And it's also

better for the hospital to free up that bed for someone who's sicker.

So that's a model we've got in production that has been quite successful in flagging it because it used to be more of a referral model and pick it up.

But now that's prompting people and it's not pushing that decision.

It's just saying, here's a patient who might be suitable, have a look at them.

And then the referring team and the hospital and home team can have a look at them.

So we've got a similar model that we're getting quite promising results around falls prediction.

Can you talk us through that scenario because that's often harder to do?