A colleague recently sent me a remarkable video – of Professor Lawrence Weed giving Medical Grand Rounds at Emory University in 1971. It’s fun to watch for many reasons: the packed audience composed mostly of white men in white jackets and narrow ties, the grainy black and white images a nostalgic reminder of Life Before High Def.
But the real treat is seeing Weed, then 47 years old, angular and frenetic, a man on a mission. He begins his talk by rifling through a typical medical chart, thick as a phone book. It is filled with garbage, he says disdainfully; “source oriented” rather than “problem oriented.” Weed was promoting his new vision for the medical record – one organized around patients’ clinical problems.
In 1964, in an article in the Irish Journal of Medical Sciences (reprised, rather more famously, in the New England Journal in 1968), Weed described his new model for patient care records, known as the SOAP note (“Subjective, Objective, Assessment, and Plan”). The idea was to begin with the patient’s history, then to present the objective data (physical examination, and results of labs, radiographs, and other studies), and finally to describe an assessment and plan for each of the patient’s problems. SOAP notes were designed to populate what Weed called the POMR: problem-oriented medical record.
This was revolutionary stuff at the time, and Weed was ready for pushback from doctors who argued that their random jottings were sacred totems of the “art of medicine.” At 51:30 in the video, Weed addresses these objections:
Art… is not a scribble in the middle of the night…. We debase the word art itself when we call what we’ve been doing art… As Stravinsky says, ‘art is nothing more than placing limits and working against them rigorously’ …and if we refuse to place them… you do not have art, you have chaos, and, to a large extent that’s what we’ve had.
I like Weed’s problem-oriented format – so much so that one of the reasons I’m pleased when my patients leave the ICU (other than the fact that this usually means that they’re getting better) is that my trainees’ oral presentations morph from being organ-system based (“Neuro: sedated, moving all fours, head CT negative for bleed; Cardiac: MAP 75 on 2 mics of Levophed, heart rate 85, lungs clear, 2 over 6 systolic murmur at apex, good systolic function on echo….”) to problem-based (“Problem 1: dyspnea. Patient remains short of breath, O2 sat 92% on 5 liters, lungs clear on exam and chest x-ray negative. Plan is for chest CT to rule out PE…”). When I hear an organ-based presentation, I find myself struggling to translate it into a problem framework, like someone who isn’t quite fluent in a foreign language trying to make sense of a song in that language.
Whatever the method used to divide patients up into manageable chunks, there is always a tension between a reductionist view of a patient’s problems (or organs) and a big-picture view. Just as we are, biochemically, simply the sum of our cells, even atheists know that humans are far more than that. So too are patients more than the sum of their problems.
Note that I’m not being touchy-feely and holistic here, decrying the dehumanizing aspects of modern healthcare. No, I’m saying that even if you are a coot who doesn’t give a damn about what the patient is feeling, even if you gloss over the social history in a mad dash to the liver function tests, even if you think that “patient-centered care” is mostly an empty slogan, even if you’re the kind of doctor who simply wants to figure out your patient’s problems and deal with them effectively, you must balance the simplicity and practicality of a systematic approach with the need to see patients as more than the sum of their problems.
With paper notes, this tension usually managed to work itself out. Even as we embraced Weed’s problem-oriented approach, there was something about the act of writing things down that made you realize that there was a person attached to the problems, and that each patient needed an über-assessment – a paragraph or two summing up his or her issues. The reason for this was not so much to honor the patient’s humanity (although that’s nice too) as it was to offer a crucial synthesis of what was otherwise a jumble of facts and impressions.
At UCSF Medical Center, we went live with our version of the Epic electronic medical record three months ago. It beats pen and paper, and it beats the EMR system that we traded out (at a cost of a hundred million dollars or so) by a long shot. The implementation went well overall, notwithstanding a few snafus (several thousand missing billing charges, a few patients temporarily unaccounted for, that kind of thing). I’m certain that these glitches can and will be ironed out.
But I’m less confident that we can fix what Epic is doing to our notes, and our brains.
The system, you see, places the problem list at the core of the patient’s clinical world – in a way that goes well beyond what Larry Weed imagined. One really doesn’t “write a note” anymore; rather one charts on each of the patient’s problems, one by one. At the end of a session, the computer magically weaves these fragments into what outwardly appears to be the patient’s progress note. But it’s not really a note, it’s a series of problems (each accompanied by a brief assessment and plan) held together with electronic Steri-Strips. In other words, it takes Weed’s vision of the POMR and hypertrophies it. As with muscle, while some hypertrophy can improve function and be attractive, there comes a point when more hypertrophy becomes constrictive, dysfunctional, even grotesque.
Why did Epic and our UCSF IT gurus structure things this way? The primary virtue is that this charting-by-problem approach allows the patient to be followed longitudinally, since one can track problems such as “hypertension” or “ovarian cancer” over years, seeing how they have been managed and observing the response to therapy. It isn’t a bad conceit, and it probably makes tons of sense when described in a fishbone diagram on an informatics seminar whiteboard.
But the effect I witnessed on patient care and education was less positive. When I was on clinical service in July and read the notes written by our interns and residents, I often had no idea whether the patient was getting better or worse, whether our plan was or was not working, whether we need to rethink our whole approach or stay the course.
In other words, I couldn’t figure out what was going on with the patient.
If Epic was the only thing promoting this kind of reductionist approach, it might be survivable. But it’s not. In the face of duty-hours limits, our trainees are increasingly programmed to operate in a “just the facts, ma’am” mode, to approach patients as a series of problems to be addressed expeditiously and algorithmically. This “if X, then Y” mode of thinking isn’t wrong, per se, but – particularly in the hospital – when unaccompanied by an effort to paint a coherent overall picture, the notes (and accompanying presentations) can become data without information, empty e-calories.
(Note that this problem comes on top of the copy-and-paste phenomenon so cleverly skewered by Hirschtick a few years back in JAMA. While copy-and-paste must be addressed, I’m less worried about it than I am about the impact of the EMR on clinical synthesis and reasoning.)
Larry Weed was acutely aware of another objection to his problem-oriented approach: the concern that each problem would be viewed in a vacuum. In his 1968 article, he wrote:
Fragmentation of single diagnostic entities resulting from listing separately single related findings is not a legitimate complaint against a complete list of problems. If a complete analysis is done on each finding, integration of related ones is an automatic byproduct. Failure to integrate findings into a valid single entity can almost always be traced to incomplete understanding of all the implications of one or all of them.
In the old days, failure to connect the dots between problems 1, 3, and 6 may well have been due to cognitive gaps. But the modern IT system can prevent even smart physicians from performing this essential act of synthesis. The patient with cough, sinus problems, and kidney failure cannot be thought of as the sum of the differential diagnosis of each of these problems. Instead, as Occam insisted, these problems must be placed in a Venn diagram, accompanied by strenuous attempts to figure out what lives at the intersection. This is damn hard to do when one is electronically charting each problem independently. Monkeys and typewriters come to mind.
Over the past few years, Epic has “won the game” in the competition among IT vendors trying to sell to large teaching hospitals. This is fine – it is a robust system and an impressive company. But something needs to be done to preserve the essential act of clinical synthesis, and soon.
What would I do? I’d build into each Epic note a mandatory field, and call it “Über Assessment” or “The Big Picture.” Mousing over a little i icon would reveal the field’s intended purpose:
In this field, please tell the many people who are coming to see your patient – nurses, nutritionists, social workers, consultants, your attending – what the hell is going on. What are the major issues you’re trying to address and the questions you’re struggling to answer? Describe the patient’s trajectory – is he or she getting better or worse? If worse (or not better), what are you doing to figure things out, and when might you rethink the diagnosis or your therapeutic approach and try something new? Please do not use this space to restate the narrow, one-problem-at-a-time-oriented approach you have so competently articulated in other parts of this record. We know that the patient has hypokalemia and that your plan is to replace the potassium. Use this section to be more synthetic, more novelistic, more imaginative, more expansive. Tell a story.
All in all, I am pleased that UCSF went with the Epic system and I remain a fan of electronic health records. And Larry Weed was right: we must have a structure to record what is happening to our patients, and his problem-oriented approach remains the most appealing one. (Ultimately, one wonders whether natural language processing will make such a structure less important, in the same way that I no longer pay much attention to filing documents on my Mac now that its search function is so powerful.)
But the time is now – before our trainees build habits that will be awfully hard to break – to recognize that electronic medical records do more than chronicle our patients’ histories, exams, and labs. They are also cognitive forcing functions, ever-so-subtly modifying our approach and language into something that can either improve our clinical care and teaching, or not. Let’s show these computers who’s boss, and put the “A” back in SOAP.