Let's examine this myth.
First, there is an assumption that, prior to seeing you, a doctor would read the EHR to get the "big picture", making the visit more efficient. My experience, and that of everyone else I've talked to, differs. Another person, such as a nurse, often has a sort of mini-interview, capturing data to put it INTO the EHR -- presumably so that the very busy doctor can be spared the effort to ask those questions, as he simply needs to read the EHR to see the answers you just gave.
What actually happens is that the doctor starts with "So why are you here today?" or some such thing, indistinguishable from what they would ask if you hadn't just talked to the nurse and answered all those questions.
To put it very succinctly, the (EHR + doctor) hybrid unit fails the "OMG" or "Oh, my God!" test. In situations where any of your good friends who hadn't seen you for a while, meeting you, would go "Oh my God what's wrong?!!", the doctor asks "What brings you here today?"
Worse, this does not seem to change with time. Familiarity with your "normal" state is neither captured by the EHR, even over time, nor is it somehow communicated from the EHR to the physician at the point of care where you'd expect it to. Reliance on the EHR has replaced memory.
Hmm. Well, how about the value of all those prior visits and what the EHR has captured about those, now that it's all electronic and legible and stuff? (a) are those read? and (b) if read, what do they change?
In answer to (a), yes, probably, if you have one prior visit, the information from might be read. Far more likely, if available, a very short 1 page summary of it might be read, listing allergies and previous diagnoses and major events.
Let's suppose that on a prior visit to a prostate surgeon specialist the specialist wrote that your prostate was "precancerous" and "immediate surgery was recommended." Further, let's say you are familiar with this surgeon, who has a reputation that "He never saw a prostate that didn't need to be removed."
So, are you as a doctor going to follow that advice and get the patient admitted and off to surgery? More likely, you will raise an eyebrow and discretely suggest the patient "get a second opinion."
In fact, for pretty much anything that is asserted as a "fact" in the EHR, your opinion may be that the source of that information is biased, inexperienced, working off an entirely different model of health, out of date, or otherwise not to be believed and acted upon.
So, exactly why then was it worth $100,000 to get this information in front of you?
Or, let's take a case where an older patient, seen over 100 times by a health system, has multiple problems, sees multiple specialists, and has over 1000 medical documents in her file from these visits.
Again, of the 12 minutes the health system allows the doctor to deal with you, and the two minutes of that the doctor might choose to spend reading the prior EHR, what fraction of these 1000 documents do you suppose he'll read? The most likely answer is: zero. In fact, as in the rest of life, the MORE extensive the EHR is, in terms of number of documents and complexity of issues described per document, the LESS likely it is that your current physician at the current visit will elect to READ any of it.
A graph of blood pressure historical data might be glanced at. Details about blood type might be looked at, with a note that this should be redone before giving blood, "just in case the prior value is wrong."
So let's back up a step and think about what we think should be going on. There is a lot of information, encoded into text or structured text or forced-choice fields in the EHR. For the most part, this information is effectively divorced from meta-information, such as the name and qualifications of the source of that data, or any qualifications they might have put on it or caveats regarding it.
Maybe the supposition is that a doctor, in the two minutes allocated, is going to "process" all that information and produce a "big picture", a mental concept that includes all of the relevant parts of what has been done before, so that, today, he can look at you and advance "the big picture" even further in understanding what might be wrong with you and the plan of addressing it. I don't think so. No one can read 100 documents and process them well in 2 minutes.
So, let's say there is some process that summarizes the prior documents and sort of encapsulates a distilled "big picture" of the patient. By definition, unless this is a wiki or otherwise heavily linked electronic document that allows "drill-down" into what lies behind assertions, the summary is going to leave out most of the details.
However, it is precisely the small details that don't quite fit, the nuances that aren't quite right, that a doctor's thousand of hours of training can spot and realize that prior diagnoses are incorrect. These details are suppressed in the summary, because it is a summary.
So, whoever writes the summary and elects which details are "relevant" and which matter enough to be put into the summary actually, in effect, determines the outcome of anyone reading it. (The same problem is true of War Rooms in the Pentagon, by the way, and has been studied there.) By the time some low level person, who has the hours, has used THEIR judgment to filter out all the "irrelevant details" in this report to his superior officer, there isn't really any room left for the superior to question.
We have, in effect, by using the hierarchical summary feature of the electronic record, removed the doctor from the loop. An army of low-level staff members have, in summarizing, removed the need for a highly-trained doctor at the top to read the summary, because there's no details there left for the highly-trained person to respond to differently than a low-level person would.
So. let's summarize our own thinking so far. The EHR is alternately too thin to be of value, as it is missing too much, or too thick to be of value, because no one can possibly read it in the 2 minutes allowed. The solution to this problem with text-based concepts is to have low-level people (the only ones with time to do so) do a summary of the case ,which could now be read in the two minutes allowed. In almost no EHR system are the details of the summary cross-linked with hot-web-links back to the source of the data, in case the source is changed, deleted, addended, or questioned. Doctors may be given an opportunity to challenge the summary, but to do so would require going back and doing the summary themselves, which, by assumption, they don't have time to do.
So, the clinical picture that can EVER be embraced by an EHR is actually quite simplistic, and has to fit in a short series of listed bullet-point items. The nuanced, net, effective clinical picture of your complex medical condition is limited to what fits on a Poweroint slide, effectively a cartoon that discards all traces of uncertainty or ambiguity or conflicting readings that might open the door to a realization that your primary diagnosis is incorrect. Alternative framings of your condition, alternative diagnoses that might be relevant are forcibly discarded since there is ":no field for them on the form." Clinical impressions of "maybe" are forced into one of "Yes" or "No" to simplify billing or to satisfy the mental model of some low-level non-clinical programmer somewhere who was trying his best to "validate data."
Furthermore, the clinical picture stored in the EHR does NOT have the property that it improves with time, or with use. No facility is included to allow a doctor to highlight relevant sections of a document to save themselves time the next time they come back to this patient. No facility is provided to let them select a section of document and "drag and drop" it into a summary document, pulling along with it all the cross-references to the work cited. No facility is included for them to put on a yellow sticky with a note to self challenging some fact in the existing record.
In fact, there is nothing in EHR systems that would look at text descriptions and recognize and flag that completely inconsistent conclusions are drawn in different places in one document or across documents.
I would be most astounded if any hospital had a section of the summary which revealed, let alone highlighted for attention such conflicts. Picture reading "Well, Doctor Smith thinks X is true, but Doctor Y thinks Smith is an idiot and X is clearly false. Doctor ignores them both and assumes Z is true. " It may be a true summary of reality, but it is very very unlikely to ever be clearly articulated in an EHR for lawyers to find. So, instead, it will be covered up and buried. ALL such conflicts will be suppressed, and even their existence whited-out of the summary report, as if everyone happily and confidently agreed with each other. The EHR facilitates this, because it has no room for "conflicting opinions" in the structured field, which has to be either "YES" or "NO".
This is the conflict within a hospital. Imagine what will occur when different physicians at different practices or hospitals have to contemplate and respond to conflicting opinions from the competing practice or hospital, in order to come up with the "master, nation-wide health summary for this patient." Imagine the heydey attorneys will have if the differences and discrepancies are revealed and highlighted. Imagine the fraud and damage to clinical truth that will occur if the differences and discrepancies are shoved under the rug and made to "go away".
So, I challenge the designers of these regional EHR summary databases. What IS your plan when you run into conflicting and incompatible diagnoses by different doctors for the same patient? As you surely will, and very quickly indeed.
Are you going to highlight them, so it's clear that none of them can be considered definitive? Are you going to code them "under dispute"? Do you even have capacity to store such a code? Are you going to use your own judgment or your own people to override one, or the other, or both doctors? Are you going to refuse to show anything until the two doctors reach a consensus opinion? Who is going to pay for the costs of resolving such discrepancies in "the master patient chart?" Who is even capable of resolving such disputes?
Are you going to pretend that such situations don't exist, or only exist "very rarely" in the hope that funding will not be held up on such a little thing? Nail it down people. Do you admit that these problems will occur (and therefore open yourself up to questions about how you intend to deal with them?) Or do you deny that these problems will occur (and therefore open yourself up to a delay in funding until you say how you will deal with them?)
In point of fact, this "unclean data" problem will present not just a problem to regional health warehouses. It will document, clearly, for all to see, just how BAD clinical records actually are. It will document, for all attorneys to discover, just how much disagreement there is among professionals. It will document, for patients, that their unqualified trust in any given doctor should be tempered with the evidence.
And it will document for all that there has been, to this date, a conspiracy of silence about this problem. Did no one know about this? When exactly were you planning on mentioning it? Only AFTER we'd spent $100,000,000 getting to that point?
Each time a doctor opens up even his own records about his own patients, he is faced with documents he's not allowed to mark on, cross-link, color-code, put post-it notes on, etc. If he attempts to go into length about complex conditions, he is punished by failure to meet his scheduled case load as well as called by the transcription department about having documents that are "too long" and cost way more than other doctor's document to transcribe and summarize.
The text stream called an EHR, therefore, may have a good ability to persist pixels, or facts such as a blood pressure reading, but as the complexity of the concept or medical condition gets higher, the EHR is unable to follow along and store, in any kind of retrievable fashion given the 2-minute-rule, the "big picture" and all the nuances that picture should be resonating.
What will get passed on to the next shift, or the next doctor, or the next visit, is at best a cartoon summary of things to date, prepared by a non-physician with all trace of nuance and uncertainty removed. One hospital I visited told surgeons they couldn't store the normal pictures with circles and arrows they used to plan a surgery or summarize what happened, as the computer system wasn't sophisticated enough to do what the paper chart system did, ie, allow pictures to be attached to the patient chart. Again, what is stored is getting dumbed down and reduced to what is easy to fit in a computer form.
Which perhaps explains why the doctor doesn't bother to read it, or, often, even to read the notes his nurse made at the start of the visit. He may ask the same questions again, not because he is interested in the "answer" (as seen by the EHR) but because he is interested in the nuances, the body language, the uncertainty or certainty that surrounds those answers. He cares about the meta-data, because a large part of clinical judgment is based on intuition and reading the meta-data. Sadly, none of the EHR has room for such metadata. A transcribed document codes an emphatic "YES!!!" the same as a neutral "yes" the same as a hesitant "um... yes, I suppose, sort of..." To humans, these are very different answers. If my girl asks me if I love her and I say "YES!" versus (pause) (ponder)( delay)( fidget) "... yes?" I am conveying very different (and actionable) information.
The EHR throws out all this meta-data. If you're going to do that, you might as well just have clerks sitting and following a flowchart or the computer have a set of rules that guide the "next question" to each issue down some tree that comes up with "the diagnosis" or "the proper recommendation", at the expense of throwing out every point at which a trained doctor would say "wait, that's the wrong question. That's not quite right."
What it doesn't explain is why the country is so gung-ho on spending billions of dollars to install Electronic Health Record systems in every nook and cranny of the so called "health care system" especially for Medicare. Children may have relatively simple things wrong with them that "fit" in the EHR. A broken arm. 65-year olds probably have at least 3 chronic conditions and are taking over ten different prescription medicines for a variety of interlocked an inter-related system problems.
Apparently some programmers, managers, and insurance companies think that can all be neatly and correctly summarized in a few Diagnostic Related Group codes (DRG's) and everything is fine.
In the real world, it's hard to even imagine how such a system could possibly deal with the complexity of even one older patient over multiple visits.
In reality, of course, every actor in the health-care is multiplexing and distracted. Doctors, nurses, labs are acting like short-order cooks, starting on one patient, taking one step, leaving them to go deal with some other patient or crisis, trying to remember where they were, reprioritizing, re-triaging, going back to the first patient for a minute, etc. None of that interrupted action-coordination is contemplated by the programmers who designed systems as if the doctor or nurse, with all the time in the world, sat down and did everything for one patient before even beginning to think about the next one.
A short order cook who did orders one at a time in serial order would be fired by the end of the first day. You just can't operate that way, you have to overlap, predict what's coming, allow for lag times, etc.
On this account EHR's are equally out of touch with reality. The EHR expects you to sit and do everything for one patient at one time, so it can do "validation" and help "support your decision."
There is no way that serial text capture and summarization can possibly do that job, in a real environment, with real medical conditions.
The IT people don't need to force clinicians to "get with the program" and "stop resisting computerization." They need to go back to the drawing board with a better sense of how badly they have conceptualized and modeled what goes on in a hospital, and design a system that supports real people doing real work with patients with truly complex clinical conditions, in the fragmented, interrupted, and multiplexing distracted mode that clinicians are forced to accept as terms of employment.
There are other issues as well, that I won't go into in this post. One of the biggest ones is inappropriate persistence or stickiness of a diagnosis. Once one doctor, whoever goes first, states an opinion and a diagnosis, regardless how tentative, there is some legal and professional courtesy and psychological pressure on the next doctor to agree with, or by silence not challenge it, even if they believe in their hearts that the diagnosis is pretty suspect. The third doctor to see the record will have even a harder time going against the flow and disagreeing with the first two doctors. From then on, very few doctors would challenge the "consensus opinion" about the first diagnosis. The diagnosis has been electronically locked-in-stone by the EHR process. If the second doctor had gone first, a different diagnosis would have been locked-in-stone.
You have to worry about any process where the order in which people see data changes the outcome. The order drugs are listed in a pull-down menu, for example, has a strong impact on which drug a doctor using an EHR will select. By itself, that should be a big "WHOA." until THAT gets sorted out.
No comments:
Post a Comment