Early Warning Scores and Predictive AI Are Answering Different Questions

Hospitals that bring AI into patient monitoring often hear the same question from their nurses: we already have an early warning score, so what is the difference?

It is a fair question. Both look at vital signs. Both are meant to help staff spot patients who need more attention, and both produce something that looks like a rating. They answer different questions, and a hospital gets the most from them by knowing which is which.

The question an early warning score answers

An early warning score adds up points from a handful of routine checks. The best-known example is NEWS2, the National Early Warning Score from the Royal College of Physicians in the UK, last updated in December 2017. It scores six measurements: breathing rate, oxygen level, blood pressure, pulse, alertness or new confusion, and temperature. Each one earns points depending on how far it is from normal, and the points are added up.

Its strength is simplicity. Anyone trained in the score can work it out, understand it and act on it, and a higher total triggers a faster response under rules the hospital has agreed.

The question it answers is: how worrying do this patient's readings look right now?

The question predictive AI answers

Predictive AI looks at the same patient differently. Instead of adding points by fixed rules, it reads the patient's own readings over time, including small rises and dips that a points system ignores, and weighs them alongside whatever else is known, such as lab results or medicines. It then estimates how the patient's risk is changing and can explain which changes drove that estimate.

The question it answers is: is this patient heading towards trouble?

Where the two can disagree

Because they ask different questions, the two can point in different directions.

A patient can have a low score while their heart rate and breathing creep up over several hours. Each reading stays in a range that earns few points, yet the overall direction suggests a body under growing strain. The AI, watching that direction, may raise the patient's risk before the score moves.

The reverse happens too. One high reading, perhaps taken while the patient was in pain or had just walked back from the bathroom, can push the score up even though nothing underlying has changed.

In both cases, each tool is doing its job. They are measuring different things.

Why hospitals benefit from both

Early warning scores give a hospital a shared language and a clear plan for who gets called and when. That consistency is valuable, and AI should not replace it.

Predictive AI adds what a points system was never built to show: the direction a patient is moving in, and why. In Lifesigns Nexus, the platform a ward team uses to follow its monitored patients, the two sit side by side on purpose. NEWS2 can be worked out automatically from the recorded readings, provided the information the score needs is available, such as how alert the patient is and whether they are on extra oxygen. That saves staff adding up points by hand and keeps the score up to date. Beside it, the AI in Nexus shows how the patient's risk is shifting and which readings are behind the change.

In practice, that can look like this. On a night shift, a nurse is caring for a patient whose NEWS2 score is low. The score has not reached the level that calls for urgent escalation, and the patient carries on receiving routine checks and care. Meanwhile the AI has moved the patient up a risk category and shows why: over five hours, breathing has risen from 14 to 20 breaths a minute and heart rate from 74 to 89. Every one of those readings still sits in the range that scores no NEWS2 points, yet together they are moving in one direction. The nurse now has extra context and a clear reason to look more closely.

So the score tells the team how worrying a patient looks now, and the AI suggests where things are heading. Both support the team's decisions, and the decisions stay with the team.