The five-part instruction: trigger, owner, action, proof, escalation

01

Why the usual document does not survive contact

The most useful research I have read on this is not about clinics at all, it is about pharmacies. A 2016 qualitative study in BMJ Open by Thomas, Phipps and Ashcroft looked at how community pharmacy staff actually used their written procedures. Two findings matter here. The first is procedure overload: staff were handed more written material than any person could realistically hold, and one dispenser described the practical outcome as "you never look anything up" after the first read. The second is the gap the authors describe between work as imagined by the people who wrote the document and work as done by the people standing at the counter on a Saturday.

That gap is the whole problem. It is not a compliance problem and it is not a training problem. It is a drafting problem, and it is fixable at the desk.

The PLOS Neglected Tropical Diseases review by Barbé and colleagues, also from 2016, is the best practical guide I know on the drafting side. It sets language rules that sound almost insultingly simple: plain words of no more than three syllables, sentences of no more than twenty-five words, active voice, address the reader directly. It also notes that the standard laboratory template runs to seventeen sections, while the study team in the field actually ran on five. That ratio has stayed with me. Most of what we put in an SOP is there to satisfy the approver, not the operator.

02

The five parts

**Trigger.** What starts this instruction. A time, an event, a state. "At the start of every session." "When the patient arrives for the second review." "If the room temperature reading is outside range." An instruction with no trigger sits in the document waiting for someone to remember it, which is another way of saying it will not happen.

**Owner.** A named role, never a department and never a plural. Not "the clinical team", not "staff", not "we". One role, singular, that a person can point at on a rota. The moment you write a plural you have created a step that three people assume the other two are doing.

**Action.** One verb, one outcome. If you need the word "and" in the middle of the action, you have two instructions and you should split them. This is where most SOPs quietly bloat, because a writer compresses six real actions into one elegant sentence and the reader has to decompress it while wearing gloves.

**Proof.** What is left behind afterwards that shows this happened. A time and initials, a photograph, a batch number, a signed page, a system entry. The proof must be smaller than the action. If recording takes longer than doing, it will be back-filled from memory once the last patient has gone, and a back-filled record is worse than no record because it looks like evidence.

**Escalation.** What the owner does when they cannot complete the action. Who they call, what they say, and what happens to the patient in the meantime. This is the part that gets cut, and it is the part that decides whether a bad day becomes a serious incident.

03

Escalation is a permission, not a phone number

A 2021 qualitative evidence synthesis in BMC Emergency Medicine, looking at why clinicians fail to escalate under early warning protocols, separated the failure into two: not recognising that something has gone wrong, and recognising it but not telling a senior. Among the barriers to the second was fear of a negative response from the senior being called.

You cannot fix that with a phone number. You fix it by writing the escalation line so that it removes the judgement call from the junior person. Not "escalate if concerned" but "call the on-call clinician if any of these three readings appear, regardless of how the patient looks". Then add the sentence that does the real work: no one will be questioned for making this call.

An escalation route that requires courage to use is not a route, it is a trap with a phone number written on it.

04

Proof is the part people cut first, and it is the part that pays

The clearest evidence that recorded checks change outcomes is still the surgical safety checklist study published in the New England Journal of Medicine in 2009 by Haynes and colleagues, across eight hospitals in eight countries. Inpatient complications fell from 11.0 per cent to 7.0 per cent, and deaths from 1.5 per cent to 0.8 per cent, after the checklist was introduced. The checklist did not teach anyone surgery. It made a small number of steps visible, owned and confirmed out loud.

That is what proof does. It converts a private intention into a public fact.

05

A Monday morning pass through your worst SOP

Take the procedure you already suspect is being ignored. Do not rewrite it yet. Do this instead.

1. Number every sentence that tells someone to do something. Ignore everything else. 2. For each numbered sentence, write five letters in the margin: T, O, A, P, E. Cross out the ones the sentence does not answer. 3. Count the crossings. In my experience O and P are missing from more than half of the instructions in any document over two years old. 4. Take the three instructions with the most crossings and fix only those. Not the whole document. 5. Walk the fixed version with the person who runs it, standing where they stand, on a normal busy day. Not in a meeting room. 6. Ask one question at the end: what would you have done differently if I were not here. The answer is your real procedure. 7. Date it, version it, and put the operating steps within arm's reach of the work rather than in a folder on a server.

Seven steps, one morning, one document. You will not fix your manual this way. You will fix the three instructions that were going to hurt someone.

06

What you actually control

You do not control whether your staff are busy. They will be. You do not control the day a regulator changes what has to be recorded, or the day two people call in sick and the rota is thin, or whether a patient behaves the way the procedure assumed. Those arrive on their own schedule and no amount of drafting keeps them out.