Marcus Welby, meet Siri

by liberal japonicus

Was interested in this NPR article, which had this:

Millennial physicians like Ho are taking over hospital wards and doctors' offices, and they're bringing new ideas about life-work balance and new technologies.

One time, a patient asked Ho if it was OK if he recorded her performing a minor surgical procedure.

"He Instagram-videoed the entire procedure," she says. "It's not that a senior physician couldn't do it — I think that they might not have the comfort level."

She means comfort with technology. Millennial doctors want offices that are high-tech. Many have never worked with paper charts and they don't read dusty medical journals — they look at them online.

"We absolutely consult Wikipedia, not the library to find the most up-to-date medical research," she explains.

As you might imagine, that last line became one of those typical arguments about Wikipedia as the font of all human knowledge versus Wikipedia as Satan's ape. But I'm sure that the folks here are going to bring a lot more nuance to this. So prove me wrong...

Comments

[Hartmut]

I find wikipedia extremely useful as a starting point but it's highly dependent on the topic. As for libraries the problem is that scientific journals these days are so extremely expensive that your typical university can not even afford to hold the full set of the necessary ones. The three (main) universities of Berlin have some years ago started to pool their resources and/or to agree which one will hold which journal, so the range could be kept as wide as possible. Still there is a lot of semi-legal stuff going on along the line of "I know your institution holds journal X. Could you make a copy of paper Y for me?" or "I hear you will visit institution Z. Could you sneak into their library and make a copy of paper Y in journal X?" The publishers hate that and try everything to suppress it. But even Nobel laureates become thieves and breakers of copyrights when a single issue of a journal can on occasion cost a worker's monthly wages.
On many occasions scientific authors have to pay to use their own works because the publisher owns all the rights to them.

[Brett Bellmore]

Wikipedia is fine, so long as the topic is not politically charged, and thus subject to the edit wars.

Personally, though, I'm looking forward to Marcus Welby meets Watson. The brute fact about modern medicine, is that there's too much relevant information for any one person to learn, and if you're a practicing physician, you'll hardly be able to pretend to keep up. It's disturbing when you meet with your doctor about something, and you're more up to date on the latest relevant research than they are, because you only had your own condition to research, and he had every patient's.

Oh, and annoyed me that the local cancer center hadn't sprung for the recording feature on their Da Vinci surgical 'robot'. I was so hoping for home videos of my cancer surgery. At least I got to watch my cataract surgery... They keep you awake for it, but give you some remarkable tranquilizers!

[JakeB]

I find myself remembering the cosmetic surgery scene in Logan's Run now.

[Older]

"Remarkable tranquilizers' -- that is the truth! I couldn't stop talking about them, they were so amazing.

Brett, how did you get to watch your surgery? Well placed mirror? I used to be the kind of person who had to, really *had to*, watch things that were done to me, especially surgeries. But then there came a day when I really truly could not, and I had to adjust. It helps that every possible surgery is on YouTube, and that's where I saw my cataract surgery. (I also e-mailed links to all my friends and relatives, but that was before I finally learned that most people don't actually want to know the details.)

[thompson]

I pretty much agree with Hartmut, here:

I find wikipedia extremely useful as a starting point but it's highly dependent on the topic.

And I will say, almost anything regarding medicine is woefully out of date and or overly simplistic, at least for research purposes. At least, any subject I reach for, I certainly haven't performed an exhaustive analysis.

Giving the quoted doctor some benefit of doubt, she probably doesn't limit her research to Wikipedia. However, I will say I've met some clinicians that would.

And Brett:

It's disturbing when you meet with your doctor about something, and you're more up to date on the latest relevant research than they are, because you only had your own condition to research, and he had every patient's.

Disturbing or not, this is pretty much right, and why patients should be given resources to manage their own conditions. I've had family members catch prescription interactions that their doctors missed...its especially dangerous when there are multiple physicians. Even with the best intentions, it can be difficult to coordinate between doctors for each patient.

[Brett Bellmore]

"Brett, how did you get to watch your surgery?"

How do you think I got to watch it? I was conscious, and they were operating on my eyes. Not like I could have that operation with them closed.

Being extremely nearsighted prior to the surgery, I had a perfectly clear view of what was going on until they started removing my lens. Then things got a bit blurry.

[Turbulence]

Even with the best intentions, it can be difficult to coordinate between doctors for each patient.

Really, it is not. Any electronic medical record system should be able to easily do that. It is genuinely weird how technologically primitive so much of American medicine is, especially since we have a really awesome EMR for automobiles.

[wj]

Turb, the explanation is really pretty simple. The automobile systems got set up, and the mechanics were not given a choice about using it.

But with medical record systems, the doctors have the power to refuse to use them. And, in a lot of cases, have done exactly that. People, including doctors, really do dislike changing the comfortable way they have always done things. So they will only do so if they see a benefit to them. If the benefit is merely to their patients? Less of an incentive to deal with the pain of changing.

[russell]

They keep you awake for it, but give you some remarkable tranquilizers!

This Thanksgiving season, I am thankful for Versed. It makes colonoscopies entertaining.

"What's that, doctor?"
"That's your poo".
"Wow!"

True story.

[dr ngo]

My son - known to some of you as the inestimable Anarch - works for a company that designs electronic medical record systems, one of the best. And he troubleshoots for them here and there across the USA. And its a very good system BUT there are teething problems, not to mention (as WJ already has) the reluctance of some health care providers to fully buy into the system, and idiosyncratic kinks unique to each hospital (and sometimes to each patient) . . . In short, this is an extremely good step in the right direction, but it is NOT a panacea.

And we (my wife and I) get our health care from one of the hospitals served, and they're very very good (people come from miles and miles away for treatment here) and she still does her own research on drug interactions, because she's got so many diverse ailments, and nobody but nobody can keep track of all the combinations unless they're looking for them, even if everything is properly listed on the electronic system . . . In fact one of her (many fine) doctors advised her to keep on doing this, just in case she finds something that everyone else has missed.

And the human body is apparently much more complicated than the automobile. Who knew?

[wj]

Not only is the human body more complicated than an automobile. But the manufacturers don't seem to have embraced the installation of computer chips (with diagnostics) throughout the mechanism as original equipment. Obviously an oversight which will require retrofitting at some point....

[thompson]

Turb:

Any electronic medical record system should be able to easily do that.

No.

Mostly, what dr ngo and wj said. I'm all for EMR, but they aren't some magical panacea. There are some drug interactions that can be readily caught and screened against by an automated system, but there are many things that are just too context sensitive and there are no hard and fast rules.

Especially with an aging population that has multiple chronic conditions progressing at once. It's not simply a matter of 'is patient taking X? Don't give them Y'. It's calibrating the dosing of X, Y, and Z because their kidneys are failing, as is their heart, but you still need to treat their severe arthritis without aggravating their colitis too much while treating them for their pneumonia. That's damn hard to do for a single doctor, and harder still when each of those conditions is being treated by a separate specialist.

It is genuinely weird how technologically primitive so much of American medicine is, especially since we have a really awesome EMR for automobiles.

As noted, people are more complicated than cars. So much so, that comparison is downright cartoonish. There are many ways a good EMR system would benefit us, but it is not a replacement for involved doctors and informed patients.

[thompson]

In fact one of her (many fine) doctors advised her to keep on doing this, just in case she finds something that everyone else has missed.

To sum it up: pretty much this. I have been told this by doctors, and I have heard doctors say this during rounds and at conferences.

[Countme-In]

"Wow!"

Indeed.

http://www.newyorker.com/magazine/2014/12/01/excrement-experiment

[CRHayne]

"In fact one of her (many fine) doctors advised her to keep on doing this, just in case she finds something that everyone else has missed."

That is so cool.

I once had doctors refuse to read the latest literature from top journals when I'd printed it out for them. (They were mis-treating chronic c. diff in my elderly father-in-law.) Saved his life by going with the literature instead of doctor's orders.

[thompson]

They were mis-treating chronic c. diff in my elderly father-in-law

If you haven't read Count's link yet, very relevant to C. diff.

[CRHayne]

"If you haven't read Count's link yet, very relevant to C. diff."

Fecal transplant was already in the literature, but no doctor within a thousand miles would do it. (This was a number of years ago.)

We might have tried it on our own next, but tapering off the vanco and adding a specific probiotic did the trick.

[Turbulence]

As noted, people are more complicated than cars. So much so, that comparison is downright cartoonish. There are many ways a good EMR system would benefit us, but it is not a replacement for involved doctors and informed patients.

Of course, no one suggested that an EMR replaces involved doctors or informed patients. Indeed, an EMR is useless without both. And of course, the mechanism by which EMRs help is by checking for potential interactions and escalating to doctors and pharmacists. What EMRs buy you is the ability to put all data in front of all clinicians.

People are in fact more complex than cars, but medical data is not actually substantially more complex than vehicular data. A listing of doctor visits with text notes plus lab reports and diagnostic imaging is really not rocket science.

Moreover, "informed patients" are a mixed bag. It is really great that dr ngo's wife has the social capital, as (I assume?) a scholar, with scholar family and friends to read the PDR or wherever she gets drug interaction data. But many people don't have the social capital or the cognitive resources to do that. So while I'm glad that dr ngo's wife can do things to help manage her care, this process can't really scale: we can't expect everyone, or even most people to do that. We need to design a medical treatment system that can work for everyone, not just really smart professionals from academic families.

[dr ngo]

I agree, Turbulence. All I was saying was (as other people have noted) simply having electronic medical records - even the best currently available - isn't enough. "A listing of doctor visits with text notes plus lab reports and diagnostic imaging is really not rocket science" - no it isn't, and that's already available. My son's company (among others) provides it, and I've seen it in action. Should be more widely available, and probably will be; at a time the national economy is generally stagnating or idling, the medical records industry is booming, and my son's company keeps taking on new hires so fast they have to throw up new buildings for them all on an almost annual basis. (Fortunately they're surrounded by farmland that they already own.)

But that's not the tough part. The tough part is having enough doctors with enough time and wide-ranging expertise to sort through all the variables. And that can come pretty close to rocket science. It may technically be true that "medical data is [are - but I'm a pedant] not actually more complex than vehicular data," but the underlying mechanism - that which the data purport to describe - is FAR more complex, and if you don't believe that, then you should definitely become a mechanic rather than a doctor. The medical data are fleeting glimpses into an organism we scarcely understand, and we don't have an "owner's manual."

Thus "A medical treatment system that can work for everyone" is a noble aspiration, but one that's unlikely to be realized in our lifetime. (Well, maybe yours, but not mine.) So we do what we can with what we got - which (as you correctly surmise) in our personal case, involves deploying our "social capital" in the interest of survival, or at least quality of life. I trust you do not begrudge us that.

[Brett Bellmore]

A while back, my mother was tracked down by a pharmacist, due to an automated drug interaction system they had at Riteaid, which noticed that she'd bought an (I think) arthritis medication at one of their stores up in Michigan, and a foot fungus medication at one in Florida. Seems the two of them were, when taken at the same time, prone to cause heart damage.

He put in considerable effort to track her down, and caught her before the effect was irreversible.

So, while waiting for the singularity, you might make a practice of always buying your meds at the same pharmacy chain, because they DO watch out for that sort of thing.

[wj]

I think you may be underestimating what an ERM system can do. Or, at least, what it can allow to be done.

With drug interactions, especially when several drugs are involved, you have a far better chance of writing a program, based on expert information, which can consider all of them and work out the interactions than you do of having any given physician be aware of them.

After all, the whole point of Expert Systems in general is that you are getting expertise which is very limited into the hands of lots more people. You do have to have a group of experts who can work out the details originally. But once you do, if the necessary input information is available, you can let anyone (including the patients!) with access to the system see what the results are. And without spending months or years gaining great expertise in all of the fields which are needed as a basis.

[Marty]

The problem with EMR systems is, generally, the House principle. Everybody lies. More important, people don't focus on updating it or making sure its accurate beyond today's symptom.

[thompson]

Turb:

Mostly what dr ngo said, especially:

but the underlying mechanism - that which the data purport to describe - is FAR more complex, and if you don't believe that, then you should definitely become a mechanic rather than a doctor.

This. An EMR is a great tool, but that's all it is. Managing complex chronic conditions, especially ones that interact with each other, requires far more than good record keeping.

I have no doubt we can (and have) built a system that can track blood tests and when your last vaccination was. Or, in Brett's example, can catch certain types of drug interactions.

But that is a far cry from being able to effectively coordinate treatment across multiple doctors. An important tool, certainly, but not a panacea.

A listing of doctor visits with text notes plus lab reports and diagnostic imaging is really not rocket science.

No, it certainly isn't. It also isn't medicine. It is a small aspect of an incredibly large and complex field. Which, I assure you, is *far* more complex then rocket science.

wj:

I think you may be underestimating what an ERM system can do. Or, at least, what it can allow to be done.

To be blunt, you are dramatically overestimating our handle on medical science. Your example of drug interactions is a good one, but managing multiple chronic conditions often go far beyond checking for known drug interactions. Its not simply the drugs that interact with each other. They interact with the patient and their microbiome, and all of these interactions are often poorly understood, as well as heavily varied between patients.

After all, the whole point of Expert Systems in general is that you are getting expertise which is very limited into the hands of lots more people.

Which would be viable if the experts were in agreement, and the parameter space was relatively confined. But the experts are often not in agreement, and the parameter space is gigantic. Especially with complex case management.

All of that being said, there is certainly value to an EMR, and more efficient dissemination of new best practices to frontline physicians. It's just not nearly as simple as people think to condense that expertise down into a digestible form.

[thompson]

Thus "A medical treatment system that can work for everyone" is a noble aspiration, but one that's unlikely to be realized in our lifetime.

This also needs to be emphasized. There are many barriers to a medical system that works for everybody. An EMR gets around some of them, but leaves a lot left.

[Marty]

"But the experts are often not in agreement, and the parameter space is gigantic"

However, the majority of medical treatments today are protocol driven. So the limitation is getting the expertise in the overall EHS to update the protocols in something that resembles quickly. The exponential growth in the inputs and the limitations of the process in gaining concurrence on their impact on protocols are the primary limiting factors in "keeping up".

[wj]

managing multiple chronic conditions often go far beyond checking for known drug interactions.

Certainly, it does. But then, if we don't know about an interaction, no doctor is going to be able to check for it either. The issue, as I see it, is to make doctors generally able to check for the interactions that are known . . . to anybody. That is, even if only the top expert in the field has found an interaction, then an expert system can make that expert knowledge widely available.

It obviously can't make available information that we don't yet have. But just making the information that we do have available more widely would be extremely beneficial.

[thompson]

wj:

That is, even if only the top expert in the field has found an interaction, then an expert system can make that expert knowledge widely available.

As I have maintained, EMR are and will continue to be useful. But even that small thing you describe, which sounds very reasonable, is a lot harder then you might think.

Experts often disagree, and the data are often muddy. There is rarely a singular 'top expert' who just happens to be better informed than all the other experts. A lot of good research is done, but its often hard to have fully controlled studies...leading to contradiction in the literature. And many interactions and side effects of drugs aren't uniform, they can happen in only a small subset of patients, making it hard to tease out those effects from random chance.

All of this is a round about way of saying there is still a great deal of 'art' to medical science. Doctors are going to make treatment decisions in conjunction with a patient based on no small amount of experience and intuition. Which means, multiple doctors that don't necessarily understand and/or agree with each others' decisions might have difficulties that aren't readily addressed by an EMR.

And all of this can be on the backdrop of a patient that will decline regardless of the treatment, and the doctors are just trying to make that decline as long and painless as possible...which is going to involve a lot of patient specific calibration and considerations about quality of life. Those patient specific aspects, both medical as well as value judgements, are going to be further colored by how they are perceived by the doctor. An EMR can help, but I return to my original statement:

Even with the best intentions, it can be difficult to coordinate between doctors for each patient.

So, especially when you are looking at chronic management, there is a lot that is impossible to automate at our current level of medical science.

None of this should be taken as a criticism of EMRs, just as a reminder that medical care can become almost arbitrarily complex, and is often conducted in a muddy gray zone of our knowledge.

[wj]

Certainly the decision will have to get made by a human being who can apply the "art." But even the best art depends in part of adequate knowledge. And it is the knowledge that the expert system can supply. (Including the knowledge that various experts disagree and how.)

Nobody, at least nobody that I know of who knows anything about computer systems, would advocate giving the system absolute authority. It is a tool for the doctor, not a substitute for him. But (done right) a very valuable tool nonetheless.

[thompson]

wj:

I think we are perhaps in more agreement then it seems. What I'm trying to illustrate (apparently poorly) is how complex even the small thing you are asking for is. So for example:

But even the best art depends in part of adequate knowledge.

Yes, and having ready access to an EMR aids that, certainly.

And it is the knowledge that the expert system can supply. (Including the knowledge that various experts disagree and how.)

What I'm trying to emphasize is outside of some well defined areas, this rapidly balloons into *textbooks* of knowledge that evolve quickly.

That 'expert system' exists: Pubmed/Medline/MeSH indexed peer-reviewed journals and medical trade publications. It's possible we could index or curate it better...but I'd really like to hear how you plan to do that.

And that's just what's published...I've worked closely with many clinicians that 'know' things that are most certainly not published. Just something that has spread around the community based on observation and experience.

Getting expert opinion and cutting edge information out to doctors is a nontrivial problem that many smart people are working on.

And finally, none of that considers how this becomes arbitrarily complex and unique in some patients, especially while managing multiple chronic conditions. The list of 'possible' interactions and side effects balloons, making it extraordinarily difficult to create a database that can return a useful list to the clinician.

Again, I'm not arguing the EMR are not useful. I'm simply pointing out that they aren't a solution to much of what makes medicine 'not work' for people today.

[Countme-In]

Negotiating one's path thru the complexity of modern technological medicine and its attendant and lucrative specialization, not to mention the Kafkaesque billing procedures one gets to unravel not too soon after coming to from the anesthetic is one thing, but have you had the chance lately to find your way to the circumlocution ward in a multi-towered, multi-winged metropolitan hospital?

http://dish.andrewsullivan.com/2014/12/02/the-costs-of-poor-hospital-design/

I have. In one I visited recently, once you find your way out of the parking garage into the hospital structure itself via a series of elevators and escalators, you are faced with a maze of endless hallways and banks of elevators, which upon exiting onto a mezzanine and walking the sky bridge to another tower, one is confronted with several more banks of elevators and on .... then you sit in an acoustically impossible waiting room way past appointment time and the only name the receptionist calls repeatedly is "Will Mr. Joseph K. please report to the admitting desk", and you are not sure you heard right, and not a soul moves because we no one knows which one of is Joseph K. until several people rise and ask the receptionist what name was called, cupping an ear, and she asks, "well, what is your name, dear?" and you tell her, and she shakes her head no, that's not it, and you return to your seat, and the receptionist calls out again "Mr. Joseph K,, your doctor will see you now", and you look down at the name tag they clipped to your collar near the entrance to the structure, a part of the building you will never see again because it's undergoing renovation just since you've arrived, and the name affixed there, indeed, is Joseph K., and you rise and follow a nurse down another interminable hallway, and she weighs you on a scale, and you ask should I remove my shoes, and she says No, and you think, but I removed my shoes three years ago the last time I was weighed by a nurse, and isn't the disparity between then and now going to queer the diagnosis somehow, but she says not to worry, having read your thoughts from the worry lines on your forehead, and then you enter a room and are asked to strip and sit on a cold metal and plastic table/bed and wait.

Once you find your car again, it's been ticketed by a traffic matron who looks exactly like the nurse who weighed you hours before, plus you are now wearing someone else's trousers.

Complexity is America's lifeblood. In twenty years, 50 million dementia-suffering baby boomers, wandering about endlessly in little eddies of the demented, in the purposely convoluted structures of hospitals, the dead end cul de sac aisles of major grocery stores, the maze of suburban street developments, our cell phones clapped to our good ear, negotiating the phone trees, as we squint at scraps of paper we carry with our PIN numbers and passwords, and the name of our first pet written on them, gibbering, and tinkling into our Depends because the bathrooms are about a mile and a half away.

You must take charge of your situation.

Now go down that hallway and take your next left, and you'll find the consultant's station. Pay the fee, and receive the next clue to where your destination is in this .... where am I again? Whatever you do, don't even glance at the medical specialists hurrying to their next lucrative assignation, who seem to know exactly where you are even if you don't, because they employ ruthless bill collectors.

Good luck.

[Countme-In]

You know those little signs inside the entrances to major structures that depict a stick figure of you with an arrow pointed at you connected to the statement "You are here"?

It's time to worry, if then later, on yet another elevator, there is an identical diagram on the elevator wall of you in the little square of an elevator which says "You are here," and you think, yes, but how come the other guy in the elevator with you doesn't have a sign that tells him where HE is, within the 8' by 8' confines of the elevator.

It's disconcerting too when you come upon a similar sign in a long hallway, one fluorescent light flickering maddeningly and apt to trigger your seizures because your medications are interacting in a way not predicted by the separate specialists who assigned them to you, that depicts a stick figure with an arrow connected to the words "You are NOT here, because that's not you. It's the guy who passed by here several minutes ago."

Then you reach the receptionist's desk and there is a sign with the stick figure, which is dressed exactly like you, striped trousers, plaid shirt, and each sock a different argyll, and the arrow points to the statement "We've no idea if you are here or not because the doctor is not here, and the receptionist is out to lunch, and we don't mean eating lunch. If YOU think you are HERE, take a seat."

Years ago, a dear friend of mine, who was much older than I was at the time, was placed in a nursing home by her grown children and during my visits, she was usually in various states of confusion and befuddlement, probably from some form of dementia combined with whatever drugs she was on for her physical maladies, if she wasn't altogether out like a light.

She hated being there.

The last time I visited her, the aide wheeled her out in a chair, just fresh from a bath, and my friend was slumped over, her chin on her chest, seemingly asleep, and I leaned over, touching her wrist lightly, and gently said, "Hi there, dear, do you remember who I am?" and she brought her head up, opened her eyes directly into mine, and said, with the tone of an exasperated teacher (which was her profession) speaking to an imbecile, "Yes, you're Countme-In. Who do YOU THINK you are?"

The aide standing by had to quickly pull a tissue from his pocket and catch the stuff rocketing out of his nose as he stifled a guffaw.

In swift recovery mode, I quickly straightened up (causing a shooting pain in my lower back), shot my cuffs, straightened my tie, and said, "Well (insert her name), I wasn't quite sure, but I knew you would be able to tell me."

Then she fell asleep immediately.

As I was leaving, I stopped and looked at the little "You are here" diagram inside the entrance, and it said "Your friend won't be here the next time you are here, whomever you are."

And she wasn't

[wj]

thompson, I think you are right. More in agreement than not.

What I am picturing is a medical record system which interfaces to an expert system on medications. The MRS tracks what medications the patient is currently taking. The expert system checks proposed new perscriptions for conflicts.

Are there lots of things that some doctors know that would not initially be in such a database? Absolutely. But something like this gets built incrementally. First, you put in all the conflicts that you know of. Then you ask everybody using the system to look for errors and omissions. And make updates frequently. (There are lots of processes and procedures needed to get all this to work properly. But that's the high level idea.)

It's a long journey, and this would be a single step along it. But it would, I think, be a step in the right direction. And one which, if we don't take it, we don't get where we want to go.

[Doctor Science]

Do those of you who use non-USan medical systems find that the problem of med-tech-wrangling is better where you are, or just as bad? I suppose it *could* be worse ... but we have The World's Worst Medical Payment System, so it would be weird if physician-coordinating and -wrangling was better here.

[thompson]

wj:

The MRS tracks what medications the patient is currently taking. The expert system checks proposed new perscriptions for conflicts.

Yeah, that is very doable, and is deployed in many ways (for example, the pharmacy chain that Brett's mom shops at). It's just not a unified system yet (which, I'd agree, would be a massive and achievable step forward for patient care).

But even that has its limits, because drug interactions can be dependent on genetics, age, condition, diet, etc etc. You could make an exhaustive list that the MRS flags every possible interaction...but you run the risk of having too much data that's poorly categorized. That's something to work towards and likely achievable, but its not a trivial problem.

But mostly, I'm just trying to emphasize (and perhaps this is perfectly understood, you just don't know WHY I keep bringing it up), that this isn't going to resolve the problem of coordination across multiple specialists. It would ease it, certainly, but as our population ages and more and more of our medical system targets managing multiple chronic diseases, we're going to need solutions beyond EMR.

[wj]

I get it. And I agree (even if my agreement has been buried in my other words).

But while we will certainly need solutions beyond EMR, we will need that as a base to build those solutions on.