The Question That Ends With a Period

"How do you feel about stopping the PCA?" "What would you like to do?"

While these may sound like good communication, they might just be dishonest if I have already concluded that continuing the PCA is not a safe plan.

Let's say patient has had a PCA through a stretch of severe pain. The acute phase has resolved enough that the transition to an oral regimen is now the clinical work in front of us. The patient does not want to do it.

Not because they're confused. From where I stand, a PCA is a delivery mechanism for opioid. From where the patient lies, it may be the only thing in the building they control with a message that shouldn't be underrated: the team believes your pain is real, and they gave you a machine that proves it. Patients do not need to misunderstand a plan in order to dislike it.

So, if I ask them how they feel about it, I might appear to hand them a decision I have already made.

Here's what I teach: as much as possible, questions should end in periods, not question marks.

The question mark can lie

An open-ended question is not respectful by default. Its ethical value depends entirely on whether the options are actually open.

Jack Wennberg came at this from a different direction entirely. Working out of Dartmouth in the late 1960s, he and Alan Gittelsohn showed that procedure rates varied wildly between demographically similar populations, and that the variation tracked physicians' beliefs more than patients' needs. The taxonomy that came out of that work still holds up. Effective care has proven value and no significant tradeoffs; every patient with the indication should get it. Preference-sensitive care involves real tradeoffs, and the right answer depends on how a particular patient weighs them. Lumpectomy or mastectomy. Dialysis or conservative management. The fifth line or the last good month.

Shared decision-making belongs in the second category. But Wennberg was sorting decisions, and a single bedside conversation almost always holds both categories at once. "The PCA needs to come off" is effective care — the evidence about transitioning off parenteral opioids before discharge doesn't turn on this patient's values. "Tonight or tomorrow morning" is preference-sensitive — it turns on almost nothing but this patient's values. One conversation. Two decisions. Two structures.

Our failure isn't that we apply shared decision-making to the wrong decision. It's that we don't notice we're holding two decisions and we put a question mark after both.

Thirty years ago Tim Quill and Howard Brody made most of this argument and named the barrier precisely: the "independent choice" model, in which the physician lays out options and odds and withholds their own experience to avoid contaminating the patient's judgment (I prefer to call this "mandatory autonomy"). They wrote that this model confuses independence with autonomy, and that it sacrifices competence for control. That was 1996. It is a measure of something unflattering about medicine that the analysis still lands.

Grammar conceals power. It does not remove it. I have already exercised judgment by deciding what to offer, what to continue, what to stop, and turning that judgment into a question just makes the patient responsible for excavating a conclusion I already reached.

Which is the uncomfortable part. Sometimes (often?) "what do you want to do?" isn't shared decision-making at all. It's a clinician declining to show their work.

Who gets told and who gets asked

Everything above tells clinicians to state opioid conclusions more confidently. In a specialty that has spent a decade watching stewardship curdle into undertreatment. About a class of drug where the evidence says our clinical judgment is not reliable in the first place.

In 2012 Salimah Meghani and colleagues synthesized twenty years of evidence and found Black patients carried both more analgesic treatment disparities and larger ones than any other group — consistent across pain types, settings, and time period. Their verdict on the trajectory was blunt: the gap was not closing. Twelve years later Meghani co-authored the update, covering 2011 to 2021 — the exact window in which our profession supposedly got serious about prescribing equity. Black patients were still less likely than White patients to receive opioid analgesia; Hispanic patients still less likely than non-Hispanic White patients. In acute pain it runs larger: Lee and colleagues found Black patients substantially less likely to receive analgesia at all.

Kelly Hoffman and colleagues put it in PNAS in 2016: half of a sample of White medical students and residents endorsed false beliefs about biological differences between Black and White people — that Black skin is thicker, for instance. Those who endorsed them rated a Black patient's pain lower and made less accurate treatment recommendations. Not less generous. Less accurate. People with medical training, reasoning from folklore, filing it under clinical judgment.

So the objection writes itself. If I tell clinicians to convert their conclusions into declarative sentences, and their conclusions about opioids are demonstrably inequitable, I have not improved communication. I have made bias more efficient and given it better grammar.

The objection is correct about the danger and wrong about the remedy — and the reason matters more than the conclusion. The question mark is not a safeguard. It is camouflage. When I ask "how do you feel about stopping the PCA?" and I have already decided, nothing about my reasoning is on the record. The patient can't contest it, because I never stated it. A colleague can't audit it. I can't be held to it, because I never claimed it. If my judgment was contaminated by a belief about whose pain is real, the question mark is the perfect place to hide that.

A declarative sentence with reasoning attached is a claim. Claims can be wrong out loud. "Continuing the PCA past tomorrow morning is not a safe plan, because it will make it harder to establish a regimen you can actually use at home" is falsifiable, arguable, answerable. The patient can say that's not true, my pain is a nine right now. The nurse can say have you looked at the overnight scores? I can be caught.

Which produces the real rule, and it is a burden rather than a license: if you cannot defend the conclusion, you have not earned the period. A clinician who wants to be done for the day can call anything medically unacceptable — and the fastest way to catch them at it is to make them say the sentence and finish it.

The stewardship overcorrection didn't happen because clinicians were too directive about opioids. It happened because a system built pressure to taper and gave clinicians vague language to do it with. "Let's think about weaning" carries no reasoning and no author. That's not humility. That's a decision with the fingerprints wiped off.

Carrots or peas

So the period goes after the clinical determination. Not after the conversation.

"Continuing the PCA beyond tomorrow morning is no longer a safe plan." Period. That sentence is mine, I'll defend it, and the patient shouldn't have to pry it out of me.

"We can stop it tonight or tomorrow morning. Which do you want?" Question mark. That one is genuine, because either answer is fine with me.

I call this the carrots-or-peas approach, deliberately and somewhat obnoxiously, because I have shepherded 3 kids through their 2 and 3 year-old stages. The point is not that hospitalized adults are toddlers. The point is that a useful choice has edges. A parent narrows a child's options partly because the child's judgment isn't mature. A clinician narrows an adult patient's options because some paths are medically wrong — a fact about the medicine, not about the chooser. The adult keeps every right the analogy doesn't touch: reasoning, argument, questions, refusal, and my continued attention when they use all four. The structure carries over. The status of the chooser does not.

The rule is: offer materially distinct, clinically reasonable choices — and no imaginary ones. Often that's a binary. Sometimes three. Sometimes it collapses to one recommendation plus the patient's standing right to refuse it, which is not nothing.

Then the sequence, which takes about forty seconds. State the boundary. Explain the reasoning. Name what it costs them — "I know giving up that button feels like giving up control, because it is." Then hand back what's left.

What's left is small in clinical magnitude and can be enormous in subjective weight. Tonight or tomorrow morning. Before or after your daughter's flight lands. Whether your family is in the room when we do it. Size is not what makes a choice autonomy-supporting; whether the person endorses what's happening is. A small real choice beats a large fake one.

Fake ones are easy to manufacture, because they cost nothing and look generous. A choice is counterfeit when:

• You have a preferred answer and you're presenting both as equal.
• The boundary was drawn by operational convenience — bed flow, staffing, your clinic schedule — and you're describing it as clinical.
• A reasonable third option got left out because it would be annoying.
• The choice won't actually be honored when the night team comes on.

One test catches most of these: if either answer would frustrate you, you are not offering a real choice.

Run it before you open your mouth. If the patient says "tonight" and you feel relief, and they say "tomorrow morning" and you feel your jaw tighten, you were never offering a choice. You were running a preference through a question mark and hoping it came back the way you wanted. That's a nudge wearing a lab coat, and patients — particularly patients who have spent a lifetime being managed by institutions — are better at detecting it than we are at performing it.

None of this means the patient must agree, or that I can't be questioned, or that my assessment can't change by morning. It means I reached a conclusion, I'll say it, and I'll carry it.

Where I might be wrong

The extension is mine, not Wennberg's. He built those categories to explain population-level variation in utilization, using claims data and small-area analysis. Nothing in that work is about how to construct a sentence at a bedside. I think the logic carries — the epistemics of a decision don't change with the unit of analysis — but I'm stretching a policy taxonomy across a gap it wasn't designed for, and someone who knows the variation literature better than I do may tell me the seam shows.

The best data on the nearest question cuts against a universal rule. Doug White and colleagues showed 169 surrogates two videos of a life-support conversation, identical except for whether the physician offered a recommendation. Fifty-six percent preferred to receive one. Forty-two percent preferred not to. I'd note the decision they studied was preference-sensitive — whether to limit life support — which is exactly where I'd keep the question mark anyway. But 42% is not noise, and if that split holds for effective-care decisions too, my rule needs a front door: would it help to hear what I think? That one genuinely ends in a question mark, and I don't think it's a dodge.

I don't have PCA-specific disparity data. I looked. It isn't there. I'm reasoning from the general analgesia literature to a specific device decision, and that's an inference, not a finding. If someone has the data, I want it — including if it points the other way.

Put the punctuation where it belongs

The patient picks tomorrow morning.

They're still unhappy. They want to know what happens if the oral regimen doesn't hold. They ask about breakthrough dosing. They try to negotiate the time to noon, and I say no, and they're annoyed about it.

None of that invalidates the period. All of it is the conversation the period made possible, because we're now arguing about the actual question instead of performing a decision that was never open.

And nobody has to like it. Validation does not require moving the boundary; empathy does not require inventing an option that doesn't exist. An unhappy patient who feels heard, understands what happens next, and knows precisely which parts were theirs is not a failure. In our field it is frequently the ceiling, and pretending otherwise is how we end up manufacturing choices nobody can deliver.

There are two options that are good for you. Both are disagreeable. My job is not to invent a third or to make you pronounce the medical conclusion so I don't have to be the one who said it. My job is to tell you why the boundary exists, in language you can argue with. To say plainly what it costs you. And to hand back every piece of this that can safely stay yours.

The period closes the recommendation. It does not close the conversation.

Related: I've written about how the vocabulary of shared decision-making is being hollowed out from outside the field, and about the autonomy reflex itself in the Beyond Mandatory Autonomy series — Part 1 and Part 2.

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.