Missing and fixing pull against each other
A missed promise makes nearby ones more likely to be missed too. A confirmed fix pushes the other way. The model treats those as two opposing forces rather than one flat score.
OLRHM v2
pressure, not snapshots
Working out what is about to go wrong, before it does.
Open-Loop Repair Hazard Model - OLRHM v2
Most risk scores take a snapshot: they look at a patient once and give a number. That misses the thing that actually matters, which is movement - promises pile up, one slipping makes the next more likely, and fixing one takes the pressure off the rest.
Our model watches that movement instead. It tracks pressure building around each open promise, and how much of it spills onto the others. The result is an estimate of how likely it is that something gets missed, and roughly when.
λᴘ(t)=Σᵢ ∈ O(t) λʙᵢ(t | Hₜ)
P(T ≤ τ)=1 − exp(−∫₀^τ λᴘ(u) du)
In words: add up the pressure on every promise still open for this patient, and that gives the chance something gets missed by a given moment.
Each promise carries its own pressure, based on what has already happened to it, how long its window is, whether anyone has tried to fix it, what it depends on, and anything hitting the whole system at once. The final number comes out of that maths - it is not a score someone capped at a convenient place.
What is new
four bets
Four things we think are new
Promises are connected, not a list
Some things can only happen after others. We keep that order, and separate a patient who is personally fragile from a hospital that is having a bad week.
A claim narrow enough to be tested
If you fix things faster across the board, fewer get missed. That sounds obvious, but we have stated it precisely enough that it could be shown to be false.
Knowing when to stop gathering data
Before spending money proving something, we ask whether the next piece of evidence is worth more than it costs to collect.
The claim
the idea, not the maths
The new part is the idea, not the maths.
The mathematical tools here already exist and are widely used. What we think is new is treating a promise as something with both a missing force and a fixing force, mapping how those promises depend on each other, and building the software so the whole thing can actually be measured. Independent statisticians and patent counsel still have to agree, and neither has reviewed it yet.
What we hold back
ask us
What is not on this page
The above is the simplified version. The full workings, how the pieces depend on each other, the urgency-routing part, and how we plan to test it are covered in the investor briefing rather than published here.