· 10 · Vital Care
project 10, the working out
An ICU ward that watches itself.
Remote ICU monitoring. Continuous NEWS2 scoring, risk bands, automatic escalation. Selected for TIDAC 2024.
- My role
- Everything, solo build
- When
- 2024
- Context
- NSBM research project
- Sector
- Healthcare
Deterioration in an intensive care patient is rarely sudden. It shows up first as a drift across several vital signs at once: a respiratory rate creeping up while blood pressure quietly falls. Any one reading looks survivable on its own. Together they are a warning.
The catch is that spotting the pattern depends on a clinician being at the right chart at the right moment. On a busy ward that is a matter of luck. NEWS2, the National Early Warning Score, already exists to turn those readings into a single number, but it is usually calculated by hand, sporadically, by the person who has least time.
So the question was not whether patients could be scored. It was: what if the score never stopped being calculated, and the ward told you who to walk to next?
The whole idea
sketched this on the back of a lab sheet, then just built it
Vitals in
Respiration, oxygen, blood pressure, pulse, consciousness, temperature
NEWS2 score
Each reading weighted and summed, recalculated on every update
Risk band
Low, medium or high: the ward list re-sorts itself
Somebody is told
Threshold crossed → alert to the doctor or nurse who can act
this last box is the one that matters: a score nobody reads is just more data
Live vital-sign monitoring with alerting against thresholds set per ward, not hard-coded.
Automatic risk classification into low, medium and high using NEWS2 scoring.
Separate views for doctors and nurses: the same ward, different jobs, different screens.
Customisable alerts for both critical and borderline readings, so near-misses are visible too.
Chart and table views of the trend, because some clinicians read shapes and some read numbers.
Notes and discharge handled in the same place, so the record follows the patient out.
What it is built on
React
Redux- RTK Query
Material UI
Tailwind CSS
Node.js
Express
Sequelize
MySQL
AWS
rule-based scoring, not a model: in a ward you have to be able to explain every alert
2024
TIDAC research symposium
Selected and presented at NSBM.
What I learned
That the scoring was the easy half. Deciding who gets told, and how loudly, turned out to be the whole design problem. An alert that fires too often is worse than none, because people learn to dismiss it.
More of it
8 more screens
Admitted patients list with risk bands
Patient records table
Vital sign trends as a chart
Vital sign readings as a table
User management for ward staff
New user registration form
Patient registration, step one
Patient registration, step two