U.S. CitiesCity-Data Forum Index
Go Back   City-Data Forum > General Forums > Great Debates
 [Register]
Please register to participate in our discussions with 2 million other members - it's free and quick! Some forums can only be seen by registered members. After you create your account, you'll be able to customize options and access all our 15,000 new posts/day with fewer ads.
View detailed profile (Advanced) or search
site with Google Custom Search

Search Forums  (Advanced)
Reply Start New Thread
 
Old 12-14-2013, 04:08 AM
 
140 posts, read 429,507 times
Reputation: 139

Advertisements

Many doctors already follow simple algorithms for most of their work (doing "cookbook" medicine). The AI is already advanced enough that it would outperform them. IBM's modified Watson is showing a lot of promise: "Wellpoint's Samuel Nessbaum has claimed that, in tests, Watson's successful diagnosis rate for lung cancer is 90 percent, compared to 50 percent for human doctors" (1). Already in the 1990, a neural network did better than physicians in diagnosing patients with acute myocardial infarction (2).

Even simple algorithms and basic knowledge base could offer large improvement in care vs human doctors in some cases. To take one example, only 53% of obstetrics and gynecology residents even knew that copper IUDs could be used for emergency contraception (3) and 30% of physicians and Title X clinic providers (such as nurse practitioners) had misconceptions about the safety of IUD for women without kids (4). Doctors often ignore evidence-based clinical practice guidelines. (5). There is no reason to think that a random doctor should be able to sensibly differ on the best way to do things with a team of experts backed by studies.

There is a profit motive in medicine as practiced by human doctors that increases costs. This recent investigation (6) illustrates how often doctors prefer to use a drug that costs $2000/injection vs another that costs $50/injection, even though six studies showed them to be largely equivalent. "Under Medicare repayment rules for drugs given by physicians, they are reimbursed for the average price of the drug plus 6 percent." Just this one case costs Medicare $1 billion per year and doesn't increase the quality of care. Properly programmed AI would not do that. And yes, I do believe that a committee of experts, insurance companies, Medicare, or the public would win out over this drug maker, when it comes to choosing what the software would do.

According to (7), 210,000 to 440,000 patients each year who go to the hospital suffer some type of preventable harm that contributes to their death. Computers will never make errors due to fatigue, unlike humans. Older doctors make a much larger amount of errors (8). This wouldn't apply to AI or properly maintained machines.

There are 1.7 million hospital-associated infections per year in the U.S. (9). Pneumonia is the second most common hospital-acquired infection, and it is often spread by health care workers (10). The infection can also be acquired in surgery due to contaminated instruments. A pilot program looking at Ambulatory Surgery Centers found that 68% of them had at least one lapse in infection control (11). Nurses' uniforms often carry dangerous bacteria (12). Many of these infections could be prevented through increased use of robots and computers.

The family doctors are likely the biggest cause for concern. For example: "The majority of the sample reported that they prescribe psychotropic medication to their patients, despite often assessing their knowledge of psychotropics as absent or marginal" (13). This should not be acceptable, and I'm quite sure that an AI-based system could already do better than these doctors. Recently, I had a sleep specialist tell me that the half-life of caffeine is 15 hours, which is of course incorrect for people without other factors, such as pregnancy or antidepressant use.

We rarely question doctors. They obviously have more knowledge and experience about the topic than we do, so it'd be stupid to do so. I haven't called out the sleep doctor about caffeine's half-life; I just noted that it didn't sound right and checked when I got home. We need to trust them, especially since we obviously see them when we feel sick and need help. We respect them highly, since they are dealing with life or death situations and help us. We think they are very smart, empathetic and humane (more on that later). Yet there is very little supervision, checks, or discipline. Errors are self-reported. The vast majority of providers employ a "deny and defend" strategy in the face of an adverse event (14). "One-third of physicians did not completely agree with disclosing serious medical errors to patients, almost one-fifth did not completely agree that physicians should never tell a patient something untrue, and nearly two-fifths did not completely agree that they should disclose their financial relationships with drug and device companies to patients." (15)

Most doctors do not follow the research, even in their specialty. I don't blame them - they simply don't have enough time to read as much as necessary. But this problem will only get worse as relatively more research gets published. Unfortunately, even if they followed the research, they would not be able to make informed decisions based on it. This study (16) shows they did just as poorly interpreting the results of a study comparing two drugs as their patients. It's not an egregious example by any means, but I was prescribed an off-label drug by a specialist who hasn't heard that a modified version of this drug was in Phase IV trials for this condition. The computers won't have this problem. The AI systems will absorb knowledge shared by top human experts in the field, not just follow published research.

Emergency department physicians spend twice as much time doing data entry vs. direct patient care (17). Even without completely replacing physicians, a semi-smart computerized assistant could greatly improve this ratio. Interns spent on average 8 minutes per day per patient (18), only 12% of their time. The largest group of practicing physicians spends between 13 and 16 minutes per patient (19). I'm sure I'm not the only one who thought that the doctor could be asking more questions that might improve the diagnosis and treatment and wouldn't mind spending more time at the doctor's office if there was a chance for improvement in these areas. And then, of course, there are other important things that patients want to know about, such as costs ("only 30% said they regularly discuss the issue"). This limitation wouldn't exist with AI-based systems. Hey, we waited 20 minutes on average in the waiting room already.

Human doctors forget things they learned in med school. The test performance decreases from 40% correct answers for students to 25-30% after many years of practice (20). Computers' memory won't have any losses.

There are many unconscious biases that affect the treatment of patients by humans that could be eliminated. For example, overweight physicians discuss weight loss less frequently with obese patients than doctors who have normal body mass indexes (18% vs 30%) (21).

We'd all like to think that our doctor graduated on top of her class from Harvard and is as smart as Dr. House. But as the old joke goes: "What do you call a person who graduates last in his class at med school? Doctor." Med school is mostly memorization and the graduation rates are incredibly high: over 96% within 10 years (22) and just 1.4% haven't graduated due to academic reasons. This compares to the 62% graduation rate in other post-graduate programs. The pass rate for the United States Medical Licensing Examination is 94% (23). This does not fill me with confidence that there is an adequate vetting process is in place. There is also no good way for patients to research doctor's skill. Are we confident enough that MCATs and GPA scores are enough to guarantee a quality MD? There is an artificial cap imposed by the government on the number of residency positions. However, even with the cap present, more than 95% of U.S. medical school seniors have matched to residency positions (24).

When it comes to surgery, there is a huge variation in operative skill between surgeons and the dexterity of a surgeon's hands can account for much of the difference in how well patients do (25). Watching videos like this (on the bottom) makes it pretty clear that robotically assisted surgery should come in the future, even if the devices in current use are still very expensive, there are serious concerns about their quality (see da Vinci systems), and are hard to evaluate (as is performance of human surgeons because of voluntary self-reporting with no oversight or enforcement). The hand tremor of humans in microsurgery would also stop being a concern. Has your surgeon exercised 3 hours ago? The luck is not on your side (26).

Doctors can be empathetic, but a machine won't. That's true. But there is a lot of dehumanization needed in order to stay an effective doctor anyway. While some students might enter the medical field for humane reasons, it often gets beaten out of them. This role can be better assumed by social workers or similar professions. "Dr. Daniel Chen has found in studies that students' empathy scores fell between the time they started medical school and the time they graduated. […]. Other researchers have uncovered similar patterns during residency training" (27). Medical students often witness their teachers act unethically and feel pressured to do so (28). Another potential "human" factor can be dismissed: patients do want to be told the truth about their condition (29).

Computer can read some human emotions more accurately than other humans already (30). There are even brainwave reading devices on the market, such as (31), so the accuracy could become very high, if that's needed.

Doctors' compensations rise faster than inflation (32). The wages of the support staff and other medical personnel are also raising quickly (33, 34). The cost of computers and robots, on the other hand, is decreasing (35). The performance and storage capabilities of AI systems and robots will only go up, and it can happen exponentially. The same cannot be said about humans in the medical profession.

Human doctors are not difficult to manipulate and often feel pressure from patients to prescribe brand name and other drugs. The TV ads for prescription drugs wouldn't have much of an effect otherwise. More than a third indicated they often or sometimes prescribed brand-name drugs when generic substitutes were available because patients requested it (36). Video-analysis software already appears to be able to detect deception better than human experts can (37).

The AI research and the robotics should revolutionize medicine by automating it, in the same way that self-driving cars will revolutionize transportation. There are differences between these two areas but some of them actually make it easier for the medical AI: dealing with unpredictable road and other drivers vs. mostly controlled environment in a medical office or hospital, having loss-of-control issues and feelings of helplessness when the computer drives vs. not having much control over the doctor's decisions anyway. We know that IBM's Watson can put together huge amounts of information. We know that the AI can beat us not only in chess, but also in games with hidden states, such as 1-on-1 limit hold'em poker. It is getting more and more difficult to create unbreakable CAPTCHA.

There should be a regulatory framework created to support and greatly accelerate AI and robotics medical research. Some states have been able to take steps in that direction for self-driving cars. The government could fund basic research in the area and create large X Prize-type competitions for companies that create inexpensive doctor-replacement level AI and robots. Savings and care quality improvements could be huge and pay for themselves quickly.

My guess is that such changes might first occur in countries other than the U.S. There are powerful lobby groups in the U.S., such as the AMA, that won't be happy with the results of these changes. There are many unwieldy regulations already in place. By the way, I'm not claiming that other unrelated changes to the U.S. healthcare system wouldn't have a large effect. I just wanted to address this topic. Let's build a better doctor.

Reply With Quote Quick reply to this message

 
Old 12-14-2013, 07:31 AM
 
Location: Beautiful Rhode Island
6,794 posts, read 11,057,429 times
Reputation: 6754
[quote=first one;32608123]

We rarely question doctors. They obviously have more knowledge and experience about the topic than we do, so it'd be stupid to do so. I haven't called out the sleep doctor about caffeine's half-life, I just noted that it didn't sound right and checked when I got home. We need to trust them, especially since we obviously see them when we feel sick and need help. We respect them highly, since they are dealing with life or death situations, and help us. We think they are very smart, empathetic and humane (more on that later). Yet there is very little supervision, checks, or discipline. Errors are self-reported. The vast majority of providers employ a "deny and defend" strategy in the face of an adverse event (14).

I agree with your description of the problems. Just the other day, in fact, I questioned a physician on some recent research in his field that he had not heard of!!

But I don't agree with your extreme solution. I think there should be integration of doctors with computer resource and backup- a trend that is in progress today anyway. Never before in human history has so much been at their fingertips in terms of information.

I always question my doctors and do plenty of independent research on their recommendations. I research every single drug extensively before taking it. I have refused to take certain medicines due to risk/benefit. All patients are their own best advocates.
Reply With Quote Quick reply to this message
 
Old 12-14-2013, 07:45 AM
 
9,762 posts, read 8,096,872 times
Reputation: 13188
Computers can't give human contact. I work at a major local hospital for the past 14 years. When patients come into the ER, it requires communication to find out what is wrong and how it happened. Symptoms can overlap many conditions. People regularly lie to ER staff. The staff listen to the patient's words and body language to get to the truth. The computers are an aid to help speed up diagnosing and treating patients. It's not a replacement for human staff inputting the data (symptoms, BP, temp, X-Ray, MRI, etc) and reading the data. Then when the patient is admitted, it takes human interaction to help speed up recovery, even if it's listening to the patient talk a few minutes at a time and a little physical contact (holding hand, touch on arm, or even a hand on the shoulder). Hoe does a computer respond to patients whose body was mangled by wrecks, penetrated by things (pipes, branches, lumber, etc), MRSA infection combined with injuries, unknown animal attack, or people with mental problems like OCD and hypochondria. With some people taking multiple drugs for recreational use, and overdose treatment for one drug may interact negatively with whatever drug the patient may have taken.
Reply With Quote Quick reply to this message
 
Old 12-14-2013, 04:10 PM
 
Location: Foot of the Rockies
86,886 posts, read 102,301,239 times
Reputation: 32946
Maybe nurses should work in the nude, since their uniforms carry all this dangerous bacteria.

From the source cited: "To be sure, the study doesn’t verify a link between the germy garb and actual patient infections, the authors say. . . .

But other infection experts say that there are some contamination sources that are far more worrisome than clothing or accessories.

“Uniforms could be a source of contamination, but there is more concern about other surfaces around the patients,” said Russell N. Olmsted, president of the Association for Professionals in Infection Control and Epidemiology.

“What we don’t want to do is direct a lot of energy to sterile attire,” he added.
"
Hospital garb harbors nasty bacteria, study says - Health - Infectious diseases | NBC News
Reply With Quote Quick reply to this message
 
Old 12-14-2013, 10:15 PM
 
48,516 posts, read 83,688,108 times
Reputation: 18036
Germs can be transmitted by air and surfaces of all types. but as to robot i think in time they will do much more common tasks. That is take care of a aging population in advanced countries. Japan is leading the research in that regard.With boomers being such a large market there are fortunes to be made in care services of any type they want/need.8K a day retiring to high of estimated 10k a day for next 19 years.26% of the present US population.
Reply With Quote Quick reply to this message
 
Old 12-15-2013, 11:32 AM
 
9,762 posts, read 8,096,872 times
Reputation: 13188
Quote:
Originally Posted by Katiana View Post
Maybe nurses should work in the nude, since their uniforms carry all this dangerous bacteria.

From the source cited: "To be sure, the study doesn’t verify a link between the germy garb and actual patient infections, the authors say. . . .

But other infection experts say that there are some contamination sources that are far more worrisome than clothing or accessories.

“Uniforms could be a source of contamination, but there is more concern about other surfaces around the patients,” said Russell N. Olmsted, president of the Association for Professionals in Infection Control and Epidemiology.

“What we don’t want to do is direct a lot of energy to sterile attire,” he added.
"
Hospital garb harbors nasty bacteria, study says - Health - Infectious diseases | NBC News
When treating highly infectious patients, staff don disposable gowns. Old hospitals can spread viruses if their ventilation system is outdated or set up improperly. Modern systems will not recirculate any air from patient rooms, will have a set of UV lights in duct to kill micro-organism, and multi-stage filter system. This isn't cheap and is electricly inefficient. Good staff training reduces spread of infections, from doctors down to housekeeping. I work in the boiler room and do general maintenance & even we get regular infection control training.
Reply With Quote Quick reply to this message
 
Old 12-15-2013, 12:00 PM
 
Location: Foot of the Rockies
86,886 posts, read 102,301,239 times
Reputation: 32946
^^And you think the nursing staff doesn't get such training? Good grief, for decades we've been hearing about the ten biggest ways to spread germs, our fingers. That article about nurses' uniforms was interesting, however, I think they ought to look at dr's lab coats as well. And this bit about not wearing them outside the hospital-the hospital is where all the bad stuff is. I do agree they should wear a clean uniform every day.
Reply With Quote Quick reply to this message
 
Old 12-15-2013, 04:47 PM
 
31,385 posts, read 32,016,053 times
Reputation: 14896
How strange...

I'm sitting here watching PBS's Frontline's documentary League of Denial: The NFL's Concussion Crisis. which high lights one physician in particular, Dr. Bennet Omalu a forensic neuropathologist who performed the autopsy on Mike Webster. Omalu, on his own initiative decided to keep the former Pittsburg Steeler's brain for further examination although it had nothing to do with his cause of death. It was just human curiosity. What Omalu found was the first link between foot ball brain injuries and chronic traumatic encephalopathy.

Now I am not some computer Luddite, I will accept the OP's claims on their face that computers can be far more accurate at making the correct diagnosis at a higher rate than a human doctor, but my question, asked with a great deal of skepticism, is the ability of a computer to ask the right questions or more to the point, the basic curiosity to ask questions that have never been contemplated. After all, a computer is only as good as the data that it is given.

I don't know of a computer that can coax information out of a person as well as another human and that is what the best doctors are able to do, to get patients to trust them with the most intimate details of their lives, much less the ability to take subjective information, read between the lines and dig a little further to arrive at enough information to arrive at objective facts, facts that certainly can be introduce to a computer for analysis. But a computer on its own, without a physician interacting at the same human to human level that is the epitome of what a doctor patient relationship is based upon, I don't think so.
Reply With Quote Quick reply to this message
 
Old 12-15-2013, 04:51 PM
 
Location: Texas
43,409 posts, read 52,413,699 times
Reputation: 70378
Computer can get at best 30% of the info...you have to know what they are not saying and your gut...algorithms are just where you start.

Of you think medicine is just a bunch of protocols, you are way off base.
Reply With Quote Quick reply to this message
 
Old 12-15-2013, 05:00 PM
 
Location: Beautiful Rhode Island
6,794 posts, read 11,057,429 times
Reputation: 6754
And this is why medicine is considered primarily an art and then a science.
Reply With Quote Quick reply to this message
Please register to post and access all features of our very popular forum. It is free and quick. Over $68,000 in prizes has already been given out to active posters on our forum. Additional giveaways are planned.

Detailed information about all U.S. cities, counties, and zip codes on our site: City-data.com.


Reply
Please update this thread with any new information or opinions. This open thread is still read by thousands of people, so we encourage all additional points of view.

Quick Reply
Message:

Over $104,000 in prizes was already given out to active posters on our forum and additional giveaways are planned!

Go Back   City-Data Forum > General Forums > Great Debates
Follow City-Data.com founder on our Forum or

All times are GMT -6. The time now is 01:33 AM.

© 2005-2019, Advameg, Inc. · Please obey Forum Rules · Terms of Use and Privacy Policy · Bug Bounty

City-Data.com - Archive 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35 - Top