The Work Beneath the Conversation: The Workaround

The last post ended on an uncomfortable objection: maybe articulation work prices at zero because the system has correctly worked out that most of it compensates for failures that shouldn't exist, and funding the compensation would fund the fragmentation.

That objection just might be stronger than I allowed. Researchers named the behavior by watching hospital nurses. First-order problem solving: you fix whatever is blocking the task and you get back to the patient. It works. The patch holds, the patient gets what they need, and the failure that produced the blockage stays where it was, invisible now to anyone with the standing to remove it.

Hospitals have always run on that. We are the version of it that got organized into a service. Palliative care is assigned responsibility for making care coherent, holds authority over almost nothing that generates the incoherence, and every gap we absorb well tells the institution the gap has been handled. We're the colloquial FBI, never getting credit for the problems we prevent.

What Nurses Do With Failure

Anita Tucker and Amy Edmondson put observers on nursing units at nine hospitals, most of them chosen for reputations in nursing excellence, and logged every breakdown in the work system — supplies that weren't there, information that hadn't arrived, equipment somebody else was using, orders nobody had written.

The nurses fixed nearly all of it. Their governing rule, applied to 93 percent of the problems, was to do whatever it took to continue the patient care task, no more and no less. A nurse short on linen walked to another unit and took theirs. Only 7% did anything at all that might have started an inquiry into cause...and the authors were being generous about what counted.

The nurses also felt good about it, and the feeling does structural work. Most of those interviewed described satisfaction at having worked around an obstacle. Tucker and Edmondson trace the loop: effective patching produces gratification, gratification lowers the impulse to involve anyone else, and involving nobody guarantees the failure returns. They call the resulting stability illusory. Absorption keeps pace with production, and from inside the unit that looks like a system in balance.

Ninety-one percent of the failures originated in something transferred to the nurse from elsewhere — another department, another shift, another service. The people who could have corrected them were the people least likely to hear about them, and the nurses' second rule was to ask for help from whoever stood closest socially rather than whoever held the authority to change anything.

Tucker and Edmondson state the arrangement in their opening paragraphs: hospitals have relied on a dedicated and highly skilled professional workforce to compensate for operational failures during care delivery.

The Consult as a Patch

Go back to the composite case from the last post and sort what happened.

Two services adjusted the same diuretic and neither reconciled it. A cardiologist and an inpatient team gave one family different prognoses for a month. Both are failures of a work system, and both got fixed at the level of the individual patient. The diuretic was reconciled for him. The prognostic range was aligned for his family.

Nobody asked why two services can change the same medication with no reconciliation step, or why a month of discordant prognostication produced no signal anywhere in the institution. There was no time in which to ask and nothing that would have received the answer. Tucker and Edmondson found the same about their nurses, who were neither careless nor uncommitted — nine of ten observed for a full shift stayed an average of forty-five minutes past the end of it, unpaid, to finish their patients' care. What those units lacked was any route by which a fixed problem became a known one.

The other two findings that week were different in kind. A son sleeping in the visitor lot, and a patient who no longer knows who he is meant to be in the world, are not products of a broken process, and no redesign removes them. That is the irreducible work — the part of this specialty that exists whether or not the institution ever functions properly.

Both kinds went into the chart as one consult.

That is the whole problem in miniature. Process defects and irreducible clinical work are indistinguishable in the only record anyone reads, so the half of the week documenting a system failure gets filed under the same heading as the half documenting serious illness. Both generators are still running. They produced the same week for somebody else two weeks later, and the institution's account of the first encounter says a palliative care consult answered a goals-of-care question.

Articulation work prices at zero partly because the system has never been shown the bill. What it sees is a solved case.

What We Have and What We Don't

The distance between what we are asked to produce and what we can compel has a specific shape.

We have some standing in our health systems. People often do what we suggest because they trust us, which is real and worth having. We hold epistemic authority inside the individual patient case: when we say what this patient understands and what they want, we are generally believed. Both operate one patient at a time and dissipate when that patient leaves.

Little of that is durable. We have no say over the discharge target, the staffing ratio, the contracted agency network, the capacity plan, or what an oncologist says at the next visit. Strauss's distinction from the last post covers it precisely — rights hold only where legal, financial, or organizational resources exist to enforce them. Ours reset to zero at discharge.

Absorption is what remains when a service gets handed a systems problem and given clinical tools.

Four Failures We Absorb

Each of these arrives as a consult. Each gets a first-order fix. And each distributes unevenly enough that the absorption performs quiet redistributive work nobody has priced.

Prognostic fragmentation. The patient has been given three pictures and we are asked to produce one. Reconciling them is skilled work and it changes nothing about how the next three get made. The distributional problem sits inside the reconciliation itself. Thornton and colleagues compared interpreted ICU family conferences with non-interpreted ones and found that families with non-English-speaking members received less information about the illness and less emotional support from clinicians. A systematic review of interpreter use in palliative care found clinicians routinely relying on family members, including children, to convey diagnosis and prognosis. The families most likely to have received fragmented information are the families where our reconciliation tool works worst.

Choice without feasibility. Someone offers an option the patient cannot take, and we are called to manage what happens when the family holds them to it. Home hospice presumes an unpaid caregiver in the house; Medicare covers no room and board and no ongoing custodial care, and the respite benefit runs a few days at a time. Whether home is available depends on whether somebody in that family can stop earning. We describe this as a preference-sensitive decision. It functions as an income test, and I've written about what it costs the people who pass it in Hospice Caregiver Burden and the Systematic Delegitimization of Care.

Conflict after premature offering. A procedure, an escalation, or a trial of something got offered before anyone assessed whether it fit, and we arrive to walk it back. For families with prior experience of being undertreated, a walk-back reads as rationing. Sometimes that reading is correct, which is the part the field does not say out loud. We are asked to repair a breach of trust we did not cause, cannot prevent, and are badly positioned to disprove. I like to call this flavor of consult, "please talk them out of the thing we're offering them."

Transition failure. The plan depends on a service that never materializes — the agency with no capacity, the coverage that lapsed in March, the equipment requiring a signature nobody can locate, hospice services that don't exist in Tijuana. We make the calls. Coverage churn and agency deserts concentrate those calls in the same populations every time, so the programs absorbing the most generate the least billable evidence of having done anything.

The fix is real in all four and the patient benefits from it. The mechanism that produced the failure learns nothing.

The Gratification Problem

From the outside that loop is a diagram. From the inside it can be the best part of the job.

The consult that catches a bad discharge, the meeting that stops an escalation nobody wanted, the call that finds the coverage problem on a Friday afternoon — those are the shifts we drive home satisfied with. The satisfaction is accurate about the patient and misleading about the institution. It is also why the report rarely gets written.

The delayed cost in their model is burnout, and the delay is what makes the equilibrium look stable. In the field's own survey of AAHPM members, roughly two in five palliative care clinicians met criteria for burnout.

It's news to no one that health systems largely treat burnout as a resilience problem and answered it with resilience programming, which is a first-order fix and works well enough that nobody is forced to look upstream. When a team reports the same distress about the same recurring failure, that report is an operational signal about a specific process, and it should reach whoever owns that process.

What Second-Order Looks Like Here

Second-order problem solving, in their definition, means patching the immediate problem and also telling someone who can change the cause. Three versions for a palliative care service.

Code the generator, not just the consult. The last post argued for logging between-encounter work as its own category. Extend it by one field: what system failure produced this consult. Prognostic discordance across services. Plan offered without feasibility assessment. Referral placed and unstaffed. Coverage lapse discovered at the bedside. Four checkboxes and a free-text line, filled in by whoever closes the consult. Leave the box empty when nothing failed and the patient simply has a serious illness, because that distinction is exactly what the record currently destroys.

Route the report where the levers are. Palliative care often and comfortably sits on ethics committees and clinical quality committees, which are the bodies least able to change any of this. The incoherence comes out of capacity management, discharge planning, contracting, and service-line scheduling. Ask for a seat at one of those — non-voting is fine — and bring a quarterly generator report to it. Tucker and Edmondson's first lever for change is managerial presence, on the grounds that managers hold the status to resolve problems crossing departmental boundaries. Their second finding is why the seat has to be real: organizations that never respond to escalation train their people to stop escalating, and one nurse's summary of that state was nothing ever changes around here.

Assess feasibility before presenting a choice. Establish whether there is a caregiver, whether in-house GIP is even appropriate to offer, whether the agency is staffed, and whether the coverage is active, before any plan gets offered as an option. This is a clinical change with an equity yield, because it stops us manufacturing consent to options that exist only for people with resources. It also generates the data for the first recommendation: when the answer is no for the eleventh time in a quarter, that is a finding.

None of this requires authority we don't have. All of it requires accepting that a well-absorbed failure is information the institution is entitled to and has never been given.

The Objections That Land

So we should let things fail to prove a point. No, and this is the reading I most want to prevent. Tucker and Edmondson's recommendation is patch and report, never withhold. The patient in front of you gets the fix. The difference between first- and second-order lives in the twenty minutes afterward, not at the bedside. Anyone reading this as an argument for strategic incompetence has inverted it.

Every service does workarounds. True, and my answer differs from the one I gave last week. For us the trouble is incentive. No cardiology division has ever been asked to justify its existence by counting the bad discharges it rescued. We have. So reporting the generator argues against our own volume, and the better we document that these failures are preventable, the more we invite the question of why the service exists. I don't have a way out of that which doesn't require somebody above us to want the report.

This is unfalsifiable. It isn't. Run the generator coding for four quarters. If the same generators don't recur — if prognostic discordance appears once in a service and never again — the pressure-relief argument is wrong and the consults were catching genuinely novel events.

Where I Might Be Wrong

Tucker and Edmondson studied nurses and supplies. Most of their failures involved a missing object or a missing piece of information, both of which have clean causes and clean fixes. I may be importing a manufacturing model into a domain where the share that resists redesign is much larger than I have allowed, and the composite case suggests it might be half.

The strongest version of that concerns prognosis. Some of what I've filed as fragmentation is legitimate clinical disagreement under real uncertainty, and a system engineered to produce one number would be a system lying about what it knows. I've written prognostic discordance as a defect. Some meaningful share of it is honesty, distributed badly.

The pressure-relief claim rests on a counterfactual I cannot test. Nobody has run the study where palliative care declines to absorb and we watch whether anything upstream moves. The comparison I want doesn't exist and probably shouldn't.

Reporting generators costs more than I've said. A service arriving at operational meetings with a quarterly list of other departments' failures becomes an adversary, and adversaries lose budget reviews. The recommendation is cheap in effort and expensive in political capital, and I've written it as though only the first were true.

And the objection I find hardest is about my own consistency. Two posts arguing against the unbundling of our work, and this one asks for structural influence over how other services discuss prognosis and build discharge plans. That is a jurisdictional claim on other people's practice, which is the move I object to when it lands on us. I don't think the asymmetry defends itself on principle. The best I can say is that I'm asking for a reporting channel rather than a veto, and the distinction is thinner than I'd like.

Final Thoughts

Palliative care is very good at this. That is the entire problem and none of the solution.

The institution reads our competence as evidence that the arrangement works, because competent absorption looks identical to an absence of failure. Every patch applied without a report says the same thing upstream: nothing needs to change here.

Fix the case, then tell somebody who can fix the cause. Twenty minutes and a checkbox, and it's the only part of this work the institution has never been shown.

I am a palliative care physician, educator, and professional strategery expert. Known for turning rounds into rants and rants into teaching points. Rounds & Rants represents my views — not those of any institution or professional membership organization where I hold a role. I don't write on their behalf and they don't vet what I publish.