The triage sieve and START: four questions each, and one real disagreement

Run either in TRIAGE, which asks only the questions that can still change the answer, keeps a live count in each category for the casualty line of a METHANE message, and puts a re-check clock on every casualty. It does not decide anything: it records what the person at the casualty decided, and can be overridden with a reason. Nothing is uploaded.

What primary triage is for

Primary triage is not assessment and it is not treatment. It is sorting, done in seconds per casualty, by somebody moving continuously, whose job at that moment is to categorise everybody rather than to help anybody. The only interventions in most sieve protocols are opening an airway and controlling catastrophic bleeding, because both take seconds and both change the category.

The discipline that makes it work is that you keep moving. Stopping to treat the first casualty you reach is the classic failure, and it is a failure precisely because it is the humane instinct.

The triage sieve, as taught in the United Kingdom

1. Are they walking? If yes, they are Priority 3 — delayed — and you move on. This single question sorts most of a scene in the first minute.
2. Are they breathing? If not, open the airway and ask again. Still not breathing: dead. Breathing only after the airway was opened: Priority 1.
3. Respiratory rate. Below 10 or above 29: Priority 1.
4. Perfusion. Capillary refill over 2 seconds, or a pulse over 120: Priority 1. Otherwise Priority 2.

START, as taught in the United States

1. Are they walking? If yes, Minor.
2. Are they breathing? Same airway manoeuvre, same two outcomes.
3. Respiratory rate. Above 30: Immediate.
4. Perfusion. No radial pulse, or capillary refill over 2 seconds: Immediate.
5. Mental status. Cannot follow a simple command: Immediate. Otherwise Delayed.

The disagreement

The two look nearly identical, and on most casualties they are. Two differences are real.

The second is the one that matters. The sieve has no mental-status question at all, so a casualty who is breathing normally, perfusing well and unable to obey a simple instruction — which is a head injury until proven otherwise — comes out as Priority 2 under the sieve and Immediate under START.

Neither system is wrong. They embody different judgements about what a rapid sort should capture, and both are validated in their own context. What is wrong is quietly averaging them, or assuming that because two people used “the triage algorithm” they used the same one. At a scene with mutual aid from services trained differently, this is not theoretical.

The practical answer is to record which system was used, and to keep the categories a person assigned rather than recomputing them later under a different system. A triage is a decision a named person made at a time; retrospectively re-running it under different rules rewrites a record.

The secondary sort

Once casualties reach a clearing station there is time for something more considered. The triage sort scores three observations, each coded nought to four, and adds them:

Twelve is delayed, eleven is urgent, one to ten is immediate, nought is dead. Its value is that it is arithmetic: two people scoring the same casualty get the same answer, which is not true of a clinical impression.

What actually fails at a scene

Not the algorithm. Anybody trained can run either in twenty seconds. Three things fail, and all three are bookkeeping.

Nobody can say how many are in each category

The casualty line of the message that goes up the net determines how many vehicles are dispatched and what is stood up at the receiving hospital. It is frequently a guess, because the person who triaged forty casualties was moving and counting is a different task from sorting. A guess that is wrong by ten is a wrong resource decision made on the strength of it.

The delayed casualty is not looked at again

A major incident does not usually kill the people categorised Immediate; they get attention first, by design. The person it kills is the one who walked away from the vehicle, was categorised Delayed at minute four, sat down at the collection point, and was not reassessed for fifty minutes while deteriorating.

Primary triage is a snapshot of a moving system. Re-triage is not optional, and the only thing that makes it happen reliably is a clock on each casualty that says how long since anybody looked.

The record leaves with the casualty

The tag goes into the back of the vehicle, which is correct — it belongs with the patient. What is left behind is the scene commander with no record of who was there. The board and the tag are two different artefacts and both are needed.

The expectant category

Worth stating plainly because it is widely misunderstood. Expectant means breathing, and beyond the resources currently available. It is not a clinical category, it is a resource one, and the same casualty in a hospital with a full theatre team would be treated. It is declared by a commander on the basis of what is available, never by an algorithm, and it changes when resources change.

None of this is clinical guidance

The algorithms above are the published ones, reproduced. Which applies to you is determined by your training, your service and your jurisdiction, and any tool that runs one is recording a decision a trained person made, not making one.