A clinical procedure course is not accepted on how good it looks. It is accepted against a table: the accreditation checklist, broken into rows, the same sheet the examiner marks against when the student is assessed. That makes the wording of a course a compliance question rather than a style question. A course retold in our own words — accurate, readable, better written than the source — is still assessed against wording it never used.
One row, one frame
So the rule we build to is arithmetic rather than editorial: one row of the checklist is one frame of the course, and the frame's description is the cell itself, character for character. Not a paraphrase of the cell, not a tidied version of it. Where the table breaks hand hygiene into eleven rows, the course has eleven frames. Where it puts two actions in one row, the course keeps them in one frame and asks the student for their order only when the row itself states one — assigning an order the table does not give would be invention.
Over two days at the end of August, fourteen procedure courses went through this — auscultation of the heart and of the lungs, blood pressure by mechanical sphygmomanometer, intramuscular injection, catheter care, microscopy, liquid dosing, compounding a solution to prescription and others. Each was compared against its own checklist row by row, and where the frame count disagreed with the table it was brought to the count the table actually has.
What the comparison found
The interesting part is not that texts drifted. It is what drift turns into once a course is a sequence a student performs rather than a page they read.
- A step that had gone missing entirely. In the lung auscultation course the head of the stethoscope was never treated with the sterile wipe that had just been opened — the wipe was opened and discarded. The question immediately before that block asked what the head is treated with.
- A third of a checklist that had never been carried across. The liquid dosing course stood at twenty frames against thirty rows: ten steps were simply absent. Everywhere else the error ran the other way, with one row broken into several frames.
- A clinical error. At the third to fifth intercostal space on the midclavicular line the course had the student listen on both sides. The checklist has the right side only, because the heart occupies the left.
- Listening points collapsed together. Each auscultation point is its own row in the table; in the courses they had been folded into a few long steps. Expanded back — six valve points in the heart course, fifteen in the lungs — each with its own question.
- A detail nobody had written down. One injection step had acquired "the upper outer quadrant of the buttock", which the checklist does not say. Removed.
- Half a step lost. "Remove the cap from the needle" had dropped the second half of its row: placing the cap in class A waste. Restored.
- An order reversed. The needle was withdrawn after pressing the wipe, where the row has it the other way round.
- Questions typed as "click the 3D object" in courses where the object bindings were empty, because no model had been built for them yet — a question whose declared input did not exist.
- Questions carrying two correct answers with no order between them — forty of them across four courses. The two answers duplicated each other in meaning, and the scoring broke on them.
Checked by a script, not by eye
The intramuscular injection course shows why the check has to be mechanical. Its structure was already right: fifty-five frames against fifty-five rows, numbered one to one. The text was not. Of the thirty-seven descriptions submitted for comparison, not one matched its cell word for word — "waste container class A" for "container for medical waste class A", "the preparation" for "the medicinal preparation", a digit where the table spells the number out. Every one of those reads correct to a reviewer and fails a comparison, which is why the comparison is a script and the report states how many descriptions matched.
What happens to a defect in the checklist itself
Checklists have defects of their own: typos, a mood that slips from imperative to infinitive, a stray page number left inside a line of narration. Those are corrected in the table itself against a kept backup, the course is rebuilt from the corrected table, every changed cell is listed in the change, and the corrected table goes back to the person who wrote it — otherwise the next version arrives carrying the same typos and the course quietly disagrees with the sheet it is marked against.
Where the disagreement is not a typo, nothing is changed at all. One checklist is written throughout in the infinitive where every other one uses the imperative; an earlier version of that course had rewritten it for consistency, and the rewrite was reverted — the course now reads differently from its neighbours, because matching the sheet outranks matching each other, and the right place to unify the style is the table, at its author. Where a word was genuinely missing — a unit of measurement, without which the instruction means nothing — the insertion is written down as an insertion rather than filed as a correction. And where twelve draft notes were found sitting in unused columns, they were raised as a question rather than pulled into the course: if those are forgotten steps, the checklist is short, and no course built from it can be complete.