AI in Medicine

[NateB]

Get rid of computers period! Major reason I retired at age 64. Turning M.D.'s into keypunch artists is the biggest waste of their time! I dictated my paper medical records based on my hand written crib notes so anyone could read my paper medical records clearly. I could page through my records and see clearly what I did to a patient in the past. Saved me a lot of time. Sure beat "scroll click, scroll click, scroll click" chit.
If I was filling in for another M.D. that didn't dictate and their hand writing was illegible, it was a "beech" to figure out what the heck the doc was doing.
Asked the head hospital administrator if he minded if I dictated all my hospital notes in the hospital system when I went to work at the local clinic I stayed at for my entire career. H&P's and discharge summaries were expected to be dictated. He said it was fine to do all my notes dictated. They would appear on the paper charts the next day and I'd sign off on them before "everything" became computerized.
When computers came online, records now occupy most of a primary care M.D.'s time and make primary care so dissatisfying. Was so glad I was able to
retire and get out when I did. I have no regrets. The student docs coming out now are too stupid to realize how good it was years ago.
Computers for lab work I have a different opinion of. Was happy with that as I could access lab work hot off the press and didn't have to wait for the "hard copy".
Oh, I had many a phone call from an itinerant E.R. doc for a patient admit of mine and received many compliments from the doc as they were reviewing the
patients' past medical records and they could "READ" everything I did in the past in the old records.
I DO NOT recommend to any young person to go into medicine anymore. It's a waste of time. It was so much more efficient in the days of old with dictated paper records. It's nearly impossible to do office work, hospital work and take call in primary care anymore due to the current system.
If one is in an "office only" practice with income based on production only, it's impossible to produce enough if one has a bunch of elderly complex patients to see. They take more time and hence production is down. Plus if one isn't involved with the care of a patient in the hospital, they are clueless once they're
discharged. Can follow with the specialists if they're needed and be familiar with what is going on with the patient before discharge.
Nowadays in the next big city over, primary care docs aren't allowed to see patients in the hospital anymore. That has been going on for
years now. There are "hospitalists" who never see the patient outside of the hospital and are "clueless" as to what is going on. Continuity of care
has "gone to hell".
I recommend to go into another field. O.k. some specialties are better than others but that would take another long post to go into. Kurt (used to M.D. for a living now retired)

Our ED docs dictate their narratives into Epic, but I don't think human transcriptionists are used any more. Other orders and notes are all entered by drop-down boxes or clicking check boxes. I don't know how many times I see a sedated and paralyzed patient with a check box ticked that says every procedure was discussed and the patient agreed.

I started EMS writing paper charts and not being allowed to leave the ER until it was done and left with the unit secretary. We would write on the road and spend a few minutes tidying it up before we went back into service. If we were too busy to write during the run, it still took about 20 minutes and we were done. My role now often requires more intense patient care, but I can spend 45 minutes with a patient and 2 hours documenting everything if we have a lot of procedures and meds to chart. I can only imagine it is more frustrating for physicians. At least we only have 1 or 2 patients at a time.

 

[Art Upton]

I don’t know that!! Maybe they would get more applicants if they advertised that.

Honestly I was more thinking about being waked up in the middle of the night, if I was a doctor I’d hate that the most (I felt so bad for the poor doctors who had to be dragged out of bed the one time I had a emergency surgery) but your correct it probably takes a toll working in trauma like you do.

Better not get into an important [to the customer that is] IT job then. While not as Tramatic as a Doctor's job, I have been woke up at night or bothered in the evening much to my Wife's discontent since the mid 1980s. First Pagers then Cell Phones.

In about 1983 our phone rang at home and my wife while we were still in bed about 7am answered it. It was a Doctor that bought an Apple Clone computer from the store I worked for that had looked my name up in the phone book and expected me to trouble shoot his problem over the phone while I was in bed.

No it was not a medical emergency at all, not even a clerical emergency; just that he wanted tech support. It got much worse than that as the years went on and IT folks became "exempt" from hourly pay and put on "Salary" so they could be made to work around the clock.

 

[Zeus-cat]

One big problem is that as Dr. House repeatedly told us, patients lie.

Patients lie to human doctors too, but one assumes humans can detect lies much easier than AI. And AI can't tell when a person is guessing about a symptom, which could lead the AI down a very wrong path. And doctors can touch and probe where AI can only deal with the input given to it by the patient. All of this can be worked out, but it will take time.

That reminds me of something a contractor repeatedly told me when I was a junior officer in the USAF; "We can do anything you want, all it takes is time and money."

 

[dhbarr]

Better not get into an important [to the customer that is] IT job then. While not as Tramatic as a Doctor's job, I have been woke up at night or bothered in the evening much to my Wife's discontent since the mid 1980s. First Pagers then Cell Phones.

In about 1983 our phone rang at home and my wife while we were still in bed about 7am answered it. It was a Doctor that bought an Apple Clone computer from the store I worked for that had looked my name up in the phone book and expected me to trouble shoot his problem over the phone while I was in bed.

No it was not a medical emergency at all, not even a clerical emergency; just that he wanted tech support. It got much worse than that as the years went on and IT folks became "exempt" from hourly pay and put on "Salary" so they could be made to work around the clock.

Or just get a more important-er IT job, where procedures need to be documented and the oncall rotation can be passed around.

 

[ksaves2]

Our ED docs dictate their narratives into Epic, but I don't think human transcriptionists are used any more. Other orders and notes are all entered by drop-down boxes or clicking check boxes. I don't know how many times I see a sedated and paralyzed patient with a check box ticked that says every procedure was discussed and the patient agreed.

I started EMS writing paper charts and not being allowed to leave the ER until it was done and left with the unit secretary. We would write on the road and spend a few minutes tidying it up before we went back into service. If we were too busy to write during the run, it still took about 20 minutes and we were done. My role now often requires more intense patient care, but I can spend 45 minutes with a patient and 2 hours documenting everything if we have a lot of procedures and meds to chart. I can only imagine it is more frustrating for physicians. At least we only have 1 or 2 patients at a time.

Try doing that with 20+ complex patients a day. Some of the oldster docs back in the early days saw 40 to 60 patients a day in the office. They got
away with it as the patients did o.k. most of the time but their notes sucked if someone else tried to read their handwriting. Plus they took call and did hospital work.

There weren't more modern therapies back then anyways. I retired 3 years ago and don't miss it. Miss some of the patients who listened to me and
were still alive when I retired after thirty something years of practice. Still see 'em in Walmart sometimes and it makes me very happy when they say they're still here because of me and offer thanks. There's no outside feeling better than that besides marital bliss.

I saw the number logs in the 60's, 70's and even into the 80's when primary care docs couldn't do as much to treat patients as they can now. I couldn't spite the older docs for that.

The dictating software when I practiced at the end of my career was so "efffffed" up the doc still had to "scroll click, scroll click, scroll click" to get to the right line to dictate. It wasn't a stream of consciousness note and was a PITA that took too much time. A waste of time if you ask me.

AI is a bunch of B.S. that is going to fail no matter what in medicine. It takes a human hand to take a history, examine and a human mind to sort the situation out. Type symptoms into a machine and when it screws up? WHO IS ONE TO SUE!!!!!!! I rest my case. Kurt

 

[cwbullet]

Professions such as medicine, dentistry, pharmacy, and teaching have several commonalities. They require a commitment to the profession itself, they require a lot of education, they put the individual in a hole financially. And today they are increasingly stressful due to restraints placed on them by people who have no earthly business making those restraints.

[Imagine if auto repairs were in a situation similar to medicine. $300 for an oil change, $40k for a transmission rebuild. People get auto repair insurance that brings costs down to today's levels, but no policy covers everything, deductibles for this and that, and there are weird loopholes. Mechanic diagnoses a problem but has to submit it to insurance personnel who know nothing whatever about automobiles, etc.​

​

Pharmacists and pharmacy techs are leaving the profession in droves because of the stress. The insurance is equally incomprehensible to pharmacists, techs, patients, and the insurance reps. By design IMHO.]​

To get more students to enter such professions and to keep those who are already there, start by removiong the stress produced by unnecessary interference in the profession.

That is tough. A clinician goes to medical school and that requires that they often start working to be paid 10 years later. They often come out 250-500K in debt. I fully understand why they charged so much. They have to do so to make up time and money lost so they can eventually retire. I would like to see subsidies for the training so we can lower the costs. To be honest I have not idea what the solution is.

 

[Azamiryou]

AI is a very broad term. Due to recent news about large language models (like ChatGPT), a lot of people seem to want to focus on that as AI.

There are many other kinds of AI that don't make stuff up, and are pretty good at finding and matching patterns, detecting anomalies, and assigning probabilities.

I wouldn't want ChatGPT to be my doctor, but having an unbiased* system with tons of data that can carefully consider all of my symptoms to determine possible causes, how likely each is, and the most effective treatments — that sounds like a big step forward!

*Of course it will still have some bias because of whatever data it's trained on, but it won't be prone to overdiagnosing whatever it's read about most recently, nor ignoring some symptoms because the patient is obese or female.

 

[cwbullet]

AI is a very broad term. Due to recent news about large language models (like ChatGPT), a lot of people seem to want to focus on that as AI.

There are many other kinds of AI that don't make stuff up, and are pretty good at finding and matching patterns, detecting anomalies, and assigning probabilities.

I wouldn't want ChatGPT to be my doctor, but having an unbiased* system with tons of data that can carefully consider all of my symptoms to determine possible causes, how likely each is, and the most effective treatments — that sounds like a big step forward!

*Of course it will still have some bias because of whatever data it's trained on, but it won't be prone to overdiagnosing whatever it's read about most recently, nor ignoring some symptoms because the patient is obese or female.

I think having AI assist physicians with suggested diagnoses might be a good middle ground.

 

[blackwing94]

How will insurance companies handle medical malpractice insurance for the virtual AI doc software? That will be interesting.

 

[ksaves2]

That is tough. A clinician goes to medical school and that requires that they often start working to be paid 10 years later. They often come out 250-500K in debt. I fully understand why they charged so much. They have to do so to make up time and money lost so they can eventually retire. I would like to see subsidies for the training so we can lower the costs. To be honest I have not idea what the solution is.

Back in the day, 1978 to 1982 the state did have subsidies which helped a lot for me. They eventually went away but I got my use out of them. Maybe the politicos will bring that back at state med schools. At private institutions, one is on their own if they get accepted and elect to go there.
In undergrad, parents gave me no choice. Told me I was going to college, they're going to pay for it and didn't care what I majored in. I recognized I was very lucky. I kept going and going and going! They were proud of me anyways with the final outcome though. Another thing is delayed gratification. I didn't marry until I was nearly 32 years old to a very lovely spouse until she passed away 5 years ago.
I had many college classmates who had to get student loans and the interest rates were so good, they took their time paying off their loans even after they got out of professional school and were working. I don't know what it's like nowadays.
I think Biden's loan forgiveness is a travesty for those hard working people who responsibly paid off their loans. Sorry to get political but dementia Joe needs to go period before he ruins the country. (Political opinion is turned off now) Kurt

 

[Bravo52]

Now, some will accept students with only 2 years of udnergrad.

Medical school is hard work but not difficult. It takes a strong work ethic.

My son was offered a ride to a 6 yr BS/MD program at the state school. He looked at the program and it seemed like a really good program. They trimmed all the non-necessary electives. I liked what they were doing but alas, he did not want to go into medicine. He opted for the standard 4yr ME degree. (He's not even working in that field after graduation.) We both realized there is little margin in the 6yr Med program. It seemed "work ethic" was the key for sure.

AI will fit perfectly in the medical community. It will more than likely be more accurate than most Drs....BUT...there is only one problem with that. It's people. There is no way AI can easily bridge the gap between medical diagnosis/treatment and how people will respond. Humans are best at understanding humans! The best answer is likely a combination of AI and Drs.

Some time ago, I witnessed an experiment where the AF took the best fighter pilots around and put them in the most advanced "AI" simulators around and they waxed the floor against the computers. At one point during the testing, one of the F-16 pilots was just losing his mind since he started getting beat in the sim. Happened over and over again. The problem was, he was no longer fighting the sim, they had secretly hooked up to another sim with a live pilot in it! Fun times.

 

[les]

Personally, I've experienced, or a close family member has experienced, many issues with human doctors making mistakes, and for the most part, not willing to accept/acknowledge that they made a mistake. AI may need improvement, but I have a major distrust with real doctors. People are people. Mistakes will be made. Items will be missed or forgotten. They will have their good and bad days......

But I also feel that doctors are getting squeezed by the insurance companies. And the amount of red tape they need to go through ties up resources and adds to costs.

 

[NateB]

I think having AI assist physicians with suggested diagnoses might be a good middle ground.

AI will fit perfectly in the medical community. It will more than likely be more accurate than most Drs....BUT...there is only one problem with that. It's people.

AI is only as good as the information the algorithm is trained on and the information it is given to screen. A few examples lie in our transport cardiac monitors. It will interpret a 12 lead EKG and warn of a STEMI with bold letters when it prints out the strip. It is dependent on the leads being placed properly, the patient lying still, and the absence of artifact from the machine movement. It tends to be accurate when it does predict a STEMI, but it misses some too. If I called for the cath lab purely because the machine said it was a STEMI, that cardiologist would probably never trust me again. The machine can guide us, but still takes a person to make the best judgement call.

The same monitor has some filters to filter to out road noise and try to interpret the heart rhythm while CPR is being performed, it's still not reliable. There's a stretch of highway where the road vibrations will make the monitor think the patient is in ventricular tachycardia. Good thing it can't deliver a shock by itself.

We also have AI assisting our dispatchers ... that's a mess too.

 

[ksaves2]

My son was offered a ride to a 6 yr BS/MD program at the state school. He looked at the program and it seemed like a really good program. They trimmed all the non-necessary electives. I liked what they were doing but alas, he did not want to go into medicine. He opted for the standard 4yr ME degree. (He's not even working in that field after graduation.) We both realized there is little margin in the 6yr Med program. It seemed "work ethic" was the key for sure.

AI will fit perfectly in the medical community. It will more than likely be more accurate than most Drs....BUT...there is only one problem with that. It's people. There is no way AI can easily bridge the gap between medical diagnosis/treatment and how people will respond. Humans are best at understanding humans! The best answer is likely a combination of AI and Drs.

Some time ago, I witnessed an experiment where the AF took the best fighter pilots around and put them in the most advanced "AI" simulators around and they waxed the floor against the computers. At one point during the testing, one of the F-16 pilots was just losing his mind since he started getting beat in the sim. Happened over and over again. The problem was, he was no longer fighting the sim, they had secretly hooked up to another sim with a live pilot in it! Fun times.

The deal after med school is the 3 to 5+ year residency training the new doc has to go through. It's not just graduate med school and done. Takes a lot more time thereafter to be certifiably ready for independent practice. AI will not work on its' own. Heck I used computer searches to help me with puzzling patients in the early internet age but since I had a medical degree, I could sort through what was B.S. and what wasn't. I highly doubt AI overall will be better than physicians. Docs are human and yes mistakes can be made. The key is to recognize them and do what was necessary to remedy them. Kurt

 

[Azamiryou]

Personally, I've experienced, or a close family member has experienced, many issues with human doctors making mistakes, and for the most part, not willing to accept/acknowledge that they made a mistake. AI may need improvement, but I have a major distrust with real doctors. People are people. Mistakes will be made. Items will be missed or forgotten. They will have their good and bad days......

I think it's the same as with self driving cars. People have a hard time trusting AI drivers, but the AI would have to be pretty awful to be worse than humans. If robot cars took over all driving and were so bad they killed 10,000 people a year in the US... that would save 20-30,000 lives a year.

 

[WizardOfBoz]

So, my job is helping to guide a group of engineers and scientists in making math models of disease. Think of them as flight simulators for the body. And, like flight simulators, where you can put in different parameters for each plane, in our models you put in different parameters for each "virtual patient". The equations we use are based upon physics, and also upon very focussed (reductionist) experiments. We tie the equations together with laws of physics (like "you can't create or destroy mass") and then calibrate the models with trial data.

My company's marketing group posted a fairly low-tech assessment of AI vs our mechanistic models, and a guy on LinkedIn objected. Turns out, he's an MD who runs an AI startup in CA. He sharpened my thinking, and recommended that I read Judea Pearl's "The Book of Why".

It kind of explains that AI that only curve fits existing data sets cannot really be very predictive. But AI that creates a model (like the we make) CAN be predictive of "counterfactuals" (experiments no one has every run).

So I've been working with a foundation that researches a cancer (Ewing's Sarcoma) to come up with individualized treatments, to the OPs initial post.

Would I trust AI to do a diagnosis? No. Would I want my physician, AFTER he/she gets the labs they ordered back, to have an AI model to challenge them? Yup.


Thought this might be interesting.

 

[SolarYellow]

I think it's the same as with self driving cars. People have a hard time trusting AI drivers, but the AI would have to be pretty awful to be worse than humans. If robot cars took over all driving and were so bad they killed 10,000 people a year in the US... that would save 20-30,000 lives a year.

I have limited insight, but more than most.

My sense of it is that an attentive human is still vastly superior to any self-driving AI. Not even close. Where AIs do have an advantage is in not getting inattentive. They also don't make willfully bad decisions.

 

[Bravo52]

The deal after med school is the 3 to 5+ year residency training the new doc has to go through. It's not just graduate med school and done. Takes a lot more time thereafter to be certifiably ready for independent practice. AI will not work on its' own. Heck I used computer searches to help me with puzzling patients in the early internet age but since I had a medical degree, I could sort through what was B.S. and what wasn't. I highly doubt AI overall will be better than physicians. Docs are human and yes mistakes can be made. The key is to recognize them and do what was necessary to remedy them. Kurt

Agree @ksaves2. The shortened "Med School" essentially eliminates unneeded coarse work up front. The back end is still the same. AI won't be able to do it alone for sure but it can be programmed with the same data that medical schools teach and can comb through that data that the medical community produces at a much higher rate than a single human. It is only a tool. The problem is the interface between the data and the patient. As stated, I think the problem is "people". Not necessarily Drs or patients but the ability to "communicate" between people and the different aspects of AI and medicine. It is an interesting thought piece.

Of course this is all early in the AI phenomenon. I suspect over the coming years the AI will get better and better. It's certainly better now than it was yesterday. The other aspect of this is the acceptance and involvement of the humans. That will change over time as well, but probably not as fast as the AI.

I'm not a Dr. I'm not involved with AI. I've only read about the advancements but do think it is farther along than most realize. So yeah, this is all just an uninformed opinion.

 

[Azamiryou]

I have limited insight, but more than most.

My sense of it is that an attentive human is still vastly superior to any self-driving AI. Not even close. Where AIs do have an advantage is in not getting inattentive. They also don't make willfully bad decisions.

I agree. It's not the attentive humans that are killing 30-40,000 people a year on our roads.

AI doesn't have to be better than the best human driver or doctor — if it's better than the average human driver or doctor, it can make a big difference overall.

Unfortunately, due to the Dunning-Kruger effect, many people won't accept an AI for these jobs until it is better than the best human.

 

[hartlch]

I agree the problem is the interface between the patient and the AI. You can't do medicine by just asking questions, you have to be able to see and touch the body. Even having a person describe their problem will likely be too vague or use imprecise terminology ("stomach" vs "abdomen" for example). It seems like some sort of medical person would always have to be involved anyway.

I can see AI being a good at-hand reference for doctors that has the latest information on some topic. Keeping up with current research is a lot of work. But I would hope my doctor didn't have to refer to it too often. Things to worry about would be who trains the AI, would the AI be trusted too much, would insurance companies try to bias treatments based on cost, etc.

 

[teepot]

For about the last 10 years when I make a doctors appointment I see an NP. The majority of the doctors here are in a Inter Mountain Health Facility. It's bigger than our hospital. There are a dozen pods. The NP's see the patients. There is a doctor in each pod. So just like my experience yesterday. Talk to the NP, she talks to the doctor, rinse and repeat. There is Desert Radiology in the building and they can do x-ray, CT and MRI's. There is a quick care too. I think the level of care there is better than the care at the hospital. Will AI be in medicine? Yes. Can it help with health care in small towns like ours? Probably. I believe AI is inevitable and will be integrated into every facet of the human existence whether we like it or not. There is to much money at stake for it to be otherwise.

 

[cwbullet]

I agree the problem is the interface between the patient and the AI. You can't do medicine by just asking questions, you have to be able to see and touch the body. Even having a person describe their problem will likely be too vague or use imprecise terminology ("stomach" vs "abdomen" for example). It seems like some sort of medical person would always have to be involved anyway.

I can see AI being a good at-hand reference for doctors that has the latest information on some topic. Keeping up with current research is a lot of work. But I would hope my doctor didn't have to refer to it too often. Things to worry about would be who trains the AI, would the AI be trusted too much, would insurance companies try to bias treatments based on cost, etc.

Sort of. The key to medicine is asking the right questions. By asking the right questions, I can normally make the right diagnosis based on the patient's complaint/age/biological gender/race/risks. You may call it biases but diseases run in certain groups and ages. The exam and tests confirm what I have already put in my differential diagnosis list.

 

[cwbullet]

I agree. It's not the attentive humans that are killing 30-40,000 people a year on our roads.

AI doesn't have to be better than the best human driver or doctor — if it's better than the average human driver or doctor, it can make a big difference overall.

Unfortunately, due to the Dunning-Kruger effect, many people won't accept an AI for these jobs until it is better than the best human.

It is definitely not better than the average doctor, yet. It makes too many errors and focuses on the worst-case scenario. You certainly don't want every 20- to 30-year-old with chest pain to get a cardiac cath.

 

[Art Upton]

The same monitor has some filters to filter to out road noise and try to interpret the heart rhythm while CPR is being performed, it's still not reliable. There's a stretch of highway where the road vibrations will make the monitor think the patient is in ventricular tachycardia. Good thing it can't deliver a shock by itself.

We also have AI assisting our dispatchers ... that's a mess too.

There are these things called a life vest. They are a heart monitor and will give a warning before they Shock You.
I was warned that vibrations can set them off and not to drive a motorcycle or use a jack hammer while recovering.

Edit: Yes they said a Jack Hammer, not sure if that was part of the humor or not.

 

[Knundrm]

It's a double edge sword right? On one hand, AI could be the best method for diagnosing and creating a treatment plan, having all medical knowledge available at all times, but on the other hand, if it's controlled by the wrong people, it could eliminate our ability to make well informed decisions. It's coming though, either way.

Having worked in healthcare forever, I see the benefit of AI as possibility ensuring more consistent medical treatment protocols for patients and hopefully freeing up providers from a lot of the mundane aspects of their jobs (and improving provider retention in the system). Then again, I can hear a lot of Docs complaining that "we're taking the 'art' out of medicine". But something has to change. For at least 20 years I've heard Docs say, if I knew then what I know now, I would never have gone to medical school. And hopefully AI could help avert some of the cyber attacks occurring in healthcare these days.

 

[NTP2]

And hopefully AI could help avert some of the cyber attacks occurring in healthcare these days.

I think it would actually do the opposite, it’s a complex system that is on a computer, if you use it for online consultations it’s a line in as well.

 

[Knundrm]

I think it would actually do the opposite, it’s a complex system that is on a computer, if you use it for online consultations it’s a line in as well.

Probably so, but I have seen articles stating that AI could help identify/protect against some attacks (in the future). At a minimum, if it was effective at limiting medication errors or limit adverse pharmaceutical contradictions, it would be a win.

 

[John Kemker]

Better not get into an important [to the customer that is] IT job then. While not as Tramatic as a Doctor's job, I have been woke up at night or bothered in the evening much to my Wife's discontent since the mid 1980s. First Pagers then Cell Phones.

In about 1983 our phone rang at home and my wife while we were still in bed about 7am answered it. It was a Doctor that bought an Apple Clone computer from the store I worked for that had looked my name up in the phone book and expected me to trouble shoot his problem over the phone while I was in bed.

No it was not a medical emergency at all, not even a clerical emergency; just that he wanted tech support. It got much worse than that as the years went on and IT folks became "exempt" from hourly pay and put on "Salary" so they could be made to work around the clock.

Back in the 90s, I worked for the VAMC in Decatur, GA in IT (IRMS, at the time). Was called to Surgery to fix a non-working terminal. (This was back in the days of dumb terminals, mostly DEC VT420s and Wyse knock-offs.) Looked at the terminal and tried to reset it, no luck. Got under the desk and found the jack number, then went to the wiring closet to find out what terminal server and port number it was connected to. Found it, reset the port, then went back to the terminal. Only, it was gone. I asked, in a somewhat irritated, loud, voice, "Does anyone know what happened to the terminal that was here?" One of the doctors piped up, "Oh! It wasn't working, so I moved it to another spot to see if it worked there!"

I calmly told the doctor, "Dr., I don't move your patients. I would appreciate it if you wouldn't move mine." /smh

 

[cwbullet]

I see some mistakes made by medical personnel. Although, I am a skeptic, I welcome AI if they can make it reliable.

 

[hartlch]

Sort of. The key to medicine is asking the right questions. By asking the right questions, I can normally make the right diagnosis based on the patient's complaint/age/biological gender/race/risks. You may call it biases but diseases run in certain groups and ages. The exam and tests confirm what I have already put in my differential diagnosis list.

Yes, I totally agree. My point is that knowing the right questions can be influenced by the patient's appearance and actual hands on examination. A doctor can notice things that a patient wouldn't think to bring up or didn't notice ("that mole on your back should be looked at"). I can't picture how playing twenty questions with an AI would be effective way of doing medicine. As a doctor, you probably use the patient's appearance to guide your questions - "this guy is complaining of a cold, but sure looks like something else is going on". And that is the "art" part of medicine. I don't see AI replacing that any time soon. Maybe AI could be effective at some sort of triage to determine what level of medical attention you need?
My primary care person is a PA because the doctors in the group are overloaded. That has been fine for me but what passes for a "physical exam" these days is not that great. Mostly some questions like "how are you feeling" or "anything going on" and then a quick listen to the chest with a stethoscope. I asked if they were going to check ears/nose/eyes and they said "I'm not sure if I can find an otoscope around here". For us older folks, growing up the doc always checked all the orifices and felt around the abdomen. Not so these days. So maybe an AI could do what passes for a general physical and order some standard lab tests? Don't think I would trust it for anything beyond that.