1. Why a written protocol stalls in practice

Because a document describes the plan and a clinic runs on tasks. A protocol in a PDF has to be re-read, re-typed and re-remembered at every visit, and each of those is a place for it to drift. The version in the folder, the version the nurse learnt and the version the newest prescriber is using are usually three different protocols after a year.

What goes wrong What it looks like six months later
The document is the source of truth Two clinicians run different versions and both believe they are current.
Steps live in a person’s memory The programme works until that person is on holiday.
Nothing records which version a patient is on You cannot answer “what were we doing for this patient in March” — which is exactly what somebody eventually asks.
Every visit re-types the same orders A twenty-analyte panel becomes a nineteen-analyte panel and nobody notices for a quarter.
The protocol has no end Patients stay on step three forever because nothing says what finishing looks like.

2. The seven parts a runnable protocol needs

If any of these is missing, the protocol degrades into a document again.

Part What it holds What it prevents
1. Entry criteria Who this is for, and what must be true before starting — baseline labs drawn, consent signed, contraindications reviewed. A patient starting a programme two weeks before the baseline that was supposed to precede it.
2. Steps with timing Each action and when it is due, relative to the start rather than to a calendar date. A programme that cannot be paused or restarted without being rebuilt by hand.
3. Orders attached to steps The panel, the prescription, the administration — as an order set, not as a description of one. Re-typing, and the quiet drift that comes with it.
4. Checkpoints The visit or review where a human decides whether to continue, change or stop. A protocol running on rails past the point where somebody should have looked.
5. What is measured The specific markers this protocol is meant to move, and their cadence. A three-month review with nothing to show, which is how members are lost.
6. An end Completion criteria, or an explicit “continues until reviewed”. Indefinite enrolment nobody ever revisits.
7. A version Which revision this patient started, and when it changed. ⛔ The unanswerable question: what were we doing for this patient in March.

3. Versioning, and the mistake everybody makes once

⛔ Editing a live protocol must never rewrite the history of the patients already on it. This is the single most common protocol-tooling bug, and it is invisible until the day it matters: a clinician improves the protocol in March, and every chart now reads as though the March version was always in place — including the patients who finished in January.

Rule Why
A new revision is a new version, not an overwrite The record of what was done has to survive the improvement of what is done.
Enrolled patients stay on their version until moved deliberately Migrating somebody mid-course is a clinical decision, not a side effect of an edit.
Steps are never deleted from an enrolled patient Removing a step from the template must not delete the record that it happened.
⭐ The chart shows the version So that a year later, the answer to what was being done is a fact rather than an archaeology exercise.

4. What must stay outside the protocol

A protocol that runs without a person somewhere in it is not a protocol, it is a script — and that is a regulatory problem as well as a clinical one.

Stays with a human Why it cannot be automated away
Whether the patient is a candidate at all Entry criteria narrow the field; they do not decide the case.
Every prescribing decision An order is prepared by the system and signed by a prescriber. The protocol never signs.
Each checkpoint The point of a checkpoint is that somebody looks. A checkpoint that auto-advances is a step with extra ceremony.
Acting on an out-of-range result The system flags; the clinician decides. ⛔ It must not hold a result back from review because the protocol did not expect it.
Stopping early Always available, always recorded, and never something the patient has to argue for.

Where a protocol involves controlled substances or compounded products, the prescribing route is its own subject — what transmits electronically and what does not is in which peptides can clinics still prescribe.

5. The half of the protocol the patient sees

A protocol that exists only in the clinic gets half the adherence of one the patient can see. The clinic’s version is orders and checkpoints. The patient’s version is what to do this week, what is coming, and what is supposed to change.

The patient should see The effect
Where they are in the course Turns an open-ended prescription into a programme with a shape.
What is due this week Removes the most common adherence failure, which is not refusal but forgetting.
The markers being tracked, and the last value Gives the three-month conversation something to be about.
What the next checkpoint decides Sets the expectation that a review is a decision, not a formality.

Which numbers are worth showing — and how often — is in outcomes reporting for longevity clinics, and the panel behind them in which age biomarkers a longevity programme runs.

6. Frequently asked questions

What is a protocol engine, exactly?
Software that stores a protocol as steps with timing, orders and checkpoints, enrols a patient onto a version of it, and then tracks where each patient is — rather than storing a description of the plan and leaving execution to memory. The practical test is whether the system can tell you, today, which patients are on step three and which are overdue.

Does a protocol engine prescribe?
No, and it must not. It prepares the orders a protocol calls for; a prescriber signs them. Every checkpoint is a human decision, and stopping early is always available. A system that advances a medication step without a clinician is not running a protocol.

What happens to patients already enrolled when we change a protocol?
They stay on the version they started unless somebody moves them deliberately. An edit must never rewrite the history of a patient who already finished — the chart has to keep showing what was actually done, and which version it came from.

How detailed should a protocol be?
Detailed enough that a competent clinician who has never seen it can run it, and no more. Over-specification creates protocols nobody follows exactly, which is worse than a looser one everybody does — because then the record no longer matches the medicine.

Do patients need to see the protocol?
They need to see their half of it: where they are, what is due, what is being measured and what the next review decides. Adherence failures are mostly forgetting rather than refusing, and a visible plan fixes far more of them than a reminder does.