The one rule#
nobody argues with#
SLOP2710 · Week 7 · Adaeze and Priya
Six people, a few minutes apart#
sprained wrist · sore throat · cut hand
back pain · chest pain · a rash
Two bays. Ninety seconds to decide the order.
- W category 1
- —
- W category 2
- —
- W category 3
- —
- category 1 waiting over 30 min
- —
- sorted down a day
- —
Every other priority rule gets resisted.#
Why does this one get consent?#
Five classes, and a target each#
| Cat | Meaning | Target |
|---|---|---|
| 1 | immediately life-threatening | immediate |
| 2 | imminently life-threatening | 10 min |
| 3 | potentially life-threatening | 30 min |
| 4 | potentially serious | 60 min |
| 5 | less urgent | 120 min |
The target is what turns a wait
into a failure.
Where it came from#
trier — to sort
Larrey, Napoleon’s surgeon: treat by severity of wound, disregard rank.
And the part nobody discusses#
Mass-casualty tags: immediate · delayed · minor · expectant
Sickest-first and most-lives-saved are different rules.
In a disaster they disagree.
The same machine, elsewhere#
- ambulance dispatch — a category assigned over the phone, before anyone has seen the patient
- P1 to P4 — the same scale with servers instead of patients
- maintenance contracts — line down is a category 1
- safety queues — classified on arrival, at a scale nobody can audit
Priya’s half#
The category is not given. It is produced.#
Three documented failure modes#
Atypical presentation sorted down — women having heart attacks is the canonical study
Pain under-read in people accustomed to not being believed
The bottom category is unbounded — last week’s trolley, worse consequences
And a fourth, belonging to the institution#
The classifier is an input to the performance report.
A borderline case carded down is a borderline case that cannot breach.
15% wrong. How much does it degrade?#
Category 1 mean: 6 → 9 minutes
Category 3 mean: 98 → 91 minutes (better!)
Category 1 waiting over 30 min: 0.7% → 3.0%
The cost of a bad classifier#
does not spread#
Two things to take with you#
“92% accurate” tells you nothing. Ask for the confusion matrix. Ask which cell is the one that kills.
A cleverer rule or a better measurement? A department choosing between them is usually looking at a false choice.