So far the argument has been about information: where it lives, how it decays, what the note does to it. This chapter is about what the arrangement does to you.
9.1 Two kinds of work
Consider a patient with heart failure, chronic kidney disease, and diabetes that’s slipping. Deciding how hard to push the diuretic when the creatinine is drifting up, and what that means for the SGLT2 inhibitor, and how to explain the tradeoff to a frightened seventy-eight-year-old — that’s genuinely hard. It requires everything you have. It is irreducibly hard; no technology will make it easy, and none should.
That’s the work.
Now consider everything you had to do to get to the point where you could think that thought. Locating the nephrology note. Determining which creatinine was most recent, and whether it was drawn before or after the last dose change. Remembering, unprompted, that you’d asked for a BMP two weeks ago and never saw it. Figuring out where in this chart the previous clinician recorded her reasoning — if she recorded it at all.
That’s the work before the work. And unlike the first kind, it is entirely reducible. No clinical value is created in any of it. It is friction, and every minute of it is a minute stolen from the thing you’re actually for.
Sociologists have a name for this. Anselm Strauss called it articulation work — the labor of assembling tasks, sequences, and clusters of tasks so that the real work can flow — a concept he developed through research on the social organization of medical work. Cognitive scientists would call much of it extraneous load: difficulty imposed by how information is arranged, as opposed to intrinsic load, the difficulty inherent in the problem itself. The patient with three interacting chronic diseases is intrinsically hard; that load is the job, and we signed up for it. The four minutes spent locating a cardiologist’s assessment is extraneous — pure format tax. It contributes nothing to the patient, teaches you nothing, and consumes the same finite attention needed for the intrinsic part.
We like the blunter name: the work before the work. Physicians do an enormous amount of it without ever having agreed to.
9.2 Hunting is not reading
Here is the texture of that work, in one contrast.
A basic metabolic panel takes an experienced clinician about two seconds to review. Not because we’re fast readers — because we’ve seen that grid ten thousand times. Sodium is where sodium always is. The eye lands on the creatinine without being asked. We aren’t reading it so much as absorbing it: the numbers arrive pre-sorted, and every ounce of attention goes straight to the only question that matters — is this different, and does it change what I do?
Now hand that same clinician a three-page consult note from a cardiologist at a health system on a different EHR. Four minutes, sometimes longer. Not because the note is harder to understand — the medicine in it is often simpler than the BMP. It takes four minutes because before you can evaluate a single clinical statement, you have to find it. Where did this practice put the assessment? Is the plan at the bottom, or embedded mid-text? Is the med list current, or imported from an intake form in 2021? Is that bolded heading meaningful, or template garnish?
That’s not reading. That’s hunting.
Chart review, in other words, is a re-orientation problem, not a reading problem. We talk about it as though the difficulty were volume — too many notes, too many results. Volume is real, but it isn’t what exhausts you. What exhausts you is that every document you open asks you to rebuild your mental map from scratch. Each specialist has a house style; each EHR renders outside records differently or dumps them as undifferentiated text; each institution templates to its own billing needs. So before any clinical thinking can begin, you run an unconscious subroutine: what kind of document is this, who wrote it, what’s its shape, where is the part I need? Then you run it again on the next document. Twenty times a day, across twenty patients, for years.
The clinical reasoning — the part you trained for, the part that’s actually hard and actually valuable — is the small bright thing at the end of a long dark hallway of orientation. Most of the effort is spent walking the hallway.
9.3 The reconstruction tax
The same hallway runs through every visit with a complex patient.
Every patient has a story — not in the sentimental sense, but the clinical one: a sequence of problems, interventions, responses, and adjustments that, taken together, explains why they are sitting in front of you today. That story exists in the chart. But the chart doesn’t tell it. It archives it. The physician does the telling, reconstructing continuity from fragments, visit after visit.
Before making a good decision about a patient you haven’t seen in six months, you need context. Which medications did we try for the hypertension before this one? What has the A1c trajectory been — improving, stable, drifting? Did cardiology ever weigh in on the spring’s chest pain? All of it is in the chart. Retrieving it is archaeology: prior notes opened in reverse chronological order, scanned for relevant details, held in working memory while the next patient waits. A thorough review of a complex chart takes five to ten minutes. At twenty patients a day, that time does not exist. What happens instead is partial reconstruction — good enough to proceed, not the same as knowing.
Call it the reconstruction tax. It is one of the most expensive, least discussed inefficiencies in medicine, and it has a reliability cost as well as a time cost: human synthesis is inconsistent. The physician who knows the patient and carries context in memory practices differently from the covering colleague seeing the patient cold. Neither is failing — they’re operating with different inputs. But the quality of care should not depend on who happens to remember what.
9.4 The private operating system
Why does memory carry so much of the load? Because somewhere in the third or fourth year of practice, every good physician stops relying on the system and builds their own.
You develop a shorthand. A place in the note where you always put the thing you’re worried about. A way of phrasing an assessment that tells you, on re-read, exactly how confident you were. You learn that Mrs. R’s daughter is the one who actually fills the pillbox, so any med change gets relayed through her. You learn which cardiologist calls back. You learn that when this patient says he’s “fine,” it means something different than when that one does.
None of that is in a field. There is no discrete data element for the daughter is the real historian or I don’t trust this A1c. It lives in your notes as shorthand, if it lives anywhere — and mostly it lives in your head.
This private operating system is not a bug. It is a mark of expertise — the accumulated, compressed judgment of thousands of encounters, and exactly what separates the physician who has known a patient for eight years from one meeting them cold. It is also completely non-transferable, and that is the problem. Consider where it fails:
- Coverage. Your partner sees your patient and misses the thing you would have caught — not through incompetence but through inaccessibility.
- Handoff. You compress eight years of context into ninety seconds of sign-out and hope.
- Team-based care. The pharmacist doesn’t know you were already planning to back off the diuretic. The care manager doesn’t know why the referral matters.
- Turnover. A physician leaves and takes a thousand patients’ worth of tacit knowledge with them. We treat this as unremarkable. It is catastrophic.
- You, eight months from now. The most common failure. You reread your own note and the shorthand no longer decompresses, because the mental state that produced it is gone.
Every one of these is the same failure: a system that works brilliantly and privately, encountering a world that requires it to work publicly. Medicine is being asked to care for more patients, with more complexity, with fewer physicians per capita, across more distributed teams — a rotating cast of partners, coverage, NPs, pharmacists, care managers, and 2 a.m. hospitalists. The old model assumed a single physician holding a single patient’s story in a single mind. That model is quietly, thoroughly obsolete — but we never replaced its infrastructure. We just added people to a system whose intelligence still resides inside individual skulls.
Brains don’t scale. Systems do. And the note can’t hold what the brain is holding, because the note was never built to carry it: it is an artifact produced by reasoning — compressed for billing, for legal defensibility, for the author’s own future reference — not a vessel for shared thinking. Read a colleague’s note on a shared patient: you can extract the facts, but you cannot reconstruct the thinking. Why did she stop the beta blocker — considered decision, or reflex to a soft blood pressure? Is “will monitor” a plan or a placeholder? You don’t know, and the note can’t tell you.
9.5 Why nobody sees it
The work before the work is structurally invisible, and the invisibility is what keeps it from being fixed.
It doesn’t bill. There is no code for reorienting yourself to an unfamiliar note format.
It doesn’t get measured. Time-and-motion studies capture “EHR time” as one undifferentiated block — 5.9 hours of an 11.4-hour day in the Arndt data. That number is cited constantly. What it cannot tell you is how much of the time was spent deciding versus hunting for something to decide about, because no category exists for the difference.
It doesn’t get taught. No one trained you in chart archaeology. You developed your methods alone, badly, under time pressure — while assuming everyone else has a better system. (They don’t.)
And expertise disguises it. A senior physician performs this work so fluidly it stops registering as work. Watch a good attending open a chart: they know where to look, what to ignore; they reconstruct three years of care in ninety seconds. It looks like mastery. It is mastery — of something that should never have required mastering. We have spent decades treating physician skill at navigating bad systems as evidence that the systems are acceptable.
9.6 Where it actually hurts
In 2018, Simon Talbot and Wendy Dean published an argument that changed the vocabulary of this conversation. Physicians, they wrote, are not burning out. Burnout implies exhaustion, cynicism, decreased productivity — and worse, it implies a failure of resilience in people who have spent decades proving they have more resilience than almost anyone alive. What physicians are experiencing, Talbot and Dean argued, is moral injury: the injury of being unable to provide the care and healing you know the patient needs. Not one catastrophic failure — the steady accumulation of routine, incessant betrayals, each of which would heal in isolation but which arrive daily.
The conversation that followed has mostly stayed at altitude: prior authorization, productivity targets, private equity, the business of medicine. Those are real, and they were the authors’ primary targets. We want to argue for something smaller and closer to the keyboard.
Some of the injury is in the friction.
It’s knowing the patient needs a repeat BMP in two weeks, and knowing that whether she gets it depends on whether you personally remember — across hundreds of patients, across a month, with no system holding it for you. It’s opening a chart and knowing the answer to your question exists somewhere in it, and that finding it will cost four minutes you don’t have — so you decide without it. It’s the moment you realize you’ve spent your evening as a human indexing service. Each instance is small, individually forgettable. But you make dozens of these micro-compromises every day, and each one is a gap between the care you know how to give and the care you actually gave. That’s not fatigue. Fatigue is what happens when you’ve done too much of your work. This is what happens when you’ve been prevented from doing it.
Days like that don’t leave you tired so much as unspent — full of capacity you never got to use.
9.7 The objection we have to answer
Talbot and Dean were emphatic that the usual remedies are worse than useless: wellness officers, resilience training, mindfulness modules. Their objection wasn’t that these are unpleasant; it’s that none of them touch the patterns doing the injuring. They ask the wounded party to cope better.
So when we claim that better documentation infrastructure would reduce moral injury, we owe you a reason this isn’t just a nicer-smelling aromatherapy inhaler.
The distinction is this: does the intervention change the work, or does it ask the physician to tolerate the work?
A mindfulness module doesn’t remove a single click; it asks you to be calmer while clicking. A resilience seminar doesn’t shorten your inbox; it asks you to be sturdier while emptying it. These interventions locate the problem in the physician — which is precisely the move that makes physicians so bitter about them, and rightly.
Removing articulation work is a different category of act. When the follow-up you set in your assessment becomes a tracked task without a second entry, and resurfaces when it should, that work is gone — not tolerated, not reframed, not meditated upon. Gone. When every note you review arrives in the same shape, the four minutes of hunting don’t become easier to bear; they cease to exist. That’s not wellness. That’s changing the work — exactly what the moral-injury argument said the situation demanded.
And one thing has become true since 2018 that’s worth saying plainly: this class of problem is tractable in a way that the business of American medicine is not. We don’t know how to fix prior authorization. We have a fairly good idea of what it would take to stop making physicians hunt for the assessment.
9.8 What’s left when you take it away
We want to be careful about what we’re promising, because medicine has been promised a lot.
Removing the work before the work will not make clinic easy. The patient with three interacting chronic diseases will still be hard. The goals-of-care conversation will still be hard. Uncertainty will still be uncertainty. If all of it were fixed perfectly, the job would remain one of the most cognitively and emotionally demanding things a person can do.
That’s the point. What’s left, when you strip away the hunting and the remembering and the double-documenting, is the thing you actually trained for. The hard part. The part that was worth the decade.
Part III is about building that. Two directions out, which the next chapters develop in turn: externalize the system — build infrastructure that holds the work of care, so it doesn’t have to be held in a person (Chapters 10 and 11) — and normalize the review surface — bring every document into a predictable shape so attention goes to content, not navigation (Chapter 12). Externalizing makes the work transferable. Normalizing makes it reviewable. Together they attack the same target from opposite ends: the enormous, unpaid, unmeasured, unglamorous labor that stands between a physician and the practice of medicine.