built into EPIC within the decade. My intuition is that it will be based on flawed studies from AI companies and hospital administrators believing it means they can a) give midlevels more autonomy, and b) make physicians work faster.

So they will pay some enormous sum of money to have EPIC-GPT that makes you click 4 extra times when you sign your note and put in orders to make sure you thought super, duper carefully about hereditary hemochromatosis. You'll likely also have to click a box that says, "I understand that all plans generated by EPIC-GPT are suggestions and may be flawed... etc."

This style of medicine will probably work really well in specialty clinics w/ patients who just need a bunch of checked boxes before surgery or a procedure. It will probably be a disaster in primary care when the the midlevel, who not only has no real education but also has spent their entire career relying on the GPT, completely misses or misinterprets key physical exam findings or fails to ask relevant questions in the history.

 

[RangerBob]

We will almost certainly have AI-aided diagnosis and management built into EPIC within the decade. My intuition is that it will be based on flawed studies from AI companies and hospital administrators believing it means they can a) give midlevels more autonomy, and b) make physicians work faster.

So they will pay some enormous sum of money to have EPIC-GPT that makes you click 4 extra times when you sign your note and put in orders to make sure you thought super, duper carefully about hereditary hemochromatosis. You'll likely also have to click a box that says, "I understand that all plans generated by EPIC-GPT are suggestions and may be flawed... etc."

This style of medicine will probably work really well in specialty clinics w/ patients who just need a bunch of checked boxes before surgery or a procedure. It will probably be a disaster in primary care when the the midlevel, who not only has no real education but also has spent their entire career relying on the GPT, completely misses or misinterprets key physical exam findings or fails to ask relevant questions in the history.

The "advice" and "suggestions" as well as the warnings I get from our EMR (which is admittedly not Epic) are so laughably poor/inaccurate that >95% of them I ignore. At least 20% of the time it's the system not even realizing the patient was transferred from a different hospital in the same system to us, when when I do a med rec on admission to acute rehab, it flags every med as a duplicate.

So yes, I very much agree we can likely expect to click more keys/buttons without much, if any, meaningful benefit to us or the patients.

The day our jobs disappear due to AI we have much bigger things than our jobs to worry about. When docs' jobs are truly threatened, we can be sure most other white collar workers were threatened already, and there would be mass unrest from the sheer volume of jobs that become redundant/unnecessary.

 

[operaman]

We will almost certainly have AI-aided diagnosis and management built into EPIC within the decade. My intuition is that it will be based on flawed studies from AI companies and hospital administrators believing it means they can a) give midlevels more autonomy, and b) make physicians work faster.

So they will pay some enormous sum of money to have EPIC-GPT that makes you click 4 extra times when you sign your note and put in orders to make sure you thought super, duper carefully about hereditary hemochromatosis. You'll likely also have to click a box that says, "I understand that all plans generated by EPIC-GPT are suggestions and may be flawed... etc."

This style of medicine will probably work really well in specialty clinics w/ patients who just need a bunch of checked boxes before surgery or a procedure. It will probably be a disaster in primary care when the the midlevel, who not only has no real education but also has spent their entire career relying on the GPT, completely misses or misinterprets key physical exam findings or fails to ask relevant questions in the history.

Maybe. I just see it as a pretty hard sell to a cash strapped hospital when that product won’t actually help bring in any additional revenue and especially if the additional time makes their providers even less efficient. Even for midlevels, a level 3 visit and a level 3 visit with AI support pay exactly the same thing.

I just don’t find diagnosis to be remotely challenging at all. Honestly that’s maybe been my biggest disappointment with medicine in general - I thought that would be a much bigger part of my job than it is. Most of the time I know the diagnosis before entering the room and certainly within the first 5 minutes. I don’t need any support on that front.

Now, an AI that reviews my documentation before signing to make sure I’m hitting maximum level billing - that I could use! Something like “I saw you reviewed that CT scan in pacs but you don’t provide your own interpretation in this note. If you do, this would qualify as a level 4 rather than a level 3. Click here to add to your note.” And then based on my note, it drafts a sample blurb to insert that pre populates my findings.

I just don’t think anyone wants diagnosis/treatment decision support, it will only slow us down, and it won’t make anyone more money.

Scribes, inbox mgmt, chart review, coding/billing support - those are areas where it will pay for itself fast. Remember that most hospital execs are bonused on fairly short term goals, so some big purchase that may take years to pay for itself is going to be a tough sell upstairs unless all the docs are pushing for it and threatening to walk if they don’t get it.

 

[ProReduction]

I feel like one of the main reasons that I went into medicine was the ability to solve problems, have a cerebral career etc. Now everybody talks about the future in which AI algorithms are integrated into Epic and automatically come up with the assessment and plan. I’m not even worried about not having a job, but to me this would make medicine extremely boring. I wish it were the 1980s/1990s when doctors would round on their own patients. Is surgery going to be the only intellectually stimulating specialty in the future (aside from academics who can get grants/do research).

"Boring" is certainly very relative. In 20 years will AI supplement or replace much of what physicians currently do? Absolutely. Will this cause massive shifts in the physician job market? Absolutely. Is any of this unique to physicians? No. Lawyers, delivery personnel, Uber drivers, Actuaries, Writers, Actors, Artists . . . you get the point . . . will all see the same disruptions. What I find to be unique to physicians is how irreplaceable we think we are.

 

[collegestud2013]

AI will not replace pathologists, only reduce tedious stuff like counting mitoses or screening acellular pap smears. The biggest threat to pathology (or any medical specialty) is crappy leadership.

Oh no not at all. You also have to think how AI gets rolled out: models will get developed and then sold to private practices and hospital systems. Hospitals won’t buy it unless they can see some sort of financial benefit. So if an AI scribe means their docs can see an extra hour of patients each day, then that adds up fast. If an AI can handle an Epic inbox and save a staff FTE form doing it, the that’s cost effective too.

I have a hard time seeing how an AI replacing the thinking and counseling aspects of doctoring can be financially viable in the current system. Remember: docs are seen as revenue generators. An AI replacing a doc will never actually be able to bill for that service which means it doesn’t make the hospital any money.

For image-heavy specialties like pathology or radiology, a common prediction on the effect of AI is not that it would be make replace pathologists or radiologists overnight or make them obsolete, but the more likely scenario is that it will make them more efficient so that less are needed (as hospital systems are always trying to cut costs to improve the bottom line). The ones still with a job will have to read higher volumes to make the same amount of money, (probably from a combination of CMS reimbursement cuts as well as the costs of the AI system itself that will need to be offset), though AI will be assisting in their reading. Radiology is a good example of this from the past; in the past there have been steep CMS reimbursement costs over the last few decades corresponding to advances in technology that have allowed for increased efficiency of reading images, and nowadays that volumes radiologists must read to make the same amount of money (adjusted for inflation) are much higher than 20-30 years ago.

At least with radiology the job market is good at the moment because AI hasn't taken over just yet, while imaging volumes are higher than ever (mostly because clinicians nowadays don't bother doing thorough physical exam and also want to CYA from medicolegal liability). However, even CMS thinks imaging is overused and going after it and trying to cut down on imaging use (for example mandating hospital EMRs to have block inappropriate imaging indications in certain settings or at least require the ordering provider to read through some prompts and override the order).

Pathology has never had a good job market in the past few decades, with new graduates usually having to do one or sometimes even two fellowships, and even then would have to geographically flexible to get a new job. This is why it's not competitive residency to get it despite having a good lifestyle with some of the lowest burnout rates, and reasonably good pay for the amount of work (pay usually in between primary care fields and higher paying surgical specialties). I would suspect that with AI being incorporated into pathology, the job market will only get worse.

I suspect AI will have less impact in reducing demand for physicians in patient-facing non-procedural clinical fields initially, as a good part of those fields is communicating with patients/families which AI cannot to well (and most patients won't want either); it can help automate some of the easier clinical-decision making so clinicians are less likely to miss things, and it may be able to make some of the tedious scut-like work more efficient and there's already a fair amount of implementation for it (eg helping with charting/writing notes, drafting letters to patients/insurances, etc...). For patient-facing fields the bigger threat to physician jobs is the rapid growth of midlevels, which is seen by many as the main solution for the impeding physician shortage in some fields (usually fields with high burnout rates, such that younger physicians who are graduating are on average providing less hours of clinical care than the older ones retiring from the same field).

 

[cj_cregg]

Sweet Jesus no.

First, care has improved with hospitalists compared to traditional PCPs who did clinic and inpatient.

Second, lifestyle back then was just awful. I already make less than almost every other specialist out there, now you want me to also work the same/worse hours than many of them? Hard pass.

Third, PCPs are about the only ones who rounded a lot back then who don't know. Cardiology, anything surgical, GI, all still round.

Just as a counterpoint, I round on my own inpatients as a PCP at a critical access type hospital, and it is set up REALLY nicely for me - frankly it's downright cushy, not to mention a nice boost to my RVUs. One of my other residency classmates rounds on his inpatients as part of a larger group where everybody does it (I am the only doc in my group who does it), and two of them do OB and round on their inpatient OB/postpartum patients and newborns. I love doing it, the patients seem to really appreciate it, and I think the continuity improves the care inpatient and transitioning back to outpatient. There are several family medicine groups I know of around where I did my residency where everybody rounds on their own inpatients as well and those docs seem happy with how it's set up.

For the AI stuff - it is already being used to generate notes, and the docs I know who are using it are big fans. For clinical decision making, I think it could be used to round out some differentials and quickly review guidelines and things like that for diagnosis and treatment, maybe more so in primary care vs specialties, suggestions above from operaman regarding billing are great as well. But you still have to use your doctor brain to figure out what the right next step is for the patient in front of you, and how to communicate that to the patient. I don't think it will ruin my enjoyment of my career by any means if done well, and I'm grateful for anything that helps me spend less time googling things and clicking clicky boxes in front of/in between patients. Whether AI will actually be able to achieve that/be done well is another question, but maybe someday. I certainly don't think it will automatically be terrible.

 

[xavi22]

We were one of the first with an EMR in our town about 20 years ago, never had a single paper chart. Certainly some wonderful capabilities came along - eRxs, e-lab ordering, messaging, finding charts, offsite access, drug interactions and allergy alerts. I love all those things. Why did it stop there? Maybe if we could afford EPIC, we might have some of that, but we are not some large corporation.

20 years later, we are still there with only fundamental EHR capabilities (EPIC giants aside) as our health care system continues to serve our people less and less. Now, with electronic data I spend a great deal of time with AI prior authorizations and rejections for necessary care. The greatest updates to our EHR has been chasing the MIPS reporting which seems to change just enough every year. I spend inordinate time entering data for quality of care benchmarks that have not been proven to increase quality of care. In fact, our national quality of care benchmarks have declined overall during this time. It remains expensive and out of reach for many.

I see AI in medical news articles more of a marketing hype by private equity entrepreneurs seeking to generate more revenue without regard to solving our actual problems. I want to know how much a drug will cost my patient before I send the prescription. I want easy access decision making support tools for standard screenings and common acute and chronic diagnoses. I want diagnostic tools based on reading my soap note data (custom subjective, exam, labs, and advanced imaging. So many things have not happened over 20 years.

we are working on these and more boss. EPIC must adjust or face real competition.

 

[xavi22]

Maybe. I just see it as a pretty hard sell to a cash strapped hospital when that product won’t actually help bring in any additional revenue and especially if the additional time makes their providers even less efficient. Even for midlevels, a level 3 visit and a level 3 visit with AI support pay exactly the same thing.

I just don’t find diagnosis to be remotely challenging at all. Honestly that’s maybe been my biggest disappointment with medicine in general - I thought that would be a much bigger part of my job than it is. Most of the time I know the diagnosis before entering the room and certainly within the first 5 minutes. I don’t need any support on that front.

Now, an AI that reviews my documentation before signing to make sure I’m hitting maximum level billing - that I could use! Something like “I saw you reviewed that CT scan in pacs but you don’t provide your own interpretation in this note. If you do, this would qualify as a level 4 rather than a level 3. Click here to add to your note.” And then based on my note, it drafts a sample blurb to insert that pre populates my findings.

I just don’t think anyone wants diagnosis/treatment decision support, it will only slow us down, and it won’t make anyone more money.

Scribes, inbox mgmt, chart review, coding/billing support - those are areas where it will pay for itself fast. Remember that most hospital execs are bonused on fairly short term goals, so some big purchase that may take years to pay for itself is going to be a tough sell upstairs unless all the docs are pushing for it and threatening to walk if they don’t get it.

Documentation review is a good use case. With Epic API (if they have any), a software can monitor what you click on each patient's chart, how long you spend, run it through AI (save it). Once you click sign, assistant API functions will be activated, and compare your notes to saved data, shows you a pop up of missed opportunities. Really not that difficult to code tbh.

 

[dgdg]

Documentation review is a good use case. With Epic API (if they have any), a software can monitor what you click on each patient's chart, how long you spend, run it through AI (save it). Once you click sign, assistant API functions will be activated, and compare your notes to saved data, shows you a pop up of missed opportunities. Really not that difficult to code tbh

The coding is easy. If it's not financially driven, won't happen....

 

[Mr.Smile12]

 

[annoyedpsychiatrist]

Do patients want chatgpt to do CBT with them? Chatgpt sucks at formulating plans in psychiatry

 

[xavi22]

Do patients want chatgpt to do CBT with them? Chatgpt sucks at formulating plans in psychiatry

you can fine-tune Chatgpt to be more precise using psych resources.