It was a dark and stormy night. My computer didn’t catch fire while typing the previous sentence. No alarms were triggered warning me about the quality of such opening. I wasn’t prompted to select subjects and predicates from dropdown lists. I typed the entire sentence, letter by letter, not at all dissimilar to its first rendering back in 1830.
Computer software in general, and Microsoft Word in particular, magically removed the hassles of quills, ink, paper, blotters, sharpeners, ribbons, whiteout, carbon paper, dictionaries, and all the cumbersome ancillary paraphernalia needed to support authoring, but made no attempt to minimize the cognitive effort associated with writing well.Authoring great literature today requires as much talent and mastery as it did in the days of Edward Bulwer-Lytton.
___________________
For several decades, software builders have tried to help doctors practice medicine more efficiently and more effectively. As is often the case with good intentions, the results turned out to be a mixed bag of goods, with paternalistic overtones from the helpers and mostly resentment and frustration from those supposedly being helped.
Whether we want to admit it or not, the facts of the matter are that health IT and EHRs in particular have turned from humble tools of the trade to oppressive straightjackets for the practice of medicine.Somewhere along the way, the roles were reversed, and clinicians of all stripes are increasingly becoming the tools used by technology to practice medicine.
A common misconception is that EHR designers produce lousy software because they don’t understand how medicine is practiced. The real problem is that many actually do, and the practice of medicine is precisely what they aim to change. These high clerics of disruptive innovation would have you believe that “resistance to change” is equivalent to the resurrection of paper charts, thick ledgers, and medical information coded in secretive hieroglyphs.
The truth is that physicians want to use modern computers, but they resent being used by computers.And the truth is that if we shed the orthodoxy imposed on us by self-serving “stakeholders”, computer software can indeed help address various problems in health care, some in the here and now, most in a distant future.
One thousand and one elements
This may sound strange to some, but the first step towards putting EHRs back on the right track should be to stop trying to help physicians practice medicine. Clinical decision “support” in the form of alerts, disease specific templates, mandatory checklists, required fields and rigid workflows are some of the things that must be removed from EHRs for two reasons. First, most of these “features” don’t work very well anyway. Second, more often than not, the real purpose of said support is not clinical in nature. For example, alerts about generic substitutes for brand name medications, data fields that must be filled and checkboxes that must be clicked to satisfy billing codes, PQRS or Meaningful Use, and the wealth of screens to be traversed before an order can be placed, have no clinical value. And in most cases the opposite is true.
Some experts argue that EHRs are failing because they are nothing more than an old paper chart rendered on a computer screen.Many others are outraged by the fabled lack of interoperability (dissemination of information) or the lack of EHR usability, i.e. number of clicks, visual appeal, color schemes and ease of information retrieval. I would suggest that these dilemmas are peripheral to
the one foundational problem plaguing current EHR designs – the draconian enforcement of structured data elements as means of human endeavor.When Google mapped the Earth, it did not begin by mandating how to build and name roads and buildings. When we indexed and digitized books and articles, we did not require that authors change the way they write prose or poetry. When we digitized music, we did not require composers and performers to produce binary numbers at equidistant time intervals, and we did not make changes to musical instruments to allow for better sampling. We built our computerized tools to ingest, digest, slice, dice and regurgitate whatever humanity threw at us, without inconveniencing anybody. This is why good technology seems magical.
EHRs on the other hand, are obnoxiously demanding that people change how they think, how they work, and how they document their thoughts and actions, just so that the rudimentary software prematurely thrust upon them can function at some minimal level of proficiency.
People don’t think in codified vocabularies. We don’t express ourselves in structured data fields.Instead of building computers that elegantly adapt to the human modus operandi, EHRs, unlike all other software tools before them, demand that humanity adjust itself to the way primitive computers work. The self-appointed thought leaders, who are taking turns at regulating the meaningful clicks of EHRs, are basically demanding that we discard the full spectrum of human communications, in favor of gibberish that supposedly serves a higher purpose.
All the pretty horses
What is the purpose of EHR documentation templates? There is practically no EHR in use today that does not include visit templates. Visit templates are a list of checkboxes, some with multiple nested levels, which allow documentation by clicks instead of by typing, writing, drawing or dictation. Visit templates are created for each disease and contain canned text for findings judged pertinent to that condition by template creators.
In all fairness, many physicians like documentation templates because with just a few clicks you are able to generate all the documentation required nowadays to get paid for your work, pages and pages of histories, review of systems, physical examination, assessments and plans of care. Do doctors like templates because they believe this extensive documentation is necessary, or do they like templates because the checkboxes alleviate the pain of typing thousands of meaningless regulatory words?I suspect the latter.
Clinical templates, along with the automated clinical decision support they enable, are advertised as time savers for physicians. The time saved is the time previously spent with patients, and most importantly the time spent thinking, analyzing, and formulating solutions. For most, it’s also the time spent rendering thoughts in a manner that can be understood by another person.
Furthermore, when your note taking is template driven, most of your cognitive effort goes towards fishing for content that fits the template (like playing Bingo), instead of just listening to whatever the patient has to say.Even in “efficient” practices where staff does the clicking and physicians have the luxury of asking “open ended” questions, the patient story, the quirky details that are irrelevant to the template, are not documented (highlighted, circled, noted on the margins, etc.) anymore. Is this a good thing?
If we proceed on the assumption that IBM Watson and the likes are eventually going to be artificially intelligent enough, and big data are eventually going to be big enough, to respectively analyze and represent a complete human being, then yes, we can safely dispense with old fashioned human expertise. However, we are most certainly not there yet, and regardless of industry rhetoric, we are not certain that we will ever be there, and we are not even sure that we want to ever be there. While this utopia (or dystopia) is portrayed by interested parties as “inevitable”, chances are that for at least several generations we will be forced to contend with imperfect digital renditions of medicine, instead of allowing EHRs to follow the growth of underlying technologies. This is akin to summarily confiscating and shooting all the horses, on the day Henry Ford rolled the first Model T off his assembly line. Where would America be today, if we did that on October 1, 1908?
Furthermore, what type of doctors are we producing when we teach medicine by template, supported by clinical decision aids based on the same template, and assessed by quality measures calculated from template data?
Medicine does not become precise just because we choose to discard all imprecise factors that we are not capable of fitting into a template. Standardization of processes and quality does not occur just because we choose to avert our eyes from the thick edges were mayhem is the norm.Dumbing physicians down is not the optimal strategy for bringing computer intelligence closer to human capabilities. EHRs should not be allowed to become the means to stifling growth of human expertise, the barriers to natural interactions between people, or the levers pushed and pulled at will by greed and corruption.
Bildungsroman style
Instead, EHRs could be the scaffolding for IMB Watson and other emerging contraptions to grow and become truly useful tools for both doctors and patients, and yes, also for legitimate and beneficiary secondary uses of clinical information.
Instead of mandating that doctors think and work in ways that serve Watson’s budding abilities, we should require that Watson learns how to use the normal work products of humans. Instead of enforcing templated thought and workflows, whether through direct penalties for doctors or indirect certification requirements for software, we should work on teaching Watson how to parse and use human languages in all their complexity.Watson should grow up to be the multi-media scribe behind the computer screen, the means by which the analog music composed by physician-patient interactions is digitized into zeros and ones without loss of fidelity and without interference with actual performance.
Billions of years of evolution endowed the lowliest human specimen with cognitive abilities that machines will most likely never attain. The glory is in the journey though. We need to accept delayed gratification, and we need to accept that the challenge will span centuries, not just one boom-bust cycle of a fleeting global economy. We need to accept the fact that we will all die long before the ultimate goals are achieved, instead of declaring victory whenever each negligible incremental step is taken. If we are going to create a new form of intelligent life on earth, we need to assume the same humility Nature, or God, has been exercising since the dawn of time and counting.
Otherwise, we are all just a bunch of hacks looking to make a quick buck on the backs of our fellow men and women.__
'eh?
More reflections on this in a bit. For starters, good stuff in the comments over on THCB, where Margalit's post was also put up. From the ever-dour iconoclast Nortin Hadler, MD:
You say, “The self-appointed thought leaders, who are taking turns at regulating the meaningful clicks of EHRs, are basically demanding that we discard the full spectrum of human communications, in favor of gibberish that supposedly serves a higher purpose.” But you imply that there will be clinically meaningful clicks in our future when self-appointed thought leaders learn to digitize the multivariate, interactive, inexact, unreliable and temporally varying factors that comprise the idiosyncratic predicaments of life. Really? Ever?Yeah, OK, but this is not exactly news, and is really not about charting per se. As I wrote 17 years ago in my essay about my late daughter:
First, the many physicians I have come to know in the past few years are in the main acutely sensitive to the problems of clinical conceit and "paradigm blinders." Indeed, the Utah pediatrician's"$100 bill" wisecrack was offered to an audience of doctors and their allied health personnel during quality improvement training. Second, the body of peer-reviewed medical literature does not constitute a clinical cookbook; even "proven" therapies-- particularly those employed against cancers-- are generally incremental in effect and sometimes maddeningly transitory in nature. The sheer numbers of often fleeting causal variables to be accounted for in bioscience make the applied Newtonian physics that safely lifts and lands the 747 and the space shuttle seem child's play by comparison. Astute clinical intuition is a necessary component of a medical art that must, after all, act and act quickly-- so often in the face of indeterminate, inapplicable, or contradictory research findings.__
UPDATE: THE DEBATE GOES ON AND ON...
From The NY Times,
Your New Medical Team: Algorithms and PhysiciansI'll cite just 2 comments:
Austin Frakt
Can machines outperform doctors? Not yet. But in some areas of medicine, they can make the care doctors deliver better. Humans repeatedly fail where computers — or humans behaving a little bit more like computers — can help. Even doctors, some of the smartest and best-trained professionals, can be forgetful, fallible and prone to distraction. These statistics might be disquieting for anyone scheduled for surgery: One in about 100,000 operations is on the wrong body part. In one in 10,000, a foreign object — like a surgical tool — is accidentally left inside the body.
Something as simple as a checklist — a very low tech-type of automation — can reduce such errors. For example, in a wide range of settings, surgical complications and mortality fell after implementation of a basic checklist including verification of patient identity and body part for surgery, confirmation of sterility of the surgical environment and equipment, and post-surgical accounting for all medical tools. Though simple procedures would all but eliminate certain sources of infections in hospitals, thousands of patients suffer from them in American hospitals every year.
Limits on how much information we can process and manipulate make it hard or impossible for even the smartest and most adept doctors to keep up with new evidence. In 2014 alone, more than 750,000 additional medical studies were published. Granted, a physician might need to keep up only with the evidence in her specialty, but even at a fraction of this rate, it is unrealistic to expect even the best physicians to assimilate every new development in their fields. In cancer alone, 150,000 studies are published annually.
Computers, on the other hand, excel at searching and combining vastly more data than a human. I.B.M.’s Watson — the computer that won Jeopardy! — is among the best at doing so. Teams of physicians at Memorial Sloan Kettering Cancer Center in New York, the University of Texas MD Anderson Cancer Center in Houston, and the Cleveland Clinic are helping to train Watson to apply humanity’s huge store of cancer knowledge to the delivery of more personalized treatment...
eblair rochester nyRecall from Margalit's thoughts above?
Charles Friedman PhD has framed that the true advances in physician cognitive support is represented by the equation: computer + physician brain > physician brain alone. Larry Weed MD the father the modern Problem-Oriented Record and the SOAP note method also made the point over 40 years ago that physicians cannot only remember all the diagnoses, they cannot remember all the questions to ask each unique patient. The idea of computer assisted decision support is to off load the wrote memory tasks to the computer, which will allow the physician to have more time, not less with the patient. Watson has made progress with looking for "best evidence" or best therapies for unique cancers but the real strides within the area of diagnostic support are happening outside of the media spotlight. The story of doctor being replaced by computer is obviously more attention getting, however computer as cockpit instrument is more realistic and is what is really exciting in healthcare information technology today.
Art Papier MD
Associate Professor of Medical Informatics
University of Rochester
CEO VisualDx
__
J.M.O'Belly KS
I'm glad I have a doctor who listens to me and then asks questions (and follow-up questions) that actually pertain to me and my health, rather than staring at a screen, reading questions from a script, and typing away in order to qualify my visit for a higher reimbursement level.
In all fairness, many physicians like documentation templates because with just a few clicks you are able to generate all the documentation required nowadays to get paid for your work, pages and pages of histories, review of systems, physical examination, assessments and plans of care. Do doctors like templates because they believe this extensive documentation is necessary, or do they like templates because the checkboxes alleviate the pain of typing thousands of meaningless regulatory words?And so it goes. As long-time readers know, I've cited the work of Dr. Larry Weed multiple times.
apropos, coming up soon, more thoughts on "AI/IA." in particular as they pertain to medical science and health care delivery.
From supermarket supply chains to consumer goods to construction to exploring for minerals and oil, the ability to crunch bigger and bigger data sets and make sense of them is improving pretty much every type of human endeavour. Kevin Kelly, the founder of Wired magazine, said the business plans of the next 10,000 startups are easy to predict: “Take X and add AI.” To coin a phrase, blessed are the geeks, for they shall inherit the Earth.I bought this book in lieu of Nick Bostrom's book “Superintelligence: Paths, Dangers, Strategies” (for now). I'd first got onto Bostrom via a New Yorker article "The Doomsday Invention: Will artificial intelligence bring us utopia or destruction?"
Healthcare is an interesting industry in this respect, because it has so far appeared to lag behind the general trend to improved performance from better information. It has been observed that our healthcare systems are really sick-care systems, often spending 90% of the amount they ever spend on an individual during the final year of their lives. We all know that prevention is better than cure, and that problems are most easily solved when identified early on, but we don’t run our healthcare systems that way.
Two major revolutions are about to sweep across the healthcare horizon, and we will all benefit. One is the availability of small instruments which attach to our smartphones, enabling each of us to diagnose early symptoms of disease, and transmit relevant data to remote clinicians. These instruments are the result of cheaper and better sensors, and the application of AI algorithms and human ingenuity to huge data sets. They will cut out millions of time-consuming and expensive visits to doctors, and enable us to tilt sick-care towards healthcare.
The other revolution is the ability to anticipate and forestall medical problems by analysing our genomes. The Human Genome Project was completed back in 2003, but it soon turned out that although sequencing our DNA was an essential first step to enabling the practical improvements to healthcare we hoped for, it was not enough. We needed to understand epigenetics too: the changes in our cells that are caused by factors above and beyond our DNA sequence. The application of AI algorithms to the data which scientists are generating about gene expression are now bringing those improvements within reach.
There is almost no aspect of life today which is not being improved by artificial intelligence. It is important to bear that in mind as we look at the potential downsides of this enormously powerful technology, and avoid a backlash which could prevent us benefiting from those improvements...
Chace, Calum (2015-08-31). Surviving AI: The promise and peril of artificial intelligence (pp. 38-40). Three Cs. Kindle Edition.
Relevant to the topic, from THCB, "What Would Issac Asimov Do?"
...As we’ve seen in the medical world–e.g., with regard to robotic surgery, femtosecond lasers, and proton beam therapy–there is an inexorable push to adopt new technologies before we determine that they are safer and more efficacious than the incumbent modes of treatment. Corporations have a financial imperative to push technology into the marketplace, employing the “gee whiz, this is neat” segment of early adopters to carry out their marketing, leading to broader adoption. All this happens well before society engages in the kind of thoughtful deliberation suggested by Eric. Meanwhile those same corporations take advantage of the policy lacunae that emerge to argue for less government interference. Unnecessary harm is done, and then we say, “These things happen.”...MARGALIT UPS THE ANTE
From one of Margalit's comments in a subequent THCB post:
I think Watson type of software should perhaps go to coding school instead of medical school and take care of the drudgery of billing, authorizations, and all the red tape, which can and should be automated. That’s what computers are for.
As to the notion that doctors are in dire need of help with thinking things through and diagnosing, why don’t we do the most basic thing that any developer should do, and ask doctors if they feel incapable of diagnosing disease? Why don’t we ask them if they need help with figuring out what to do after they diagnose something?Wow. Well, for one thing, surveying docs as to their opinions on documentation and digital decision support can be done, and has been done, repeatedly. As with nearly all opinion polling, you get wide distributions of sentiment pro and con. Moreover, all of this latest naysaying misses one fundamental point. It's not about information technology per se, it's the "productivity treadmill" imperative. If the typical physician only had to see an average of one patient per hour (8-10 pts per day) rather than 25-30, adequate documentation would be way less onerous. Let the mids handle the banal cases.
We have never posed these questions to physicians (and I will go out on a limb here and say that I think I know what the answer will be). We did however make a unilateral decision to intervene based on opinions from very interested quarters stating that medicine has become too complex for the human mind (or some other cliche along these lines). To me, the reasons behind this decision are suspect, to say the least. But even if I assume good intentions, the reality remains unchanged – we are in the way...
Unless you're really arguing that physicians should be absolved from any documentation duties whatsoever.What proportion of patient encounter bookings requiring MD attention are scheduled mainly to keep the doors open? By some credible estimates it's more than half.
I cannot but help be reminded of the Weeds' book "Medicine in Denial," which I've cited multiple times:
A culture of denial subverts the health care system from its foundation. The foundation—the basis for deciding what care each patient individually needs—is connecting patient data to medical knowledge. That foundation, and the processes of care resting upon it, are built by the fallible minds of physicians. A new, secure foundation requires two elements external to the mind: electronic information tools and standards of care for managing clinical information.MORE PUSHBACK AT THCB
Electronic information tools are now widely discussed, but the tools depend on standards of care that are still widely ignored. The necessary standards for managing clinical information are analogous to accounting standards for managing financial information. If businesses were permitted to operate without accounting standards, the entire economy would be crippled. That is the condition in which the $2 1⁄2 trillion U.S. health care system finds itself—crippled by lack of standards of care for managing clinical information. The system persists in a state of denial about the disorder that our own minds create, and that the missing standards of care would expose.
This pervasive disorder begins at the system’s foundation. Contrary to what the public is asked to believe, physicians are not educated to connect patient data with medical knowledge safely and effectively. Rather than building that secure foundation for decisions, physicians are educated to do the opposite—to rely on personal knowledge and judgment—in denial of the need for external standards and tools. Medical decision making thus lacks the order, transparency and power that enforcing external standards and tools would bring about. [Medicine in Denial, pp. 1-2]
By a commenter known as "lawyerdoctor."
With all due respect, Mr. Oram, doctors do NOT “need a lot of help making the right diagnosis.”
Doctors will occasionally benefit from having access to a vast amount of easily retrievable data in order to help confirm a difficult diagnosis. But the overwhelming number of physician decisions made thousands of times every day all over the world are accurate without the assistance of Watson, or Siri, or Medscape, or Oprah. They are made with the assistance of 4 years of medical school, 3-7 years of residency, and many untold years of clinical experience seeing tens of thousands of patients.
It does not help doctors to turn them into data-entry clerks, creating a 17-page electronic documentation stream that serves only to support billing a suture removal as a level 5 ED visit. I went to doctor school, not typing school. I prefer to see patients, talk to them, examine them, and hopefully make them better.
I can assure you that no one will be happier than I to see the day when the “robot doctor” spits out prescriptions for z-paks to runny nosed kids at a kiosk in the mall. Why? Because it means that I don’t have to do it. Will I get paid a little bit less? sure, but I’m fine with that. I can then treat real injuries, serious illnesses, and difficult cases, which is what I was trained in medical school and residency to do. When Watson learns how to put in a chest tube during a trauma code, let me know.
UPDATE
From The Atlantic:
The Unregulated Rise of the Medical ScribeOK, beyond the narrow training/competency concern, my comment beneath the article:
Demand for note-keepers in doctors’ offices is booming, but standards and training haven’t caught up.
A national campaign for electronic health records is driving business for at least 20 companies with thousands of workers ready to help stressed doctors log the details of their patients’ care—for a price. Nearly one in five physicians now employ medical scribes, many provided by a vendor, who join doctors and patients in examination rooms. They enter relevant information about patients’ ailments and doctors’ advice into a computer, the preferred successor to jotting notes on a clipboard as doctors universally once did.
The U.S. has 15,000 scribes today and their numbers will reach 100,000 by 2020, estimates ScribeAmerica, the largest competitor in the business. After buying three rivals this year, it employs 10,000 scribes working in 1,200 locations...
I call chart coding "lossy compression." And, in a way, any type of charting documentation is that, whether it's digital "structured data" or handwritten physician narrative impressions in a paper chart. The only way to get at the full contextual clinical encounter would be to videotape it (and then transcribe it -- which would essentially be equivalent to a "deposition" in the legal field). Given that the cost of digital A/V storage capacity is now effectively nil (how many banal YouTube cat vids are out there?), it's technologically possible.__
Yes, there would be privacy obstacles to that -- going beyond mere HIPAA.
Maybe we could try it via explicit patient "informed consent opt-in" permission, so we could assess the reliability of scribe charting.
"This encounter may be recorded for quality assurance purposes..."
OFF-TOPIC ERRATUM
My friend Dr. Mike Painter posted this on Facebook.
OFF-TOPIC EDITORIAL
Apparent campaign strategy? Alienate everyone except Ann Coulter, Cletus, and Billy Bob, Unreal, this guy.
________
More to come...
No comments:
Post a Comment