Street Medicine: "Stable or Unstable" vs. "Sick or Not Sick"

The kid was breathing thirty-two times a minute. Every third breath had a soft grunt on the exhale you had to lean in to catch. Two years old. Dad had called 911 for “she just doesn't seem right today.” The monitor looked survivable. Heart rate 148, which in a two-year-old could mean anything from crying-in-the-driveway to pending arrest. Saturation 96 on room air.

And the medic did what a decade of textbook trained him to do. He looked at her and asked himself the question that every EMS program puts at the front of the assessment: is she sick, or not sick?

He said not sick.

She was on high-flow oxygen in a PICU four hours later.

He was not a bad medic.

His assessment was competent.

His vitals were accurate. His documentation was clean.

The problem was the question they first asked.

The Question That Answers Itself: Stable or Unstable?

Everyone who calls 911 is sick.

It might be loneliness. Chronic pain that finally broke through a tolerance they had been managing for years. An argument with a spouse that walked into a panic attack. An intoxication that started at noon. Or septic shock.

In every one of those cases the patient dialed the number that means *I need help.* Something is wrong. That is what the word means to them, and by that meaning the answer is always yes - they are sick.

That is obviously not what we mean when we teach it. When an instructor asks a student, “is she sick or not sick,” they mean is she physiologically threatened. Two different questions wearing one word.

The failure is not the ambiguity by itself. It is the collapse. Because the two meanings share a word, the clinician's verdict carries both at once — a judgment about physiology and a judgment about whether this person rates our attention.

Say “not sick” out loud on a real call and listen to what rides along with it. That is how a service ends up with an unwritten category for the patients who don't count.

“Unstable” cannot do that.

It is a claim about a system state, not a verdict on a human being. The word has no room in it for contempt.

Fuzzy Math With No Variables

Bart Kosko's work on fuzzy sets makes a point worth borrowing here: membership in a category is a degree between 0 and 1, not a switch. “Sick” is obviously fuzzy. Kind of sick. Really sick. Circling the drain. Fine — fuzziness is not the problem.

The problem is that in the entire history of the phrase, nobody has defined the membership function. What are the variables? What are the thresholds? What operation maps an observation onto a degree of sickness? There isn't one. Which means when two clinicians disagree about whether a patient is sick, there is nothing to adjudicate it with. No appeal. It is not a measurement. It is an opinion with clinical consequences attached.

Stability is fuzzy too, and I am not going to pretend otherwise. How stable is a real question with a real answer. A heart rate of 180 sits further from the middle than 110. A pressure of 110/40 is a MAP problem whatever the systolic looks like. Respirations of 8 is a different animal than 16. That is degree of membership — and it is computed from named variables against published thresholds.

Same fuzziness. Only one of them has inputs.

The Frame Contradicts the Algorithm

We teach airway, breathing, circulation, control bleeding (typicallly control bleeding before all else - but is a matter for a different day). That is a stability algorithm. It is the first thing every student learns and the last thing any of us abandons under pressure.

Then we bolt a sickness-first frame onto the front of it. And when the two conflict, the frame wins — because the frame runs in the first ten seconds and the algorithm runs after.

Take the child with croup who has been working to breathe for six hours and is now exhausted. The etiology is interesting. It changes nothing about the next sixty seconds. We are going to support ventilation, bring up the saturation, and protect perfusion before they stop breathing, and we would do exactly that if the label were bronchiolitis or anaphylaxis or foreign body. The diagnosis is downstream of the intervention.

Nobody says out loud, “he's a little unstable, let me finish my assessment and circle back to that.” But a framework that opens by asking what is wrong with someone makes precisely that sequence feel reasonable.

Instability is not a finding you work up. It is a finding you interrupt for.

The PAT Was Never a Sickness Tool

The Pediatric Assessment Triangle gets filed in the literature as structured gestalt.

Look at what it actually measures.

Appearance — tone, interactiveness, consolability, gaze, speech or cry. That is neurologic status.

Work of breathing — position, retractions, abnormal airway sounds. That is respiratory effort.

Circulation to skin — pallor, mottling, cyanosis. That is perfusion.

Neuro, breathing, circulation, assessed from ten feet away. The PAT is a stability instrument. It got filed under a sickness label because of the vocabulary that happened to be in fashion when it was written.

Its clinical logic is stability logic too. The PAT exists because children compensate beautifully and then stop compensating all at once. When all three domains are abnormal, no one asks what she has first. That question comes after, and the answer does not change what your hands do.

The tool was right.

The teaching language wrapped around it was wrong. In classroom after classroom the student is asked “is this child (or person) sick?” when the instrument in front of them is measuring whether the child is still holding.

What the Gestalt Research Actually Shows

The strongest evidence for clinician gestalt is Kline's work comparing unstructured physician judgment against Wells and revised Geneva for pulmonary embolism — gestalt with an AUC of 0.81 against 0.71 for the rule. It gets cited a lot in this debate, usually loosely. Worth being precise about what it says.

Those physicians were not guessing. They had history, vital signs, risk factors, a physical exam, and years of pattern exposure, applied to patients who had already walked into an emergency department. That is expert probability estimation for one named disease in a pre-filtered population. It is skilled work and the finding is real. It is not evidence that a global sickness impression is reliable in a driveway at 0300.

There is a second transfer problem that matters more. An emergency physician's impression is a hypothesis that gets tested within the hour by labs and imaging they ordered themselves. I have never met an emergency physician who eyeballs a patient, orders nothing, and sends them home — not even the frequent flyer, and the frequent flyer deserves that workup as much as anyone. In the ED the impression is the beginning of a process.

Prehospital, the same cognitive act is the end of one. There is no confirmation step. Nothing downstream catches a wrong call before it costs something. Same act, entirely different consequence.

But the deepest issue is not validity at all. It is this: the "sick - not sick" question can be answered without data, so it gets answered without data. Nothing in “does she look sick” requires you to touch the patient, count anything, or put a cuff on an arm. You can answer it from the doorway, and you will be right often enough to keep doing it wrongly. A question that demands no work will not get any.

Unstable You Can See. Stable You Have to Earn.

The two options are not symmetric, and we have been treating them as though they are.

Unstable is sometimes obvious from across the room. Agonal respirations. Mottling to the knees. A silent chest. An obtunded patient who was talking to dispatch nine minutes ago. No instruments required.

Stable is an assertion. It says: I have looked at this person's physiology and nothing in it is currently failing. You cannot get there from the doorway, because the only way to be lethally wrong about stability is compensation — and compensation is invisible by definition. The compensating patient looks like the well patient. That is what compensating means.

The error is asymmetric in cost as well. Call an unstable patient unstable and you spend some resources. Call a compensating patient stable and you spend the case at the next morbidity review.

Which is why the teaching question has to be two-part: are they stable, and how do you know? The second half is where all the work lives. Drop it and “vitals in range” becomes the answer, and we have rebuilt the same trap with a better vocabulary.

The Warning: “Stable” Is Already Being Hollowed Out

Before anyone adopts this wholesale, an honest caution about the word we are promoting.

“Patient is stable” already means, on a lot of radios, "nothing interesting is happening and I don't need anything from you." It is filler. It has been drifting toward a checkmark for years, through the exact mechanism that ruined “sick” — said constantly, defined never.

If we swap the front-end frame and don't publish the criteria and the reassessment obligation alongside it, we get the identical failure in five years with a nicer word. “He was stable” as a shrug. A QA reviewer with nothing to review. A stability call that is not backed by named physiology is the same vibe wearing a lab coat.

So the argument here is not use this word instead of that word. It is: make the front-end question one that demands an answer, then define what an answer looks like.

Three things keep it honest.

Named Criteria. Published, trained to, identical across the service. If your medics cannot recite what makes a patient unstable, you do not have a framework, you have a preference.

A Stated Basis. How do you know. Which numbers, which findings, which comparison. This is the half that survives into the chart.

A Shelf Life. Stable until when. Every stability call carries a reassessment interval and a named finding that would flip it. A call with no expiration date is a guess with better grammar.

A Working Triad

This is a proposal, not a validated instrument, and I would rather say so than dress it up. If you cannot declare stability without knowing physiology, then the declaration needs stated inputs. Three of them do the work:

Physiology. The numbers. Heart rate, blood pressure and MAP, respiratory rate, saturation, capnography, glucose, temperature, rhythm, GCS.

Patient State. What the body is showing you. Work of breathing, skin, mentation, position, speech. The domains the PAT formalized for children and nobody ever formalized for adults.

Trajectory. Where this patient was an hour ago and where the line is pointed. The 78-year-old whose documented 145 is now 118. Delta shock index. The second set of vitals that matters more than the first.

Physiology alone misses the compensating patient, because compensation is a normal-looking number. Pateitn State alone is gestalt with better manners. Trajectory is the leg that catches what no snapshot can, and it is the one we skip most often.

What “Unstable” Actually Means in the Field

The word gets thrown around loosely. Tighten it. Unstable in the prehospital environment means one or more of the following is true at first contact:

• Airway Compromised or Actively Failing. Not “the patient has a bad airway history” — the airway is failing now. Stridor, obstructed breathing, inability to protect against secretions, a GCS drop below the threshold where reflexes hold.

  
  

• Breathing Failing. SpO₂ under 92% on room air with respiratory distress. Respiratory rate under 10 or over 30 in an adult. Accessory muscle use, tripoding, one-word sentences, silent chest.

  
  

• Circulation Failing. Shock Index ≥ 1.0. Systolic under 90 with signs of hypoperfusion. Skin cold and mottled with a heart rate that doesn't match the story. A delta shock index of 0.1 or more between measurements. (**Circulation Failing also means we suspect or identify internal or external Hemorrhaging.)

  
  

• Neuro Failing. GCS ≤ 13 in a patient who was previously alert. New focal deficit. Active seizure. Sudden severe headache with a vitals shift.

  
  

• Metabolic Failing. Glucose ≤ 60 with symptoms, or ≥ 400 with signs of DKA. Core temperature ≥ 40°C with altered mental status. NEWS2 ≥ 7.

Any one of those and the patient is unstable, and the instability gets addressed before the workup continues — shorter reassessment interval, intervention urgency, transport priority, pre-alert format. Zero of them and the patient is provisionally stable, pending a stated reassessment interval and the structured second look.

That is a decision tree. It hands off. It reproduces. It writes down.

Five Cases, Two Frameworks

Rather than argue the frame in the abstract, run it. Each case gives you the same patient twice. Call it as sick or not sick, then as stable or unstable. The drill scores the two frameworks separately so you can see which one catches what.

You Cannot QA a Vibe

Here is the operational reason this outranks a semantic argument: your framework has to survive the seams.

Every institutional function EMS has — protocol, education, chart review, medical direction, destination selection, mass casualty — requires a claim that can be inspected. “He didn't look sick to me” is unfalsifiable. It cannot be reviewed, taught against, or corrected.: you cannot QA a vibe.

The worst day makes it obvious. START, SALT, and every mass casualty schema in current use run on stability physiology — respiratory rate, perfusion, mental status, ambulation. Nobody triages forty patients under a school bus by asking which ones look sick.

If your everyday frame doesn't translate into your worst-day frame, one of them is wrong.

Surprise Is Information You Currently Can't Use

The patient who was talking to you two minutes ago is in arrest, and you are stunned.

The surprise is due to lack of data gathing. It is telling you your model of that patient was wrong. But under sick-or-not-sick there is nowhere for the signal to go, because “he didn't look sick” is not a reviewable proposition. You cannot debrief it. You cannot teach against it. So the event lands as pure damage — the call that follows a medic home.

Under a stability call with a stated basis and a reassessment interval, the same arrest becomes a set of answerable questions. What was my call. What was it based on. When was I rechecking. What did I miss. That is the difference between an event that injures a clinician and an event that develops one. You cannot learn from a call you had no framework to be wrong about.

What This Doesn't Claim

There is no randomized trial showing that changing this vocabulary saves lives. There probably never will be — you cannot blind it and the confounders are enormous. Anyone promising a mortality benefit from a language change is doing the exact thing this article is arguing against.

What does follow, directly from the definitions: it changes what gets looked at, when it gets rechecked, what gets written down, and what can be reviewed afterward. Those are process claims. They are demonstrable. They are enough.

And the obvious objection, which deserves an answer rather than a dodge: does prove-stable drive over-triage? It shouldn't, because a stability call is not a transport priority. “Unstable” does not mean lights and siren; it means the instability gets addressed before the differential continues. The pathway consumes the call. It is not the call.

What to Do Monday Morning

If you buy the argument, here is the implementation.

• Rewrite the front-end question in your patient assessment SOP. Not “is this patient sick,” but “is this patient stable, how do I know, and until when.”

• Publish named instability criteria and train to them. Undefined “stable” is the failure mode we are trying to escape, not the destination.

• Attach a reassessment obligation to every stable call. Interval, plus the finding that would flip it. Put it in the chart.

• Teach the PAT as what it is — a pediatric stability instrument — and stop asking students whether the child is sick. Adopt a stability-based sepsis screen and shock index tracking on the adult side.

• Rewrite the standard radio report so the first word after age and chief complaint is stable or unstable, and push receiving facilities to build their routing against that category.

• Retire “sick or not sick” from documented clinical impressions. If it shows up in a narrative, send it back. It is a habit dressed as a judgment.

• Teach it forward. If you are precepting or writing con-ed, the next generation should not graduate with “does this person look sick to me” as their trained default.

None of this is radical. Every tool named here already exists. What it asks is that the field stop letting a phrase from a paper textbook do work that named physiology does better.

Closing

“Sick or not sick” is not wrong because pattern recognition is worthless. It is wrong because it asks a question with no defined answer, at the one moment when a question with a defined answer is available, and because it can be answered without doing any work at all.

“Stable or unstable — and how do you know, and until when” demands physiology.

It hands off.

It reproduces.

It survives QA.

It scales to the worst day of your career. And it catches the two-year-old with thirty-two respirations and a soft grunt every third breath roughly four hours before “she doesn't look that sick” catches her.

You have to earn stable.

That is the whole argument.

—

Street Medicine is the Black Flag EMS segment for evidence-based challenges to conventional prehospital wisdom. If you think the field is doing something wrong, tell us and we'll fight it out in print — with citations.