Sepsis Triage App

Date of Birth

2 December 2025

Date of Death

6 December 2025

It started because I was learning Python

I wanted something real to build.

Tutorial exercises are all very well, but there are only so many imaginary shopping lists and calculators I can construct before losing the will to live.

So I started thinking about an actual problem.

Carers — including people without clinical training — spend far more time with vulnerable people than healthcare professionals do.

They’re often the people who first notice:

Something isn’t right.

Not necessarily something dramatic.

They’re sleepier.

More confused.

Not eating.

Their temperature is a little odd.

They’re breathing differently.

They’re just… not themselves.

What if an app could help turn that vague observation into the right questions?

Then sepsis sharpened the idea

We discussed a real case where somebody’s wife had noticed that he was “going funny.”

She called 111.

They asked the right questions.

An ambulance was dispatched.

That was essentially what I wanted the app to reproduce.

Not:

Does this person belong to a group we’ve decided is likely to get sepsis?

I didn’t want assumptions about who should be high risk getting between the observer and what was actually happening in front of them.

I wanted:

Pure sepsis spotter. No biases. Just: could it be sepsis?

So I built one

The first version was a small Python triage engine.

Patient observations went in.

Temperature.

Heart rate.

Respiratory rate.

Confusion.

Age.

Changes from normal.

The code assessed the pattern and produced a simple risk level with corresponding advice about whether somebody needed urgent assessment or emergency escalation.

Then it started growing.

Baseline versus current observations.

Atypical presentation.

Hypothermia.

Subtle functional decline in older people.

Explanations for why something had been flagged.

Eventually it might produce reports.

Possibly store observations over time.

This was becoming an actual little application.

I was also learning a great deal of Python.

Including why P1 and p1 are not the same bloody thing.

Then reality started arriving

If this were merely a coding exercise, fine.

But if actual carers were going to use it to make decisions about actual sick people, suddenly there were other questions.

GDPR.

Data storage.

Clinical governance.

Validation.

Liability.

And the sentence no developer of a medical triage app particularly wants to hear:

“But the app said they were fine.”

Still, these were potentially solvable problems.

Then I discovered the rather larger one.

Somebody had already done it

Properly.

There were already serious sepsis-recognition tools developed through organisations including the NHS and UK Sepsis Trust.

Not one woman learning Python on her laptop.

Actual clinical organisations.

Infrastructure.

Data.

IT teams.

Resources.

Validation.

In other words, I’d identified a real problem…

…and then begun reinventing a wheel that already existed in considerably more roadworthy form.

For once, I did the sensible thing.

I stopped.

Cause of death

Prior art.

More specifically:

Other people had already solved the problem properly.

Legacy

A surprisingly useful Python learning exercise.

My first little clinical rule engine.

An introduction to functions, dictionaries, debugging and the important discovery that capital letters matter.

Also a useful lesson:

Before spending six months building something, spend six minutes checking whether the NHS already has one.

Prognosis

Very dead.

And unusually, I’m completely happy about it.

Death was a bit surprised though.