Could a computer do a doctor's job?

[GoodGame]

---

Members don't see this ad.

Exclude surgeons of course. For those, we would have to consider robots but that's the subject of another discussion.

My argument lies in the idea that the job a doctor does is based on if-then statements. If patient comes in with symptom x, do procedure y, or administer treatment z. It's all based on a huge hypothetical flowchart, and this is exactly what computers are best at. It's their modus operandi.

Additionally, collectively computers could gather much more experience than any individual doctor could. If we had 'doctor computers' installed in medical offices, these computers would all work together and share data in the cloud. Thus any patient they look at, any abnormality they observe via their cameras, would be instantly uploaded, processed, and compared with thousands of similar cases. Statistically speaking, the computers would then have a very high chance of correctly diagnosing the patients.

What do you guys think? Could a computer, *in theory*, replace a doctor?

There are like a million reasons why this idea would not work so be nice. I just want to hear your opinions.

 

[AntiHouseMD]

The last time we had a thread like this, it turned into a GIF thread. So, to start:

 

[aSagacious]

Nope. There's too much grey area that requires intuition or discretion by the doctor. For example let's consider something as simple as blood pressure. 120/80 = normal.

Under your proposed system...

If:
patientPressure > normalPressure + 20
Then:
prescribe drugs ABC
Else if:
patientPressure < normalPressure - 20
Then:
prescribe drugs XYZ
Else:
do nothing

...seems straightforward enough. The problem is some folks (like athletes) walk around with 90/60 no problem, while that might send an elderly woman into shock. Similarly an obese man might regularly hover around 160 with no acute issues while the athlete would have an aneurysm under these conditions.

If dealing with something as straightforward as blood pressure gets this messy, imagine how poorly the computer would handle more multi-faceted cases.

 

[Eric01]

You could program a computer to take in symptoms and process a decision based off of risks etc. but you would not be able to get a computer to have qualitative judgment.

This would lead to both excessive amounts of tests in uncritical situations, and lack of tests in critical situations. Also a lot of differential diagnosis don't follow a specific algorithm.

Case in point, a patient comes in suffering from chronic dizziness. That would blow a computer away.

 

[GoodGame]

... and that's where you failed.

[/thread]

Explain how it is not so.

 

[AntiHouseMD]

Tell me how quickly this computer of yours would diagnose a patient. Here's a hypothetical, a patient comes in with nausea, shortness of breath, and weakness. It's an aortic dissection (rare, I admit, but just follow me). As the computer is trying to sort everything out, the patient starts crashing. A doctor could sense the change and quickly start providing treatment. With a computer, it would have to notify a human to start emergency treatment. It takes longer and might be too late at that point. One of the doctor's most important strengths is that he/she can provide acute medical care in the case of an emergency. I can't see how a computer could do the same thing.

As anyone has said so far, there are too many variables in medicine to be left to a computer.

 

[ppfizenm]

Explain how it is not so.

If you really need to ask then medicine probably isn't for you. We already have patients thinking they can do a doctors job with web md, we don't need doctors thinking the same.

 

[Perrotfish]

Tell me how quickly this computer of yours would diagnose a patient. Here's a hypothetical, a patient comes in with nausea, shortness of breath, and weakness. It's an aortic dissection (rare, I admit, but just follow me). As the computer is trying to sort everything out, the patient starts crashing. A doctor could sense the change and quickly start providing treatment. With a computer, it would have to notify a human to start emergency treatment. It takes longer and might be too late at that point. One of the doctor's most important strengths is that he/she can provide acute medical care in the case of an emergency. I can't see how a computer could do the same thing.

Robotics aside, the computer would call the same nurse the physician would, and would give her the same instructions. Physicians don't generally actually DO anything. The only difference is that the computer could check on the patient several hundred times a second, while physicians round (hopefully) 3 times/day.

For my money, yes, this happens over the course our generation. Radiologist go first, the Gas, then it's a race to see if surgeons are replaced by robots before a medical program can do a regular, conversant history and physical. Medicine becomes a research and engineering profession, focused on what we don't know how to do rather than what we do, which ironically will justify America's currently nonsensically long training cycle for physicians.

Of course, I think that technology is going to get so wierd over the course of our generation that no notices a change as small as this. I suspect that our traditional concept of what consitutes a human, as well as the clear demarcation between human and computer, are going to be severely challenged within the next 50 years. An operatorless robotic surgery is the least of my concerns.

 

[aSagacious]

... and we have our first invoking of Burnett's Law at post #8 folks.

 

[GoodGame]

Nope. There's too much grey area that requires intuition or discretion by the doctor. For example let's consider something as simple as blood pressure. 120/80 = normal.

Under your proposed system...

If:
patientPressure > normalPressure + 20
Then:
prescribe drugs ABC
Else if:
patientPressure < normalPressure - 20
Then:
prescribe drugs XYZ
Else:
do nothing

...seems straightforward enough. The problem is some folks (like athletes) walk around with 90/60 no problem, while that might send an elderly woman into shock. Similarly an obese man might regularly hover around 160 with no acute issues while the athlete would have an aneurysm under these conditions.

If dealing with something as straightforward as blood pressure gets this messy, imagine how poorly the computer would handle more multi-faceted cases.

You're right, the whole process is 'messy.' But that does not mean a computer cannot handle it. For example, in your blood pressure hypothetical case, the computer could easily take into account the patient's age, BMI, occupation, etc.

Then it could take all of this data and compare it against a nationwide database. Thus it could check that out of one million athletes who visited medical offices in the past month, 900,000 have a BP that is within one standard deviation of this athlete's BP.

Of course, this is a simplified example. The real program would be much much more complex, but the whole idea boils down to 'teaching' the computer to correctly process this kind of information, and because computers are incredibly fast, I believe they are capable of 'learning' it.

Now before you say that's BS let me remind you that computers can only do five things: add, subtract, multiply, divide, and shift numbers around by transfering them to and from memory. Every single computer on this planet is basically a complex calculator. Who would have thought that based on these five operations we could be where we are today where we can use computers to communicate through the internet, watch movies, play advanced video games, listen to digital music, render video wirelessly, and ultimately, simulate reality through various AI engines.

At the core of it, humans are somewhat similar to computers because we use electrical signal patterns to make realizations and decisions. We are a lot more advanced because we can do a lot more operations based on our numerous amounts of neurons. Our processing 'capacity' is basically !neurons. Thus computers as they are today cannot even come close to our level of intelligence, but there are certain similarities at the core of it that can be exploited to make technology at least mimic our thoughts. We've done it so far with space travel, the internet age, and the early AI engines that are in use today. Why can't we extend the system to a well known medium such as medicine.

 

[AntiHouseMD]

Well it depends. If you wondered if a computer could "completely" take over for a doctor, then I would say no. You still have the issue of examining a patient manually, learning about their past history, etc. Now, I could see a system where you plug in the symptoms and the computer diagnoses and gives a treatment for the patient. The computer could also take symptoms like blood pressure, heart rate, and so forth. With your theory, though, I would worry about how a computer would detect something such as a tumor on the prostate, cervix, etc. Personally, I would prefer a doctor over a computer. But, it's an interesting question.

Also, just for argument's sake, let's say there is a power outage at a hospital for an extended period of time. How would you take care of that issue?

I'm not saying that your question isn't valid. Given some of the creepy tech stories I see (sex robots, dancing robots, and only God knows what else), nothing would shock me anymore. But, at least right now, I just see too many issues. In the future, maybe. However, I personally doubt it.

 

[aSagacious]

Ok, how about assessing the patient for symptoms?

Can a computer recognize when the patient experiences an altered mental state? Can it identify a lump in someone's breast? How about generalized edema? Or any qualitative metric for that matter?

 

[xadmin]

Yes it can. Look, a computer is grading your Essay now and a computer is driving you around in a car with the google car. So yes a computer can diagnose your symptoms and prescribe medicine to you. Although with physical exam, it will be harder, but soon our technology will get there and the computer can do that too.

Exclude surgeons of course. For those, we would have to consider robots but that's the subject of another discussion.

My argument lies in the idea that the job a doctor does is based on if-then statements. If patient comes in with symptom x, do procedure y, or administer treatment z. It's all based on a huge hypothetical flowchart, and this is exactly what computers are best at. It's their modus operandi.

Additionally, collectively computers could gather much more experience than any individual doctor could. If we had 'doctor computers' installed in medical offices, these computers would all work together and share data in the cloud. Thus any patient they look at, any abnormality they observe via their cameras, would be instantly uploaded, processed, and compared with thousands of similar cases. Statistically speaking, the computers would then have a very high chance of correctly diagnosing the patients.

What do you guys think? Could a computer, *in theory*, replace a doctor?

There are like a million reasons why this idea would not work so be nice. I just want to hear your opinions.

 

[AntiHouseMD]

Yes it can. Look, a computer is grading your Essay now and a computer is driving you around in a car with the google car. So yes a computer can diagnose your symptoms and prescribe medicine to you. Although with physical exam, it will be harder, but soon our technology will get there and the computer can do that too.

That would be something that would do an extremely uncomfortable prostate exam.

 

[ppfizenm]

I remeber a study being done to see whether a computer was better than a doctor at reading ekgs and determining if the patient was experiencing a heart attack or not. They both could only do it a little more than 50% of the time.

 

[GoodGame]

If you really need to ask then medicine probably isn't for you. We already have patients thinking they can do a doctors job with web md, we don't need doctors thinking the same.

What? I don't think it's my job to know every detail of a doctor's occupation as a pre-med. If it was, I wouldn't need to go to medical SCHOOL.

If my question is really so preposturous as to make you question my capabilities as an MD, then by the same line of logic I can extend the argument and say that you will not be a capable doctor due to the logical fallacies you make in your posts, which undoubtedly extend in real life, and will extend into your profession.

 

[aSagacious]

What? I don't think it's my job to know every detail of a doctor's occupation as a pre-med. If it was, I wouldn't need to go to medical SCHOOL.

If my question is really so preposturous as to make you question my capabilities as an MD, then by the same line of logic I can extend the argument and say that you will not be a cable doctor due to the logical fallacies you make in your posts, which undoubtedly extend in real life, and will extend into your profession.

Hello and welcome to the internet, we hope you enjoy your stay. 🙂

 

[kpcrew]

That would be something that would do an extremely uncomfortable prostate exam.

somewhat relevant:

source: http://ddxcomics.com/2011/07/digits/

 

[AntiHouseMD]

What? I don't think it's my job to know every detail of a doctor's occupation as a pre-med. If it was, I wouldn't need to go to medical SCHOOL.

If my question is really so preposturous as to make you question my capabilities as an MD, then by the same line of logic I can extend the argument and say that you will not be a cable doctor due to the logical fallacies you make in your posts, which undoubtedly extend in real life, and will extend into your profession.

 

[kpcrew]

What? I don't think it's my job to know every detail of a doctor's occupation as a pre-med. If it was, I wouldn't need to go to medical SCHOOL.

If my question is really so preposturous as to make you question my capabilities as an MD, then by the same line of logic I can extend the argument and say that you will not be a cable doctor due to the logical fallacies you make in your posts, which undoubtedly extend in real life, and will extend into your profession.

 

[GoodGame]

Ok, how about assessing the patient for symptoms?

Can a computer recognize when the patient experiences an altered mental state? Can it identify a lump in someone's breast? How about generalized edema? Or any qualitative metric for that matter?

NO! In the name of America, I shall never let this happen, and if necessary I will check the breasts of every woman in the country myself. You can thank me later for my sacrifices brothers:

 

[Perrotfish]

Also, just for argument's sake, let's say there is a power outage at a hospital for an extended period of time. How would you take care of that issue?

What do you think would happen in a modern hospital right now if the power went off and stayed off?

 

[AntiHouseMD]

NO! In the name of America, I shall never let this happen, and if necessary I will check the breasts of every woman in the country myself. You can thank me later for my sacrifices brothers:

Perrotfish, I would imagine that the generators would take care of it. I was thinking more along the lines of if our entire medical system was converted solely to computers and doctors were no longer used.

 

[AntiHouseMD]

NO! In the name of America, I shall never let this happen, and if necessary I will check the breasts of every woman in the country myself. You can thank me later for my sacrifices brothers:

Ah, just found the Youtube clip. Okay then, I just found your first patient. Good luck, brother.

[YOUTUBE]http://www.youtube.com/watch?v=eWiWrrNxA9s[/YOUTUBE]

 

[Perrotfish]

Perrotfish, I would imagine that the generators would take care of it. I was thinking more along the lines of if our entire medical system was converted solely to computers and doctors were no longer used.

Well if the generators could take care of it they could take care of the theoretical computers too. It's not like the doctors have any hand in running the hospital's back-up power grid.

 

[AntiHouseMD]

Well if the generators could take care of it they could take care of the computers too. It's not like the doctors have any hand in running the hospital's back-up power grid.

Very true. I'll concede.

 

[kpcrew]

Very true. I'll concede.

very gracious.

 

[thestrokes14]

There was research done at Columbia University that tested the supercomputer Watson's (from jeopardy) ability to diagnose diseases based on symptoms. It was actually extremely accurate. I think computers can be used for simpler things, but as things become more complex, doctors are certainly needed.

 

[JogoBonito]

as long as the computer still gives me a lollipop after the check up i m sold.

There was research done at Columbia University that tested the supercomputer Watson's (from jeopardy) ability to diagnose diseases based on symptoms. It was actually extremely accurate. I think computers can be used for simpler things, but as things become more complex, doctors are certainly needed.

I ve always thought that computers using symptoms to diagnose disease was like the same thing as just typing symptoms into a better version of web md. then we really don't even need doctors anymore right?

But I think people could never get the peace of mind that comes from a doctor's diagnosis then from some computer's analysis.

 

[Illinois2012]

Exclude surgeons of course. For those, we would have to consider robots but that's the subject of another discussion.

My argument lies in the idea that the job a doctor does is based on if-then statements. If patient comes in with symptom x, do procedure y, or administer treatment z. It's all based on a huge hypothetical flowchart, and this is exactly what computers are best at. It's their modus operandi.

Additionally, collectively computers could gather much more experience than any individual doctor could. If we had 'doctor computers' installed in medical offices, these computers would all work together and share data in the cloud. Thus any patient they look at, any abnormality they observe via their cameras, would be instantly uploaded, processed, and compared with thousands of similar cases. Statistically speaking, the computers would then have a very high chance of correctly diagnosing the patients.

What do you guys think? Could a computer, *in theory*, replace a doctor?

There are like a million reasons why this idea would not work so be nice. I just want to hear your opinions.

Looks like you answered your own question pretty well in the first post.

 

[ppfizenm]

What? I don't think it's my job to know every detail of a doctor's occupation as a pre-med. If it was, I wouldn't need to go to medical SCHOOL.

If my question is really so preposturous as to make you question my capabilities as an MD, then by the same line of logic I can extend the argument and say that you will not be a cable doctor due to the logical fallacies you make in your posts, which undoubtedly extend in real life, and will extend into your profession.

God, I hope I won't end up a cable doctor. Could you imagine.

I shadow a GP and he let's me conduct the physicals. There are certain things a computer just isn't capable of doing. Just like dictionary.com isn't capable of writing a post for you. If you go into an interview claiming computers are better than doctors you'll never be accepted anyway.

 

[Perrotfish]

I shadow a GP and he let's me conduct the physicals. There are certain things a computer just isn't capable of doing. .

Yet.

 

[vin5cent0]

Tell me how quickly this computer of yours would diagnose a patient. Here's a hypothetical, a patient comes in with nausea, shortness of breath, and weakness. It's an aortic dissection (rare, I admit, but just follow me). As the computer is trying to sort everything out, the patient starts crashing. A doctor could sense the change and quickly start providing treatment. With a computer, it would have to notify a human to start emergency treatment. It takes longer and might be too late at that point. One of the doctor's most important strengths is that he/she can provide acute medical care in the case of an emergency. I can't see how a computer could do the same thing.

As anyone has said so far, there are too many variables in medicine to be left to a computer.

Kind of a dumb argument IMO. How many doctors could honestly save your life if that just randomly happened on the spot while they were looking you over? I sure as hell wouldn't trust any of my regularly-seen doctors to be able to handle that situation.

 

[UnclePhil]

My argument lies in the idea that the job a doctor does is based on if-then statements. If patient comes in with symptom x, do procedure y, or administer treatment z. It's all based on a huge hypothetical flowchart, and this is exactly what computers are best at. It's their modus operandi.

My goodness, you might be on to something here. Why waste 4 years in medical school, and another 3+ years in residency when you just solved the entire process of practicing medicine in a few sentences.

 

[BigRedBeta]

Will it happen eventually? Probably, look into any discussion about the singularity or the concepts of Roy Kurzweil and there's plenty of evidence to suggest that the pace of technological advancement will exceed what we can currently estimate.

The most severe limitations presently will be in robotics.

The hardest problem will be the acutely decompensating patients, and the patient who, an experienced clinician just gets an ominous feeling about. Repeated studies have continued to show that one of the most important prognostic factors is clinical exam upon presentation. New interns call this learning "sick" vs. "not sick". There may not be any vital sign abnormalities but the experienced clinician, just by looking, can say "something's not right here". That's a level of sophistication that is probably impossible to program - however the "learning" computers currently be developed offer potential.

Any current resident can give you some common diagnostic dilemmas that require a high index of suspicion and that purely "criteria" driven protocols can have disastrous results.

An easy example is the new onset type one diabetic child presenting in DKA versus the child in septic shock. Both can appear really, really sick with delayed capillary refill and a need for fluid resuscitation. The problem is that in septic shock, you need to push 60ml/kg as quickly as possible (in under 15 minutes if possible), with the possibility of extreme morbidity or mortality likely if you don't, while in DKA that is one of the absolute worst things you can do, and in fact, if you did do it, you could actually put the patient on to a path that ONLY ends with cerebral edema, herniation and death. Making the wrong decision leads to awful, awful outcomes and even the death of a child. You're probably thinking "oh just check a bedside glucose", but in peds, given the extremely low yield since very few children have diabetes and because we actively try to avoid sticks in peds if we can and you're under pressure to act quickly, it's difficult to make a protocol out of the situation. It's not adult medicine where a 1/3 of your patients have issues with insulin resistance, pre-diabetes, etc. The combination of critical decision making, judicious use of resources, relative rarity of conditions, and severely bad outcomes for poor decisions make this a particularly great example of the nuance required of physicians.

It may be possible to have computers/robots eventually. But I think despite whatever advances may come, the real limiting factor is actually going to be medmal and liability issues. There are strong market pressures that will limit many, many, many robotics/computer companies from taking part because of their exposure to getting sued. It's likely that only one single court case may be so devastating as to set the whole field back by decades.

 

[AntiHouseMD]

It may be possible to have computers/robots eventually. But I think despite whatever advances may come, the real limiting factor is actually going to be medmal and liability issues. There are strong market pressures that will limit many, many, many robotics/computer companies from taking part because of their exposure to getting sued. It's likely that only one single court case may be so devastating as to set the whole field back by decades.

One of the few times that medical malpractice lawyers may actually help doctors.

 

[ppfizenm]

The entire point is moot anyway when you consider the fact the AMA would never let that happen. AMA lobbyists would have any tech that replaces a doctor illegal through some means.

If you need an example of this consider how the AMA got midwifery outlawed in 30 some states when research shows they do just as well as doctors when given hospital privileges.

 

[JESSFALLING]

People, correction: Most People, like people. With that in mind, I think that it's going to be a while before docs are obsolete. Rather, the technology revolution will simply continue to augment the practice of medicine. Sure, it may date a few jobs, but I don't think that the field will become entirely dehumanized in my lifetime, if ever.

 

[BJJ]

If you really need to ask then medicine probably isn't for you. We already have patients thinking they can do a doctors job with web md, we don't need doctors thinking the same.

Shut it. It isn't up to you to decide or even speculate on who should be a physician or not based on solely one statement or observation.

 

[ppfizenm]

Shut it. It isn't up to you to decide or even speculate on who should be a physician or not based on solely one statement or observation.

This is just a ridiculous statement. There are certain things that anyone is qualified of offering an opinion on, regardless of experience. If you went into an interview saying I think we should put the mentally handicapped to work on farms (actual example of someone I know from a mock interview that the interviewer led him into saying) there is no way the school will accept you. Saying "I think a computer can do your job" to a doctor is just as entangling. You might then get a question like "If a computer can do this why couldn't it run the country" or "Why should we waste our time teaching you medicine when a computer can do it better" You'll end up digging yourself into a hole and getting the rejection. It doesn't take adcom or doctor to see this. I think I am perfectly qualified to judge how this situation would turn out.

Now if you were applying for an engineering PhD this probably would be a more kosher opinion to have.

 

[235788]

The entire point is moot anyway when you consider the fact the AMA would never let that happen. AMA lobbyists would have any tech that replaces a doctor illegal through some means.

If you need an example of this consider how the AMA got midwifery outlawed in 30 some states when research shows they do just as well as doctors when given hospital privileges.

not all research is good research. They are nurses...

 

[ppfizenm]

not all research is good research. They are nurses...

Up until a few decades ago doctors made it impossible for a nurse to give a simple injection without the doctor being present. There are certain jobs that the AMA just won't allow other people to do b/c it takes money/jobs/control away from the doctors.

Regardless of what you think of the research they won't let themselves become obsolete and they have the money to make sure that doesn't happen.

 

[Bacchus]

Someone mentioned if-then statements above. Intending not to flame, this is what a lot of mid-level providers and nurses operate on. If BP is 130/90, institute anti-hypertensive. Do it several hundred times and you will follow that pattern forever. A doctor doesn't exclusively work on pattern recognition. There are algorithms, sure, but medicine doesn't always work on an algorithm. You need to assess a patient, his baseline, and modify accordingly, leaving the pattern you've established if the patient needs more customized care.

So, no, computers will not replace physicians.

 

[Stumpyman]

My dad is an engineer and hes actually working on a major project in IBM to make a completely computerized hospital.... I would love to elaborate more on the interesting topic later when im on a comp.

 

[isoquin]

What you are describing is a pattern recognition software that would lend itself well to non-procedural internal medicine. I think it's going to be difficult to convince some doctors to use such predictive software, but evidence may help. Nonetheless, we do not have the tech to have robots listening to lungs or putting in central lines. Even if we did, the role of doctor would not be obsolete, but changed.

 

[HealthHare]

 

[JogoBonito]

yeah yeah, humans will eventually be taken over by robots... or apes. or the polar ice cap melts and everybody dies except for kevin costner...i mean the fishmen.

but computers will never replace doctors!!!!

 

[GoodGame]

Eh some of you guys are saying it couldn't happen because of the logistical problem of implementing such a computer architecture into our existing healthcare network. I wanted to argue more along the lines of whether it is theoretically possible for a computer to perform the duties of a doctor.

 

[BJJ]

This is just a ridiculous statement. There are certain things that anyone is qualified of offering an opinion on, regardless of experience. If you went into an interview saying I think we should put the mentally handicapped to work on farms (actual example of someone I know from a mock interview that the interviewer led him into saying) there is no way the school will accept you. Saying "I think a computer can do your job" to a doctor is just as entangling. You might then get a question like "If a computer can do this why couldn't it run the country" or "Why should we waste our time teaching you medicine when a computer can do it better" You'll end up digging yourself into a hole and getting the rejection. It doesn't take adcom or doctor to see this. I think I am perfectly qualified to judge how this situation would turn out.

Now if you were applying for an engineering PhD this probably would be a more kosher opinion to have.

NOT the point I was making. If someone thinks that way, well then that is the whole point of the interview process to weed them out. Obviously there are people who would make better physicians than others. My point is that a lowly pre med such as yourself (and myself) have no right to say anything. You are not choosing who gets into medical school, the admissions staff is. It is up to them choose who would make a good physician, NOT a self entitled pre med.

 

[ppfizenm]

NOT the point I was making. If someone thinks that way, well then that is the whole point of the interview process to weed them out. Obviously there are people who would make better physicians than others. My point is that a lowly pre med such as yourself (and myself) have no right to say anything. You are not choosing who gets into medical school, the admissions staff is. It is up to them choose who would make a good physician, NOT a self entitled pre med.

O is that what the loosely veiled insults and the shut up meant?

I do like how you've both conceded the point I was making and admonished me for holding it in the same post.

This is a forum of premeds on a pre-allopathic message board. Of course I have a right to offer an opinion on the matter. He was asking a question and I answered it. Nowhere does it say please don't answer if you think I'm wrong for holding this opinion or even adcoms and med students only.

I'm not the one picking who get's to go to med school obviously but that doesn't make me any less able to recognize an opinion or belief or application flaw that will get someone sorted out.

Common sense isn't gold but I am sure glad I've been hoarding it.