Advertisement - Members don't see this ad
Whelp, it's been a fun career.
AI in Psychiatry
Started by PsyDr
Get help with your application
Use all the free resources available to you from SDN: articles, guides, expert advising, forums discussions, and school research.
Whelp, it's been a fun career.
Pretty insane and really discouraging for current residents who haven't had years of an attending salary and have growing debt. You go through 12 years of training just for an overnight startup company to be able to now see all your stable patients that you have worked hard to stabilize. And who know what other privileges AI will get in the near future. The current rhetoric used is that doctors can now use more of their time managing complex patients, but doing fully outpatient this year in my PGY3 I look forward to seeing my stable patients sprinkled in throughout my schedule. It's a really rewarding part of the job. Not sure how everyone else feels about that. Anyway, I'm seriously considering doing locums now after I graduate to maximize my income potential. Maybe I'm being pessimistic but I never thought changes like this would happen so soon.
I've said it before but will say it again - if all you have to offer your stable patients is a refill of their prescriptions, AI will clearly eat your lunch. You have to have some kind of value add for those people. If you don't, they should go back to their PCP anyhow.
Advertisement - Members don't see this ad
That's very valid. I guess I'm more concerned with the expansion of privileges over time. AI hasn't been out that long. How long before AI can increase lexapro from 5 to 10mg for patients it deems "low risk"
It's the midlevel story all over again.
You don't view it differently? At least with midlevels there's a limiting factor of labor. With AI it can theoretically serve an unlimited amount of patients.
I know there's a slippery slope here and clearly there should be some concerns. But from a general humanity perspective, I'm not sure this is horrible. These meds are generally available OTC in most countries. It's fairly unique to the US that we require an MD to repeatedly sign off on lisinopril and we don't really consider the downsides of that in terms of medication access as much as we should. I also don't think MDs are reimbursed in any meaningful way for most of these refills. They're likely just unbilled clicks in an EMR. The CEO isn't wrong that most MDs won't check everything the AI will. Now am I scared? Sure. However, this particular narrow thing is not game ending. It's not even low hanging fruit really as it's only profitable via AI. I definitely don't anticipate the FDA will be taking any action. It was kind of bizarre the article even mentioned that.
Last edited:
Will take a few years to go widespread even under the best scenerio. I doubt there is any meaningful inpact before 2030 and likely longer than that but for sure its coming and will have a signficant impact. I could care less becuase im spending 25-30 min and meds are literally like the last 1-2 min of my sessions with patients. They all know this and continue to state how much better there care is now vs when the old psych doc spent 5 min with them refilling there meds.
Yeah I'm more on the lets watch the slippery slope thing but it sounds like it'll literally just be reviewing prescriptions people are already on and refilling them.
Also for some other specialties this isn't even reviewed by a doctor or nurse most of the time anyway. Tons of offices just have front office staff or MAs handle refills under the physician's name. Can't say they're any more qualified to assess how appropriate a stable refill is than a standardized system.
The only reason I see people who are stable on one med is often because they WANT to keep seeing me or don't have a PCP or something....it's actually semi annoying because often I can't justify higher than a 99213 for them. This would actually be great because sweet let the AI "system" take the liability for the prescription so I don't have to argue with people why they need to keep seeing my every X months so I actually get paid to have the liability for the med. Let them sue the AI company once it takes over prescriptive care.
Also for some other specialties this isn't even reviewed by a doctor or nurse most of the time anyway. Tons of offices just have front office staff or MAs handle refills under the physician's name. Can't say they're any more qualified to assess how appropriate a stable refill is than a standardized system.
The only reason I see people who are stable on one med is often because they WANT to keep seeing me or don't have a PCP or something....it's actually semi annoying because often I can't justify higher than a 99213 for them. This would actually be great because sweet let the AI "system" take the liability for the prescription so I don't have to argue with people why they need to keep seeing my every X months so I actually get paid to have the liability for the med. Let them sue the AI company once it takes over prescriptive care.
Yeah the company seems quite confident in their liability coverage and that's understandable because they literally get to select all of their patients as the statistically lowest possible risk. Honestly I can't even tell if they will be handling psych meds. Does it say anywhere?
After reading the article I'm not as immediately concerned about this as I think many people will be after just reading the title. Couple points that are very relevant:
1. The meds being refilled are limited to 190 common meds used for chronic illnesses. While that could make up a large percentage of prescriptions, it's <1% of all meds out there and probably won't affect most "psychotropic" meds that PCPs aren't already comfortable prescribing themselves.
2. The AI is not starting or changing meds, it's asking basic questions that our nurses/MAs would and then providing a renewal if deemed appropriate. If a patient who has been stable for years needs a refill of their lexapro 10 mg before they see me in a few weeks or months, sounds like a great way to decrease the inbox clutter. Most places already have automated refills or non-AI programs doing this, so this seems like a pretty benign use of AI anyway. More like a PE trying to make in-roads to peddle their product than anything.
3. Pharmacies are already using AI for prescription refills, sounds like some doc (Oskowitz) has a basic understanding of this system and is just generalizing similar programs for use on a larger scale. Sounds like another classic PE move of taking something that already exists and repurposing it for $$$.
4. They have a separate malpractice for this program, which means they're indirectly acknowledging that this program is "practicing medicine". The FDA isn't involved yet, but I'd be shocked if there isn't a major case or legislative battle over this sometime in the near future. This could also become something that is state dependent.
5. The company is looking to expand their product to other states. If this scales up to large volume, that means the potential for lawsuits to scale is there too. I could easily see this backfiring significantly, especially if kids manage to fool the system (and we should know from other articles how easy it is to jailbreak AI programs). There are already lawsuits against AI companies for people killing themselves just from using AI without the AI actually providing a service. I think it would be incredibly easy to have large suits, including class action suits, against this company when something goes wrong if the AI programs are providing refills leading to negative outcomes. I think this is even more plausible and even expected given they've obtained malpractice insurance. If that happens, how quickly is this going to get nixed when no one is willing to provide malpractice coverage?
6. The doc/company is also using AI for other means as well. The one stated in the article uses AI as a "consultant" that potential patients can answer questions in that will then direct them to physicians and specialists. This seems much more like a doc with some AI knowledge trying get ahead of the curve to make case before other companies inevitably occupy that space than someone actually trying to revolutionize medicine, especially since these services are already broadly implemented. It's the same product with different packaging.
Overall, I think this is probably much ado about very little. It's just a doc trying to use AI to make a quick buck (his quotes in the article read as so egotistical) before others in PE inevitably try it. I think the chances that this goes smoothly are about as likely as most other PE endeavors we've seen recently.
1. The meds being refilled are limited to 190 common meds used for chronic illnesses. While that could make up a large percentage of prescriptions, it's <1% of all meds out there and probably won't affect most "psychotropic" meds that PCPs aren't already comfortable prescribing themselves.
2. The AI is not starting or changing meds, it's asking basic questions that our nurses/MAs would and then providing a renewal if deemed appropriate. If a patient who has been stable for years needs a refill of their lexapro 10 mg before they see me in a few weeks or months, sounds like a great way to decrease the inbox clutter. Most places already have automated refills or non-AI programs doing this, so this seems like a pretty benign use of AI anyway. More like a PE trying to make in-roads to peddle their product than anything.
3. Pharmacies are already using AI for prescription refills, sounds like some doc (Oskowitz) has a basic understanding of this system and is just generalizing similar programs for use on a larger scale. Sounds like another classic PE move of taking something that already exists and repurposing it for $$$.
4. They have a separate malpractice for this program, which means they're indirectly acknowledging that this program is "practicing medicine". The FDA isn't involved yet, but I'd be shocked if there isn't a major case or legislative battle over this sometime in the near future. This could also become something that is state dependent.
5. The company is looking to expand their product to other states. If this scales up to large volume, that means the potential for lawsuits to scale is there too. I could easily see this backfiring significantly, especially if kids manage to fool the system (and we should know from other articles how easy it is to jailbreak AI programs). There are already lawsuits against AI companies for people killing themselves just from using AI without the AI actually providing a service. I think it would be incredibly easy to have large suits, including class action suits, against this company when something goes wrong if the AI programs are providing refills leading to negative outcomes. I think this is even more plausible and even expected given they've obtained malpractice insurance. If that happens, how quickly is this going to get nixed when no one is willing to provide malpractice coverage?
6. The doc/company is also using AI for other means as well. The one stated in the article uses AI as a "consultant" that potential patients can answer questions in that will then direct them to physicians and specialists. This seems much more like a doc with some AI knowledge trying get ahead of the curve to make case before other companies inevitably occupy that space than someone actually trying to revolutionize medicine, especially since these services are already broadly implemented. It's the same product with different packaging.
Overall, I think this is probably much ado about very little. It's just a doc trying to use AI to make a quick buck (his quotes in the article read as so egotistical) before others in PE inevitably try it. I think the chances that this goes smoothly are about as likely as most other PE endeavors we've seen recently.
As others have said, I do think the slippery slope arguments have some legitimate basis. AI is coming, and the timeline at which it is achieving landmarks is in many cases drastically accelerated from what has been previously predicted. If this is just a means to get AI into healthcare at a higher level and the applications are able to be generalized to changing/starting meds and then to diagnosing, that is a legitimate concern. At the same time, I think a lot of people in tech dramatically overestimate AI's capabilities and vastly underestimate the complexity of things we see as common or basic.
Additionally, I also think psychiatry is going to once again be far less susceptible to replacement than other fields. Our meds interact with everything, they get blamed for every side effect and problem, and answering questions about SI/HI/AVH is going to escalate the Rx order get sent to us anyway. There's a huge difference in the eyes of the public of refilling lisinopril or metformin vs pretty much any psych med. There are still a lot of things that AI is just plain bad at, and the idea that AI is going to be able to pick up and integrate nuanced cognitive concepts or situational cues in a consistent and accurate way is something I'm extremely skeptical of. I really don't see this as anything to get up in arms about at this point. I'm sure things could become very different 10-20 years from now, but I do think we will still have a significant place at the (ignored side of the) table for many years to come.
Additionally, I also think psychiatry is going to once again be far less susceptible to replacement than other fields. Our meds interact with everything, they get blamed for every side effect and problem, and answering questions about SI/HI/AVH is going to escalate the Rx order get sent to us anyway. There's a huge difference in the eyes of the public of refilling lisinopril or metformin vs pretty much any psych med. There are still a lot of things that AI is just plain bad at, and the idea that AI is going to be able to pick up and integrate nuanced cognitive concepts or situational cues in a consistent and accurate way is something I'm extremely skeptical of. I really don't see this as anything to get up in arms about at this point. I'm sure things could become very different 10-20 years from now, but I do think we will still have a significant place at the (ignored side of the) table for many years to come.
Is the pharmacist that checks the prescription an AI algorithm, too?
I am much more concerned about the slippery slope issues. Anyone could write a prescription. However, the legal recognition of that prescription might be significant. This is evolving quickly, with David Schweikert (R-AZ) introducing the The Healthy Technology Act of 2025 (H.R. 238) in January 2025, which would allow artificial intelligence (AI) to be classified as practitioners eligible to prescribe medications.
AI is going to be here huge in psychiatry, whether we like it or not. I imagine salary depression: is an APP/psychotherapist with an AI as good as a psychiatrist? If yes (plausible, if not likely), then why are we using psychiatrists given they are 3x the cost? At that point, our salaries will need justification.
Advertisement - Members don't see this ad
I'm not worried at all.
The day they take my job - almost no one has a job 🤣
Salary depression for psych is unlikely. I keep seeing more and more demand. People are not only desperate for care and providers - they are very desperate for good providers. AI being fed a bunch of slop when so much of the field is sloppy guarantees garbage in, garbage out.
Almost all of my patients are complex, I'm afraid. Too much job security for me. The day a PCP can care for my patients is the day I retire.
The day they take my job - almost no one has a job 🤣
Salary depression for psych is unlikely. I keep seeing more and more demand. People are not only desperate for care and providers - they are very desperate for good providers. AI being fed a bunch of slop when so much of the field is sloppy guarantees garbage in, garbage out.
Almost all of my patients are complex, I'm afraid. Too much job security for me. The day a PCP can care for my patients is the day I retire.
I mean sure it's a rando bill introduced by one rep. Anyone can propose a bill about anything. So are these:
Help me out. Why is this bill, that has been introduced twice, “rando”, but a similar bill for psychologists was labeled a “bad” on here less than 5 years ago?
“...but I think the psychology national organizations are using this to crack open the door. Once they get it in the VA, it's like NPs...why not roll it out everywhere then?”- you
"Rando" doesn't mean "not stupid" it means "a random rep introduced a goofy bill". They're all stupid doesn't mean it's gonna pass anymore than the trump mount rushmore bill.
I actually think the trump rushmore bill has a better chance of passing in the near future lol.
So AI scope expansion efforts can be ignored, but not psychologists?
Unlike AI making medical decisions, prescribing psychologists are currently practicing medicine right now in several states and several other states have actual similar legislation coming up for actual votes.
I like that you keep track of my quotes from 5 years ago 😉
The Veterans COMPACT Act which originally contained that language around allowing prescribing psychologists throughout the VA was introduced in the House 9/2020 and became law by 12/2020. It was an issue because that was proposed language was going to be contained within a bill expanding VA mental health services that was clearly going to pass pretty easily.
Yes much more of an issue than a standalone bill introduced by one guy with no cosponsers that's stagnated in the House Energy and Commerce committee since last January. If this had any chance of actually getting out of committee, the AMA, AANP and AAPA would all suddenly realize they're all buddies for one day and would be going wild about it.
Last edited:
psydocc
Full Member
Even when the day arrives that congress allows AI to be a "practitioner" that can prescribe meds, I think there will still be plenty of people that value 1. talking to an actual person and 2. getting treated by an actual psychiatrist.
I could even see a scenario where being seen by the psychiatrist comes at a premium which could mean more money for us, not less.
The scenario I can't ever see happening is one where my more educated/health literate patients, many of whom cared greatly about the pedigree of my training, are suddenly okay with putting their brain in the hands of an online diploma mill NP who is being supervised by a hallucinating chatbox.
I could even see a scenario where being seen by the psychiatrist comes at a premium which could mean more money for us, not less.
The scenario I can't ever see happening is one where my more educated/health literate patients, many of whom cared greatly about the pedigree of my training, are suddenly okay with putting their brain in the hands of an online diploma mill NP who is being supervised by a hallucinating chatbox.
Correct. Do I have worries that there will be even more of a 2-tier (or 3-tier) system with AI that's already happening with mid-levels for medicaid/uninsured? Absolutely. Am I worried about psychiatrists losing their jobs or having a downward force on pay in the next decade? Absolutely not.
If AI can do our work, the white collar world is absolutely already going to be cooked, every pathologist and radiologist will be living under a bridge. PCPs will be supervising an army of bots and 3/4 of them living under a bridge. I will have to explain to my teenagers that accountants and lawyers were prestigious and important jobs back around the same time dinosaurs ruled the Earth, etc.
hell AI is still making up freaking legal citations and quotes to this day haha
Judges find suspected AI hallucinations in PA court cases
A Pennsylvania Commonwealth Court judge raised questions about the use of AI when confronted with an error-filled brief.
hell AI is still making up freaking legal citations and quotes to this day haha
Judges find suspected AI hallucinations in PA court cases
A Pennsylvania Commonwealth Court judge raised questions about the use of AI when confronted with an error-filled brief.www.spotlightpa.org
Skill issue, notice most of those plaintiffs are pro se
I mean it's gonna be a skill issue when people are putting their own prompts into Doctor AI too....the point is that it's still doing this, people are not picking up on it because they're putting too much trust into the tool and it's a known issue that AI "hallucinations" may or may not be fixable at the very least in the short term if at all with current models.
Current frontier models hallucinate far, far less than what's available for free. You're right about this plaguing people trying to use this on your own. The skill issue is having enough experience with these models to recognize the circumstances under which they are more prone to invent answers and how to and how not to ask questions. For the near future we are still in the centaur era, where an expert human paired with an LLM is far more effective than either alone. Unfortunately that may not last as long as we'd like.
Its good to keep an eye on the landscape as its evolving. Will reassess in 2030. Id like to be in a position where whatever happens with medicine I didn't slack off and didn't take advantage of the available work in front of me. Would love to be in a place to be PT by end of 2030. If people are able to do this i would encourage the best defense sometimes is a great offense.
I remember more than I want!
Advertisement - Members don't see this ad
I realize this is Google summaries vs more directly trained AI. A big part of our job imo is parsing through what patient information is relevant and what needs to be collected to make valid diagnoses and treatment plans. AI sucks at this and until AI can critically assess source material I’m still not overly concerned.
www.theguardian.com
‘Dangerous and alarming’: Google removes some of its AI summaries after users’ health put at risk
Guardian investigation finds AI Overviews provided inaccurate and false information when queried over blood tests
I realize this is Google summaries vs more directly trained AI. A big part of our job imo is parsing through what patient information is relevant and what needs to be collected to make valid diagnoses and treatment plans. AI sucks at this and until AI can critically assess source material I’m still not overly concerned.
‘Dangerous and alarming’: Google removes some of its AI summaries after users’ health put at risk
Guardian investigation finds AI Overviews provided inaccurate and false information when queried over blood tests
AI Overviews don't actually use frontier models, they are not actually an accurate reflection of the current capabilities of the technology even as they exist at this very moment.
It could lead to further delay of people gracing my office (our offices).
I got PCPs in area that manage depression anxiety whatever for several years, and patient tread water with no real relief that I can get get to remission in 1-6 months. So its another stepping stone delay...
I don't let patients go longer than q90 days for appointments and all refills, I cruise their charts when doing refills, and appointments, essentially it helps me do another chart review, which I have concerns the AI won't do as well as myself. As others pointed out, probably an extension of the MA/RN doing a refill on behalf. If it pseudo glances thru chart, refills for a year, that's essentially 3 or more missed chart reviews.
People in general don't value doctors, too few do, so I don't foresee it leading to a premium rate increase.
It'll get expansion/utilization by something like UHC, implemented by flagging any F code diagnosis and then taking over for patient care until the patient has had a PCP induced prior authorization to permit care with a Psychiatrist, in a backdoor HMO fashion...
There will be a grouping of people that pull back from AI stuff, like Will Smith in iRobot movie 'I prefer to drive myself' and almost like an Amish offshoot. Intentional technological stagnation.
So far even with phone tree automations, insurance companies or other things, more than half the time I'm growling at the phone "get me a human" as the AI is incapable of addressing the complexity of the problem.
I got PCPs in area that manage depression anxiety whatever for several years, and patient tread water with no real relief that I can get get to remission in 1-6 months. So its another stepping stone delay...
I don't let patients go longer than q90 days for appointments and all refills, I cruise their charts when doing refills, and appointments, essentially it helps me do another chart review, which I have concerns the AI won't do as well as myself. As others pointed out, probably an extension of the MA/RN doing a refill on behalf. If it pseudo glances thru chart, refills for a year, that's essentially 3 or more missed chart reviews.
People in general don't value doctors, too few do, so I don't foresee it leading to a premium rate increase.
It'll get expansion/utilization by something like UHC, implemented by flagging any F code diagnosis and then taking over for patient care until the patient has had a PCP induced prior authorization to permit care with a Psychiatrist, in a backdoor HMO fashion...
There will be a grouping of people that pull back from AI stuff, like Will Smith in iRobot movie 'I prefer to drive myself' and almost like an Amish offshoot. Intentional technological stagnation.
So far even with phone tree automations, insurance companies or other things, more than half the time I'm growling at the phone "get me a human" as the AI is incapable of addressing the complexity of the problem.
Phone tree automatons are to a first approximation never LLMs, so they are a bit irrelevant here. I don't disagree with a lot of the rest of your analysis in terms of how people may react.
Sure, and there are far better technologies than that. Maybe a better example is that we have Microsoft Abridge where I'm at which is award winning as a healthcare EMR in terms of clinical summarization, scribing, and documentation. I have used it a bit and frankly I find it very inadequate in terms of completing the tasks I'm using it for (summarization and scribing). I realize that this is still different from LLMs and some of the leading programs like Claude, Gemini, or ChatGPT, but if models that are designed specifically for our settings and purposes are failing me to this extent (and I've confirmed with several colleagues who are big into AI it's not user error on my part), then I'm not at the point where I feel worried about the utilization we're talking about in this thread.
Will we get there? I'm sure there's a very high possibility we could. But as others have noted, if we're at the point that psychiatrists aren't relevant anymore then I think most non-labor jobs in society will have already been condemned to irrelevance.
Welp, was nice knowin' you guys. I always joked about having to get a job at Best Buy if medicine took a nose dive. AI is probably going to take that job too, though.
AI is MUCH BETTER than midlevels.
This was discussed in another thread, but the folks here who are minimizing the impact that AI is going to have in the fairly near future on almost every field (not just psychiatry, or medicine in general) are burying their heads in the sand. I do not intend that as an insult; it's a natural response, especially for the younger amongst us.
I was interviewing medical students for residency yesterday, and I kept wondering if and how long they will eventually be attendings. By the way, and not unrelated, 50% of these applicants had either failed Step 1, COMLEX 1, or a full year of medical school. 😱 The program I was doing interviews for also has issues with their residents passing Step 3 and/or the psychiatry board exam. AI WILL NOT fail these. And then I hear "but the human touch of psychiatrists!" I'd say at least 50% of the psychiatrists I have dealt with are not warm or even particularly nice to patients. But we will see.
I was interviewing medical students for residency yesterday, and I kept wondering if and how long they will eventually be attendings. By the way, and not unrelated, 50% of these applicants had either failed Step 1, COMLEX 1, or a full year of medical school. 😱 The program I was doing interviews for also has issues with their residents passing Step 3 and/or the psychiatry board exam. AI WILL NOT fail these. And then I hear "but the human touch of psychiatrists!" I'd say at least 50% of the psychiatrists I have dealt with are not warm or even particularly nice to patients. But we will see.
I think this is a 'sky is falling' take. We still dont have flying cars. The phone did not replace the computer. Auto-read on EKGs did not replace cardiologists. And google search did not replace doctors.
AI is fancy google search. Nothing more, nothing less. It is pouring over countless instances of crap data and making errors like any toddler would if put in the drivers seat of a ferrari. Zero chance MD psychiatrists are out of a job unless AI has absorbed all "thinking" jobs on the planet. They day that occurs, jobs are not necessary anyway.
Its like how automation and cheap labor was going to replace carpentry. Low and behold - the opposite occurred. People are desperate for expert carpentry and they are booked out months in advanced charging hundreds of percents more each decade. There is a shortage of expertise across the board. AI will further widen the gap where expertise is almost impossible to find.
Recent AI conference i attended suggested the most patient facing specialties least likely to be AI infiltrated and the least patient facing the most. I just need my rent paid and my child support payments covered, then im golden. I will push my Acura to the max.
psydocc
Full Member
Under this very logic, surely you can understand why some patients value crossing paths with a good psychiatrist? This is the reason my cash patients will shell out $600/hour to see me instead of going to whichever Tom, Dick, and Karen NP takes their insurance.
except there are NPs who charge $750/hr and some patients prefer NPs because they think they are more holistic and person-centered!
AI is even more validating/enabling than any NP or masters level therapist.
I have noticed that nurses and NPs don't want to see NPs themselves or for their families though.
The fact that frontier LLMs are capable of independently solving novel and previously unsolved mathematical theorems, extensively programming large code bases with limited human supervision, and acting as agentic problem solvers in a wide range of domains rather belies this. They may train on the data of the Internet but to conceptualize them as a fancy Google search is a a fundamental misunderstanding of the technology. Something more complicated and interesting is going on.
The one thing AI won't be able to do, rest assured, is be human-- and if it matters to be human to be a psychiatrist, then our jobs will be resilient in at least that respect. Also, if to be human is to suffer, then our jobs are secure, as the transhumanist eutopia is then dead in the water. If either of those are false, or both are false, then yeah, AI could eventually deem psychiatrists obsolete.
The other necessary is that the economy is stable enough going forward to afford at least the standard of living we enjoy today where expendable income can be spent on psychiatric care, and not on more basal needs. If mass amounts of people are unemployed and miserable, I don't think that bupropion and psychotherapy is going to be the answer. In my imagination, the profession then competes for the few very wealthy people who exist, and whoever doesn't have wealthy patients will either need to switch to being a PCP in order to care for physical problems because society fell apart OR compete economically with mid levels empowered by better-than-us AI because they're *just tools* OR have a great life because AI utopia was achieved.
One thing I'm grateful for right now is my non-manualized psychotherapy skills. Anything algorithmic will be first to go to AI management, simply from a data processing standpoint.
The other necessary is that the economy is stable enough going forward to afford at least the standard of living we enjoy today where expendable income can be spent on psychiatric care, and not on more basal needs. If mass amounts of people are unemployed and miserable, I don't think that bupropion and psychotherapy is going to be the answer. In my imagination, the profession then competes for the few very wealthy people who exist, and whoever doesn't have wealthy patients will either need to switch to being a PCP in order to care for physical problems because society fell apart OR compete economically with mid levels empowered by better-than-us AI because they're *just tools* OR have a great life because AI utopia was achieved.
One thing I'm grateful for right now is my non-manualized psychotherapy skills. Anything algorithmic will be first to go to AI management, simply from a data processing standpoint.
Work harder through 2030. Thats the answer. Going in circles here the answer is no one knows and its possibly a legit threat.
Ill be working my tail off through 2030 and ideally be in a position that im financially indep. If ur not willing to do the same then dont complain if this turns out to be a much bigger concern.
Ill be working my tail off through 2030 and ideally be in a position that im financially indep. If ur not willing to do the same then dont complain if this turns out to be a much bigger concern.
Advertisement - Members don't see this ad
That is my thought. EKGs sound like one of the easiest parts of medical practice to automate, and yet we still really want humans to read them. Radiology and maybe path are one step above that but I remember when I was in med school 10+ years ago radiology was supposed to be dead soon due to computers and now AI, but seems like radiologists are now doing better than ever. Things will change for sure, probably for the worse for everyone except stockholders, but history repeats itself and I doubt we'll really advance much beyond a credentialed human sitting amusing a patient while the body heals itself. But the credentialed human will be using OpenEvidence to answer the patient's concerns which were raised by some ChatGPT queries. Pediatricians will continue to make no money and ERs will continue to be the dumping ground of society.
EKGs routinely misinterpret QTc in patients with abnormal rhythms, especially if there are very small or very wide T waves. I see automated reads of QTc actually measuring the QP interval and completely missing the T wave far more often than it should. You'd think this wouldn't be an issue at this point, but on C/L I find myself recalculating the QTc myself far more often than I'd like. If nothing else it gets me a point for medical complexity when I'm billing.
Man I knew this was a thing 15 years ago when I actually had to look at these, but how in the world has this not been fixed in the intervening 15 years? I thought AI was extensively programming and solving hereto unsolved math theorems, but we can't even just read a QTc accurately on all EKGs?
You would think this would have been addressed, but it’s easier to just have a physician doa. Quick read and sign their initials. Probably more for the liability coverage than anything, but one would think this would just be redundant at this point.
Bigger question: have we actually tried to teach AI to read EKGs?You would think this would have been addressed, but it’s easier to just have a physician doa. Quick read and sign their initials. Probably more for the liability coverage than anything, but one would think this would just be redundant at this point.
Because the software that's built into the EKG machines is absolutely not AI.
